dissenting.
After reviewing the record1 and decision of the district court, I am left with a *596“definite and firm conviction that the trial court committed a clear error of judgment” in awarding fees and costs to Peoplemark. Logan v. Dayton Hudson Corp., 865 F.2d 789, 790 (6th Cir.1989). A district court abuses its discretion when it “applies an erroneous legal standard, misapplies the proper legal standard, or relies on clearly erroneous facts.” Id.
In Christiansburg Garment Co. v. Commission, 434 U.S. 412, 422, 98 S.Ct. 694, 54 L.Ed.2d 648 (1978), the Supreme Court held, a court should not assess attorney’s fees against a plaintiff unless the court finds that his or her claim was frivolous, unreasonable, or groundless, or that the plaintiff continued to litigate after it clearly became so. After a thorough review of the record, I find it was clearly erroneous for the trial court to conclude both that the Equal Employment Opportunity Commission’s (EEOC) case was meritless from the outset and that the EEOC should have known that its case was meritless as of October 1, 2009. I also find clearly erroneous, the court’s conclusion that the EEOC’s prosecution of the case was so unreasonable as to justify an award of fees. For these reasons, addressed in more detail below, I respectfully dissent.
To fairly assess both the reasonableness of the EEOC’s prosecution of the case and the merit of the case itself, it is necessary to lay out the proceedings in the lower court in considerable detail. One cannot fairly assess the propriety of a fee award in this case without a close reading of the documents filed with the court. Only through a thorough review of the documents can one understand the variety and content of the employment application materials that Peoplemark provided to the EEOC during discovery. On must also thoroughly review the documents to understand the EEOC’s considerable task of processing and analyzing Peoplemark’s employment application documents. As such, and at the risk of retracing a few of the steps the majority discusses, I begin with a rather lengthy description of the presuit investigation and lower court proceedings.
I. Factual Background
A. Investigation
Peoplemark is a temporary employment agency, which refers individuals for placement with client-employers. The company primarily provides workers for light industrial jobs, such as packaging, warehousing, assembly, and some machine operations. It also assigns people to some receptionist and clerical positions. Peoplemark currently has four offices: Grand Rapids, Michigan, Memphis, Tennessee, Owensboro, Kentucky, and Orlando, Florida. Until 2006, it also had an additional office in Livonia, Michigan.
1. Charge and Statement of Position
In May, 2005, Sherri Scott filed a charge of discrimination against Peoplemark with the EEOC. Scott alleged that after she applied at the Peoplemark office in Grand Rapids, Michigan, Peoplemark’s interviewer told her she would not be hired because she had a felony conviction. Scott alleged that Peoplemark rejected her application because of her race and felony conviction.
From 2005 to 2007 the EEOC investigated the charge.
At the outset of the investigation, Peoplemark provided the EEOC with a written statement of position conceding that it uniformly rejects felon candidates. Peoplemark confirmed that its interviewer told Scott that she could not continue in the application process because she had *597prior convictions. Peoplemark also informed the EEOC that it handled similarly all candidates with convictions. People-mark argued that its client’s refusal to hire felons for “security, safety, and other legitimate business reasons” provided the company with an appropriate business justification for rejecting felon candidates.
a. Letters from Chief Counsel
Throughout the EEOC’s investigation, Judd Osten (Osten), Vice President and Chief Counsel for Peoplemark, continued to assert that Peoplemark did not hire felons. Osten sent letters to the EEOC on October 31, 2005 as well as June 12 and December 1, 2006, expressly stating that Peoplemark rejects all felon applicants.
In his first letter, Osten provided information specific to the Grand Rapids office. He stated that the office uses a bifurcated application process to screen out felon applicants. It did so, he stated, because Peoplemark’s client-employers refuse to employ felons.
The first step in the process involves an information sheet, on which an applicant indicates only his or her name and whether he or she has an existing or pending felony charge or conviction. After the applicant fills out the information sheet, a staff member searches the public record to confirm the accuracy of the applicant’s disclosures. Applicants without a felony conviction proceed to the second step — filling out an employment application.
In his second letter, Osten stated that Peoplemark, Inc., never hires felons. Os-ten did not state explicitly whether he was referring to all Peoplemark offices:
It is the practice of Respondent not to accept applications for placement in temporary assignments of individuals with felony conviction records because they pose too great a risk to Respondent and its clients ... Respondent’s clients are generally aware that it does criminal background checks on all of its candidates for temporary placement, but [Respondent] has made no independent inquiry of its clients as to which, if any, would accept personnel with a felony conviction record. It has received no request from any client for the assignment of an employee with a felony conviction record. The following [four] clients have orally advised Respondent that they do not wish to have anyone with a felony conviction assigned to their accounts: Pridgeon & Clay; Zondervan; Team Industries; and Adac Automotive.
In his third and final letter during the investigation, Osten stated explicitly that Peoplemark had a company-wide practice of rejecting felon candidates: Peoplemark will “stipulate that the practice of screening out applicants with felony conviction records, which is at the core of this matter, is used in all of our offices throughout the country.”2
Based on Scott’s charge, Osten’s statements, and 18,000 cover sheets3 covering *598one year of applicants in the Grand Rapids office, the EEOC attempted to conciliate the dispute.4 The EEOC was ultimately unable to conciliate the matter. It issued a cause letter of determination on September 10, 2007.
B. Filing Suit and Initial Case Management Order Deadlines
On September 29, 2008, the EEOC filed suit in the United States District Court for the Western District of Michigan on behalf of Scott and similarly situated African Americans.
The EEOC alleged:
Since at least May 2005, Defendant Employer has engaged in unlawful employment practices at all of its facilities in violation of Section 703(a) of Title VII, 42 U.S.C. § 2000e-2(a). The Defendant’s unlawful employment practices include maintaining a policy which prohibits the hiring of any person with a criminal record. Such policy has had and continues to have a disparate impact on African American applicants.
The effect of the practices complained of in paragraph 7, above, has been to deprive Scott and similarly situated African Americans of equal employment opportunities and otherwise adversely affect their status as applicants for employment.
On January 9, 2009, a Magistrate Judge held a case management conference. The parties set deadlines:
1) identify experts: 6/30/2009
2) produce expert reports: 8/31/2009
3) close of discovery: 12/31/2009
4) bench trial: 7/16/20105
C. Discovery Disputes and Electronic Document Production
Had discovery proceeded without incident and attended delays, the original deadlines would have given the EEOC seven and a half months to gather information and produce its expert reports.
Unfortunately, discovery was not uneventful. Two discovery disputes consumed approximately five of the seven and a half months the EEOC expected to have for gathering factual information and preparing its expert reports. Various discovery related motions, conferences, and hearings occupied the parties from February to early August, 2009.
1. Discovery Dispute # 1: Peoplemark’s Demand for a Complete Class List
The first discovery dispute arose because the EEOC could not provide People-mark with a complete class list at the outset of discovery. On February 4, 2009, Peoplemark filed a single interrogatory asking for a class list, including contact *599information, of all black candidates that had been denied placement because of their felony convictions. The EEOC responded that it previously provided Peoplemark with the name and contact information of Sherri Scott and it was too early to identify the rest of the class.
At the time of Peoplemark’s interrogatory, the only company records the EEOC possessed were the 18,000 initial information sheets applicants had completed in the Grand Rapids office during a one year period. The sheets contained only the applicants’ names and whether they disclosed a felony conviction. The sheets did not contain the race of applicants, or any other personal identification information from which the EEOC could identify race with any reasonable degree of accuracy. The sheets also did not indicate whether Peoplemark had actually placed any of the applicants who had noted a felony conviction on the information sheets.
a. Peoplemark’s Motion to Compel and the EEOC’s Response
Unsatisfied with the EEOC’s answer to its interrogatory, Peoplemark filed a motion to compel. The Magistrate Judge granted the motion and sanctioned the EEOC for failing to produce a class list in response to Peoplemark’s interrogatory. The judge ordered the EEOC to produce a class list by May 1, 2009.
In response to the motion to compel and the Magistrate Judge’s order, the EEOC explained that it was impossible to produce a complete or accurate class list without all of Peoplemark’s application records for the relevant time period. The EEOC had no record of which applicants Peoplemark had placed, nor did it have the applicants’ personal identification information — it could not identify accurately the race of applicants without their personal information.
In a word, the Magistrate Judge had, early on in the case, placed the EEOC in a classic Catch-22 situation — from which the EEOC was never able to extricate itself.6
To resolve the discovery dispute, the EEOC provided its best guess as to a list of 286 persons with individual claims for relief. On providing the list of potential victims, the EEOC reiterated its position, stated previously in its answer to People-mark’s interrogatory, that this was only a potential list of some possible victims, not a class list. The EEOC stated, inter alia:
Although the investigation uncovered a number of applicants with criminal records believed to be African American, the Commission has not yet determined the magnitude of the class and is in the process of gathering additional information from Defendant in discovery so that it may identify the definitive class for purposes of this litigation.
On March 5, 2009, the EEOC timely served its Answer to Peoplemark’s first Interrogatory:
INTERROGATORY NO. 1: Identify by name and address each and every individual whom you maintain has been discriminated against by the alleged employment policy/practice which is the subject of this lawsuit.
ANSWER: Plaintiff previously identified Sherri Scott, 718 Broadway, Apt. 2G, Niles, MI 49120. To the extent this interrogatory seeks the identities of individuals other than Ms. Scott, this request is pre-mature because discovery *600has recently commenced and is on-going. Plaintiff will develop the class of victims during discovery and supplement this response at a later time.
[I]n an effort to resolve this discovery dispute, the EEOC is willing to provide you with information regarding applicants believed to be African American and who had disclosed criminal convictions on their application with Peoplemark. Based on Defendant’s representation that no individual with a felony conviction was hired, we believe that these individuals may form part of the class in the instant case. However, please be advised that the EEOC makes no representation that this list is all inclusive as it is based upon the limited scope of information Peoplemark provided during the administrative investigation. Consequently, we anticipate that individuals may be removed from and added to the list as discovery progresses. Until substantial discovery is completed, the magnitude of the class cannot be determined.
Peoplemark argues that because it produced a large volume of employment applications and other documents, during what it describes as an “extensive” administrative investigation, the Commission should have established the class during its investigation ... Defendant is incorrect in its argument for several reasons. First, the administrative investigation was not as over broad as Defendant attempts to suggest. In fact, the Commission made several compromises regarding the scope of its documentary information requested by limiting the time frame to March 2004 through November 20067, and seeking only the first page of each employment application instead of all documents related to the application process. Second, there was nothing in the documents Peoplemark produced which revealed the race of applicants. Consequently, and as a required part of its investigation, the Commission took further investigative measures to identify applicants by race to the extent possible. In light of People-mark’s no-felony-conviction policy, all applicants with felony convictions identified during the investigation as African American were treated as the putative group of victims harmed by Peoplemark’s policy. Accordingly, the Commission attempted to conciliate this matter based on this putative group of victims who it believed were harmed by Defendant’s policy. Regardless, now that the parties are in litigation, the Commission is required to define the class with more precision and has served Peoplemark with written discovery requests (which include requests for all previously incomplete application materials and all applicant information to the present) in an effort to obtain information needed to determine the scope and identities of the class.
(Emphasis added).
2. Discovery Dispute # 2: The EEOC’s First Set of Interrogatories and Requests for Document Production
Prior to sending its own interrogatories on March 4, 2009, the EEOC underwent a comprehensive drafting process with its electronic document experts. While still working to resolve the first discovery dispute the EEOC served its first set of fifteen interrogatories and document requests on Peoplemark. Unfortunately, the format of the EEOC’s interrogatories gave *601rise to a second time consuming discovery dispute.8
For approximately eighty days, People-mark refused to answer the EEOC’s interrogatories. Peoplemark made a numerosity argument: subparts of questions counted as additional questions, and thus, the EEOC had really served seventy interrogatories, instead of the limit of twenty-five stated in the case management order. After repeated attempts to obtain answers to its interrogatories, the EEOC contacted the case manager to clarify the local rule on subparts. The case manager told the EEOC that subparts were allowed and Peoplemark should answer the interrogatories.9
On May 21, 2009, Peoplemark finally agreed to answer all the interrogatories. However, Peoplemark still refused to produce the requested employment applications and information sheets.
On May 22, 2009, the EEOC filed a motion to compel production of the applications and information sheets.
On June 2, 2009, the EEOC received a supplemental correspondence from People-mark stating that it would make the documents available for copying and inspection at a mutually agreeable time and-place at its offices in Michigan, Tennessee, Florida, and Kentucky. In response, the EEOC withdrew its motion to compel
Well into July, Peoplemark had still failed to provide specific dates when the documents would actually be available to the EEOC at the company’s various locations.
Nearly ninety days passed between the time the EEOC served the document requests and when Peoplemark agreed to produce the documents.
3. The EEOC’s Request for Electronic Document Production
After a lengthy fight to gather information about what types of electronic records Peoplemark maintained, the EEOC asked Peoplemark to produce its e-documents.
On June 22, 2009, Peoplemark served its answers to the EEOC’s request for production of electronic applicant data. Peoplemark agreed to provide the data at a time that was mutually convenient to both parties. In mid-July Peoplemark produced its first version of an e-database with limited applicant information. The database contained applicant placement in*602formation for three of the five offices, but, according to the EEOC, the agency was unable to access much of the information because Peoplemark failed to provide the agency with necessary passcodes. The EEOC alleges that Peoplemark did not provide a fully searchable database until mid-September 2009.
D. Motions to Extend/Amend Case Management Order and People-mark’s Document Production
1. First Motion to Extend:
June 29, 2009
On June 16, 2009, Peoplemark’s attorney phoned the EEOC to propose a two month extension of the deadline to name experts. The extensive discovery disputes had consumed the first five months of the discovery period. Under the original case management order the EEOC had only two months remaining to collect, copy, scan, streamline, code, statistically analyze, and report on four years of applicant data located in four different states.
The EEOC concurred with People-mark’s request and added that an extension for producing expert reports was also necessary. The EEOC stated that it would need considerable time to obtain the documents, process the data, and produce an expert analysis.
On June 23, 2009, the EEOC followed up with Peoplemark regarding extending the deadlines in the case management order. The agency provided Peoplemark with its proposed amended deadlines. Over the next week the EEOC tried unsuccessfully to reach Peoplemark’s attorney numerous times.
On June 29, 2009, one day before the initial deadline for disclosing experts, Peoplemark’s attorney abruptly reversed course and informed the EEOC that the company was no longer willing to stipulate to revised deadlines which Peoplemark had initially proposed. The same day, immediately after learning that People-mark refused to proceed as planned, the EEOC filed its first request to amend the case management order.
In its first motion to amend the case management order, the EEOC explained that two discovery disputes had consumed more than five months of discovery time, and that the EEOC had tried unsuccessfully to reach an agreement for amended case management deadlines with People-mark. The EEOC proposed new deadlines:
1) identify experts: 9/4/2009
2) produce expert reports: 2/5/2010
3) close of discovery: 6/30/2010
The EEOC’s proposed amended case management order allowed the agency a little over seven months to: 1) copy and scan information at Peoplemark’s offices in four different states; 2) send the raw applications, information sheets, and other employee documents to a document processor to create an e-database; and 4) send the e-database to its statistical expert to code, run statistical formulas, and write an expert analysis.10
*603On July 6, 2009, Peoplemark filed a motion to dismiss, which the court denied the next day.
On July 16, 2009, the Court granted a thirty day extension for both the identification of experts and the production of expert reports. The court also allowed the parties an additional fourteen days of discovery. This de minimis extension gave the EEOC only two months to collect, copy, scan, process, code, and analyze the applicant information contained in People-mark’s records.
The EEOC immediately informed the court that it would be unable to meet the first amended case management order. The court responded that the parties were free to file motions and be heard by the court on the issue of further extensions.
2. Second/Renewed Motion to Extend: July 23, 2009
Accordingly, on July 23, 2009, the EEOC filed a renewed motion requesting the same extended deadlines it had asked for in its June 29, motion to amend the case management order.11
In addition to reiterating the substantial amount of time the discovery disputes had consumed, the EEOC explained why the agency had failed to start copying and scanning the paper records in People-mark’s offices in Michigan, Tennessee, Florida, and Kentucky.
The EEOC stated that although, on June 2, Peoplemark had agreed to make the documents available at a mutually convenient time, Peoplemark had not yet provided a formal date when the EEOC would be allowed access to the documents at the four locations.
In addition, the EEOC explained that it was required by federal law to use a bidding process for hiring outside vendors and expert witnesses. The EEOC described the numerous steps that it had taken in the month and a half since Peoplemark agreed to make the documents available: 1) the EEOC consulted, as required, with the EEOC’s Research and Analytical Services Division, which determines the number of vendors and experts likely to be needed for a case; 2) the EEOC attorneys initiated, a bidding process, as required, to hire a document processor and two expert witnesses; 3) the EEOC attorneys obtained approval of the document processor contract on July 21, 2009; and 4) the EEOC attorneys recommended two experts for formal approval.
The EEOC acknowledged that it was still awaiting formal approval of the expert contracts and did not expect them to be approved in the week remaining before expert identification deadline. The EEOC further stated that the court’s suggestion in the July 16, 2009, status conference that the EEOC use its small in house research staff to complete the data analysis was not feasible.
Although the EEOC’s experts had not been formally approved within the court’s expert identification deadline, the EEOC attorneys were ultimately able to obtain exceptional approval from the government to identify two experts on July 31, 2009, to meet the court’s deadline.
*604The agency identified expert sociologist, Dr. Devah Pager, and expert labor economist/statistician, Dr Janice Madden.
a. Hearing on Second/Renewed Motion to Extend
On August 6, 2009, after the EEOC identified its experts, the court held a hearing on the EEOC’s second/renewed motion to amend the case management order. The transcript of the hearing reveals the district court’s misunderstanding of the nature and extent of the EEOC’s: 1) task of preparing an expert statistical report, and 2) presuit investigation process-the statistical proof the EEOC was required to gather before filing suit. I discuss the hearing at length because I believe the lower court’s flawed understanding is highly relevant to both the merits inquiry and procedural abuse analysis that the lower court performed before assessing fees.
The court began by stating that it applied a “good cause” standard to granting motions to amend. Under this standard, the court required a party to show that it acted diligently but was still unable to meet the established deadlines. The court expressed its concern that the EEOC acted dilatorily when it failed immediately, or at least more quickly, to obtain its experts.
The EEOC responded that it had acted diligently, but it was still unable to produce its expert reports on time. The EEOC stated that it began the process of researching, writing, and revising expert contracts in December 2008 or January 2009 at the latest (less than one month after filing the case). The agency also explained that federal regulations required it to follow a bidding process for hiring vendors and experts, and the EEOC attorneys could not legally circumvent the internal agency protocols.
In response to the EEOC’s explanation of its efforts the court stated that “protocols within the agency can’t govern the civil procedure process, and [the agency had] to have good faith at the start.”
The court went on to express concern that the EEOC waited sixty days to serve its first set of interrogatories, to which the EEOC responded that during that time the agency was working diligently with its electronic document experts to craft its interrogatories for Peoplemark. The agency stated that it needed to see what Peoplemark produced in its Rule 26 disclosures before it could begin drafting its first set of interrogatories.
The EEOC explained that after receiving Peoplemark’s Rule 26 disclosures, the agency’s attorneys and experts underwent an extensive joint drafting process. This was necessary, according to the EEOC, because the first set of interrogatories sought information regarding electronic storage systems and formats of electronically stored documents. For this reason, the EEOC attorneys could not simply draft the interrogatories themselves — they were not computer experts.
The EEOC also clarified for the court that it could not simply modify an existing interrogatory template. Rather, based on documents provided in Peoplemark’s Rule 26 disclosures, the EEOC needed to craft interrogatories specific to the electronic programs they believed were in use at Peoplemark’s offices.
The court replied that the EEOC could have limited the number of Peoplemark’s offices that it included in the suit if it was concerned about meeting deadlines. The court also stated that it had been generous with scheduling and never had such issues with a government litigant.
In the court’s opinion, “[P]art of coming to court is having the court put reasonable limits on and see which parties make it through that process and which parties *605don’t, and that’s part of the way we drive to a result.”
The EEOC responded that although the government’s internal protocols contributed to some delay in the EEOC’s ability to physically go out and start copying and scanning documents, the government’s protocols were not the sole or even the main reason the EEOC was unable to meet the court’s deadlines.
The EEOC reiterated the discovery disputes laid out in both its first and second motions to amend the case management order. The EEOC argued that if People-mark had not waited over two months to answer the EEOC’s first set of interrogatories and over ninety days to agree to produce documents, then, despite delays caused by government regulations, the EEOC likely would have been able to request and secure someone to gather the documents much sooner. Furthermore, during the time the EEOC was processing those documents it could have finalized the contract for its statistical expert who, up to that point in the process, would have been totally unnecessary.
The court’s response was:
Well, the documents part addresses whether you make the expert report deadline, and we don’t know that. I mean, when we talked in July, you didn’t think you’d make the [expert identification] disclosure deadline, and you did ... And if you have a report deadline that’s still, what, 60 days away, even though that seems impossible now, maybe you’ll make that too ... Or maybe you’ll come close or make it in part. Or maybe having the deadline will mean that you’d get a lot closer to it than you would if the deadline were extended. I mean, the reality of the deadline is it marshals resources that we didn’t even know we had. That’s just what a deadline does.
The court proceeded to reprimand the EEOC for not filing a motion to compel after Peoplemark stalled in setting a date for production. The EEOC responded that it was hesitant given the court’s resolution of its first motion to compel and the court’s earlier award of sanctions.
Throughout the remainder of the hearing, the court continued to speculate, in the face of Dr. Madden’s affidavit, that it might be possible for the EEOC to copy, scan, create an e-database, code, analyze, and produce a report in the next sixty days. The EEOC firmly maintained that it was absolutely impossible for it to do so. The EEOC stated that it was able to identify experts only because it obtained exceptional approval, but it was physically impossible to do the work necessary to produce the expert reports in sixty days.
In response to the EEOC’s firm statement that it was absolutely impossible to complete all the steps required to produce an expert report in sixty days, the court retorted, “I’m suspecting in September [the deadline for producing expert reports] you may have another vision of how powerful and effective you really are.”
After the lengthy dialogue between the court and EEOC, Peoplemark announced that it had asked all of its offices to send the documents requested by the EEOC to Memphis. Peoplemark stated that it had already scanned all of the documents from the Grand Rapids office. The company stated that it had put the information onto two discs and the day before sent the discs to the EEOC. Peoplemark stated that it would do the same for the remaining offices.
The EEOC responded with surprise to Peoplemark’s decision immediately to scan and produce all of the documents. The agency stated that the company’s offer came as quite a shock in light of its previous refusal to provide the documents and *606its reluctant agreement, after a ninety day wait, to provide the documents only after the court issued a protective order. The EEOC told the court that, assuming the discs were complete, Peoplemark’s offer to copy and scan the paper records would reduce by two weeks the time the EEOC needed to produce its expert reports.
The court closed the hearing by restating that it still believed the EEOC could “marshal the resources” necessary to meet the deadline, but if the EEOC was unable to do so, then it could approach the court again in the next sixty days to discuss the matter further.
On August 7, 2009, the court formally denied the EEOC’s second/renewed motion to amend the case management order.
I. Peoplemark’s Document Production After the Hearing on the Second Motion to Amend: 8/11-9/17/2009
On August 11, 13, 14, 17, 18-21, 25, and September 17, 2009, the EEOC received discs containing full paper application information. The EEOC made forensic copies of the information and overnight mailed the discs to its document processor.
Thus, for the first time in the litigation, the EEOC had in its hands the data it needed to compile its statistical analysis, prepare its experts reports, and ascertain with a significant degree of certainty those persons who would be class members. Despite this fact, the agency found itself between the Scylla of the work it needed to do an Charybdis of the court’s inflexible view towards its deadline. The court refused to acknowledge the fact of the long-delayed, but essential document production. Moreover, the court assumed, based on its subjective speculation that the agency was exaggerating the time needed for its experts to do their work.
The volume of the disclosure, as finally received, was far more extensive than the EEOC had anticipated: namely, 178,000 pieces of paper, rather than the approximately 50,000 pages the agency had expected to receive — an increase of more than 250%.
The fact that Peoplemark produced 250% more documents than the EEOC had originally anticipated added considerably to the amount of time the EEOC’s data processor needed to pull information from the applications and enter the information into a standardized format. Quite simply, more applications required more processing time. The EEOC’s document processor stated in its affidavit in support of the third renewed motion to extend that each additional 1,000 pages would likely require an additional one business day to process.
The EEOC’s document processor also informed the court that there had been a noticeable deterioration in data CDs Peoplemark produced. At first, Peoplemark provided the EEOC with neatly labeled documents, with images and corresponding text, which was presented in a logical, well-named hierarchical folder schema.
As time went on, Peoplemark no longer organized or labeled the documents according to any ascertainable presentation schema. Peoplemark failed to put the electronic documents in folders. People-mark’s reduced organization forced the document processor to have to wade through non-eritieal documents such as 1-9 forms and drug test consent forms to find the relevant applicant information that it needed to complete an e-database for the EEOC’s experts.
4. Third Motion to Amend: Request to Extend Deadline for Dr. Madden’s Statistical Report: 9/24/2009
On September 24, 2009, the Commission filed its final motion to amend the case management order. It requested an extension for filing Dr. Madden’s statistical *607analysis report. The EEOC stated that it would produce Dr. Devah Pager’s report under the previously established deadlines with the caveat that it be allowed to supplement that report once the information database was complete.
Dr. Pager’s report detailed the disparate conviction rates between blacks and whites in society generally and in the communities where Peoplemark has offices. The EEOC was able to produce Dr. Pager’s report because it did not necessarily require an analysis of Peoplemark’s employment records.
In its third motion the EEOC again requested that the court give the agency until February 2010 to produce Dr. Madden’s report.
The EEOC outlined its efforts to ship all discs to its document processor and its document processor’s efforts to enter the information from the raw application materials into a uniform e-database that the EEOC’s experts could code, analyze, and report.
The EEOC attached affidavits from both its document processor and Dr. Madden. The document processor outlined its progress thus far, and stated that in its professional opinion, the e-database could not be provided to Dr. Madden’s team for coding and analysis before November 4, 2009, (thirty business days from the filing of the motion).
Dr. Madden stated in her affidavit that after receiving the e-database her team would need at least three months to code, analyze, and report on the data from the 45,000 applicants the document producer identified.
Her affidavit in support of the motion explained in detail why she needed that much time to provide a report:
Once the electronic or computerized data have been provided to me, a database which is currently being prepared by a litigation support vendor ..., I must code the data into formats that a computer and a statistical program will recognize ... I must review the data entered by the vendor for education ... prior work experience, criminal records, shift desired, etc. This process of assigning a common, computer readable value for each characteristic for each of 45,000 applicants requires at least eight weeks.
Once these vast arrays of data are coded into a standard format ... I apply widely accepted statistical formulas to calculate whether there are differentials in the likelihood of hire for those with criminal records ... The application of these programs to the coded data requires at least three weeks.
Finally, when the statistical results are available, I prepare a written report summarizing the approaches taken and the findings. The writing of the report requires my unassisted time, although the extensive checking of the report against the computer programs is performed by assistants ... The writing of the report requires about two weeks, one of which must occur after the computer work has been completed.
The EEOC has advised me that the completed electronic database will be delivered on November 4, 2009. Therefore, I believe that my expert report can be completed 12 weeks thereafter, or by February 11, 2009 (given Econsult’s longstanding policy of closing during Christmas week).
On September 30, 2009, the EEOC timely produced Dr. Pager’s report.
On October 16, 2006, the Magistrate Judge declared that the EEOC’s third motion, including Dr. Madden’s affidavit, failed to make clear “how plaintiff expects to use Dr. Madden’s report to prove liability.” The court expressed concern that
*608“Dr. Madden is not routinely coding in an applicant’s race for purposes of her analysis.” Further, the court stated:
If the report is simply being used to prove that statistically people with felony convictions are less likely to be hired than people without felony convictions, defendants may not even dispute this point ... The issue in this case ... is whether the defendant categorically denies hire to any individual with a criminal record ... If [it was true that Madden only seeks to show that people with convictions are hired less frequently], a continuance based on the need for Dr. Madden’s report would not be necessary.
Instead of denying outright the EEOC’s request as it had done previously, the court asked for a supplemental report providing further explanation of why Dr. Madden’s report was absolutely essential to proving liability and why she needed all the fields of information to write her report.
5. EEOC’s Supplemental Brief in Support of Extending the Deadline to Produce Dr. Madden’s Report
On October 23, 2009, the EEOC filed its supplemental memorandum, which included a second, more detailed affidavit from Dr. Madden. The EEOC stated that Dr. Madden’s statistical analysis was absolutely necessary to prove the causal element of its prima facie case.
First, the EEOC explained that, although its pleadings asserted that People-mark maintained a categorical no-felon policy existed at Peoplemark, discovery had showed that Peoplemark had, in fact, placed a small number of applicants with prior felony convictions. Accordingly, the EEOC told the court that it did not plan to allege a categorical policy as the first element of it prima facie case. Instead, the EEOC planned to allege a policy of considering felony convictions in the pre-application and application processes used at Peoplemark’s offices.
The EEOC also attempted to explain the substantiative employment discrimination law to the court. It told the court that it was not required to assert the same “categorical” policy alleged in its pleadings in its prima facie case. The EEOC stated to the court that it could prevail on its disparate impact claim as long as it showed Peoplemark had a policy that disparately impacted African Americans.
Second, the EEOC stated that Dr. Madden needed to review the “applicant flow data and current work force data” in order “to determine whether the manner in which convictions are used at each facility has a disparate impact on African American applicants for those positions.”
Dr. Madden provided an extremely thorough explanation of the expert analysis process. The most crucial excerpts from her second affidavit stated:
[T]wo different analyses of the application and personnel information provided by Peoplemark are required:
a. An analysis of whether applicants who report conviction records on their applications were less likely to be hired than those who did not report criminal convictions.
b. An analysis of whether African American applicants were more likely to report criminal convictions than were applicants from other ethnic racial groups.
For [the first analysis] ... I must take the information that applicants hand wrote on their applications, which includes the dates of any convictions and a description of the criminal charges, and code the information into a machine readable format. While the coding of dates into years is straightforward, the *609coding of charges is more complicated ... I will use [the electronic database provided by the document processor], but I must first sort and group these entries in order to characterize convictions by their severity (felony vs. misdemeanor) and type (i.e., drugs, violent, theft, financial, etc.).
In order to be sure that any differences in hiring between those with convictions and those with no convictions are associated with the convictions ... it is necessary to analyze the likelihood of hire among ... applicants with otherwise equivalent education and job histories ... I must sort and group [education and job history information collected by Peoplemark in various ways] in order to characterize them in a way that a computer program can analyze.
For education, I must determine the level. Applicants entered their education on the application in a variety of ways. Some wrote the name of the school and “yes” in the area where they were asked to enter grade level or degree completed. Others entered a degree and no school ... Others only entered a school name ... If convicts were less likely to have completed high school or attended college then an analysis that failed to consider the effects of education would inaccurately attribute a hiring differential to convictions.
For work experience, I must determine the type and the amount. Applicants identify their last three employers, including the dates of employment, the company name, and the duties. Some enter complete dates, other seasons and years, others only years ... It is important to categorize these descriptions into a consistent set of occupations to characterize the types of experience of candidates. This work requires the judgment of a labor economist who is trained to categorize jobs.
My analysis relies on mathematical imputations to estimate the proportion of applicants who are African American, because Peoplemark’s application forms and related data do not ask individuals seeking employment to identify their race or ethnicity. People mark has provided the racial identity of persons hired, leaving the racial identity of rejected applicants unknown. Therefore, I must estimate the African American representation among applicants in order to compare the racial composition of applicants to the racial composition of hires.
... By weighing African American representation in each occupation and location pool according to the number of applicants in the pool and summing over all the pools for any hiring location, I can estimate African American representation in the pool. I then use standard statistical methodology to evaluate whether there is a difference between the racial compositions of the applicant pool and of those who are hired.
The analysis of whether African American applicants were more likely to report criminal convictions than were applicants from other ethnic and racial groups requires that I compare the proportion of African Americans among all other applicants ...
... The writing of the report requires about two weeks, one of which must occur after the computer work has been completed.
Although the court requested the supplemental briefing on the necessity of additional time for producing Dr. Madden’s report, it essentially disregarded Dr. Madden’s detailed recitation of the steps she needed to take and the time she would need to accomplish those tasks. See section F infra.
*610On November 6, 2009, Valora timely produced the e-database to Dr. Madden’s team.
E. December-January 2009: Depositions of Managers, Former Employees, and Chief Counsel
From December 15, 2009, to January 12, 2010, the EEOC deposed managers from all of Peoplemark’s locations, former employees, and Peoplemark’s Chief Counsel, Osten.
Peoplemark’s managers revealed that much of the felony-screening is driven by client preferences, and as a result, some offices hire felons while others do not.
Peoplemark’s Tennessee manager stated that she received verbal instructions not to place ex-felons, she follows those instructions in the Memphis office, and she has no clients that accept ex-felons.
Peoplemark’s Michigan manager stated that his office denies ex-felon applicants because it does not have clients willing to accept these applicants.
Peoplemark’s Kentucky manager stated that his office occasionally places applicants without first conducting background checks. If a later background check reveals a felony, Peoplemark informs the employer and allows it to decide whether to continue to employ the individual. However, if after an employee is placed Peoplemark finds a violent felony conviction or multiple convictions, then People-mark will terminate the employee itself.
Peoplemark’s Florida manager disclosed the most liberal policy for placing ex-felons. According to the manager, the Florida office accepts applications from ex-felons and engages in a client driven screening process on a ease by case basis in deciding whether to place the individual.
On December 17, 2009, the EEOC deposed two former Peoplemark employees. Both of the individuals stated that People-mark placed them even though they had felony records and informed Peoplemark of their prior convictions.
Finally, on December 16, 2009, the EEOC deposed Osten. The EEOC asked Osten why, during the administrative investigation, he repeatedly stated that Peoplemark had a blanket no-felon policy. Osten replied that he did not remember if he had asked someone else within the company before making the statements. His recollection was that it was “pretty generally well-known” that Peoplemark did not hire convicted felons and that clients did not want to hire convicted felons for various reasons.
F. Court’s Partial Grant of Motion to Extend Deadline to Produce Dr. Madden’s Report
Following depositions, on December 21, 2009, the Magistrate Judge agreed to give the EEOC until December 31, 2009, to produce Dr. Madden’s statistical analysis. This extension gave the EEOC’s expert a little over a month and a half to code, analyze, and write a report detailing the information contained in 45,000 applications. Although Dr. Madden had repeatedly informed the court that she needed three months from the time the document processor produced a uniform information database to complete her statistical analysis, the court gave her approximately half that amount of time.
In determining what it believed to be an appropriate extension, the court relied on the EEOC’s statement in its second motion to extend that it would be able to process Peoplemark’s application documents and produce an expert report in three months time. The court relied on this timeframe even though the EEOC explicitly stated in that motion that its *611time estimate was based on Peoplemark producing 50,000 papers.12
The court reasoned:
Since all of the defendant’s documents were provided to the EEOC by August 24, 2009 (with the minimal exception of 600 pages found later in September), one could reasonably expect plaintiff to have produced its expert report by November 24, 2009, based on the opinion of its own in-house expert.
In its earlier renewed motion to extend deadlines filed July 23, 2009, plaintiff stated it could provide its experts’ reports by February 12, 2010 ... of course, five months was considerably more than [the EEOC] estimated ... and even with five months to complete the expert report, it would be done by December 31.13
[It is] readily apparent that a substantial portion of the delay in this case has been the plaintiffs lack of diligence in bringing the initial vendor up to speed.
Finally, the court speculated that the EEOC’s failure to send Peoplemark’s e-records to its document processor substantially delayed processing.14
The court went on to state an alternative hypothesis for the delay in producing expert reports:
On one hand, it would appear [the document processor] should have had the overwhelming bulk of the data its [sic] hands no later than the end of August 2009, and that the EEOC was simply not being diligent in transferring it to [the document processor].
There is a more probable explanation for the latest delay, however ... [0]n July 31, 2009, the two experts identified by plaintiff had “not yet been retained” by the EEOC ... Since Dr. Madden would not even have begun developing her protocol for [the document processor] to use until she was under contract, and she was not under contract until September 2009, [the document processor] could not have begun its keypunching operation until sometime in September. Had plaintiff been diligent and hired its experts as soon as it decided to litigate the case (well over a year ago) or even in the initial months of the lawsuit (a year ago), or even during the first several months of 2009, [the document processor] could have started key coding the documents for the database as soon *612as discovery permitted (perhaps even beginning with the 18,500 documents plaintiff obtained during its administrative investigation).
And had plaintiff been ready in the first half of 2009 to process documents it knew it would be receiving, it might have been more diligent in pursuing the discovery of those documents as soon as the case management order was in place.
In light of all that plaintiffs counsel have been told by the court in this case, one would have assumed that counsel would have hand-carried the necessary paperwork around the EEOC bureaucracy to get Dr. Madden hired immediately upon her being identified. That this did not happen only reinforces the conclusion that the EEOC is continuing what is tantamount to a “too big to fail” strategy. The EEOC has undertaken a major case which it is going to investigate and litigate at its own pace regardless of whatever discretion the court might attempt to exercise. Having not obtained the pretrial schedule it proposed at the Rule 16 conference, plaintiff has, by continually dragging its feet through discovery, and more particularly by failing to obtain crucial expert assistance in any kind of timely manner, sought to achieve the same result by presenting the court with a fait accompli.
Had plaintiff made a serious attempt during the past year to adhere to the case management order, or had it made a more reasonable request for an extension of the expert deadlines, its position might be more tenable ... But it has simply been in denial of its obligations, apparently believing that where the EEOC is concerned the usual rules do not apply.
(emphasis added).
According to the court, it was reasonable to give Dr. Madden’s team only a month and a half (after receiving the uniform application information from the document processor on November 6), instead of the three months she stated was needed.15
The EEOC disagreed. On December 30, 2009, the EEOC filed notice that it was appealing the Magistrate Judge’s order. The District Judge affirmed the order on January 29, 2010.
Following that affirmance, which precluded the EEOC from being able to timely produce Dr. Madden’s report, she and her team stopped working on February 2, 2010. At that time, the team was on pace to have their expert report completed and filed with the court on February 12, 2010.
After the district court affirmed the Magistrate Judge’s final case management order, the EEOC realized that it would be unable to survive summary judgement because it could not prove the casual element of its prima facie case.16 Without Dr. *613Madden’s statistical analysis of the percentage of African American’s within the applicant pool that were placed, it was impossible to show that considering felonies caused Peoplemark to hire a reduced number of African Americans from the applicant pool.
G. Peoplemark’s Motion for Summary Judgment, Production of Expert Report, and Settlement Negotiations
On February 25, 2010, Peoplemark moved for summary judgment. On the same day, the EEOC approached People-mark about a settlement agreement. The EEOC proposed that both parties agree to pay their own fees and expenses. The EEOC confirmed its oral proposal in writing the next day. Discussions continued into March.
On March 8, 2010, the EEOC requested that Peoplemark provide a response by the close of business the following day.
On March 11, 2010, Peoplemark countered with a request that the EEOC pay $500,000 of Peoplemark’s fees. The EEOC rejected the counter proposal the same day and emailed a proposed stipulation and order of dismissal to Peoplemark.
On March 24, 2010, the parties jointly moved to dismiss the case with prejudice.17 The proposed dismissal provided that Peoplemark would be treated as the prevailing party for the purposes of determining fees.
1. Peoplemark’s Expert Report (Provided with Summary Judgment Motion)
Peoplemark produced its expert report from Dr. Malcom Cohen in conjunction with its motion for summary judgment in February 2010. Peoplemark’s expert performed two tasks: 1) he responded to Dr. Pager’s sociology study and 2) he analyzed the limited data contained in Peoplemark’s electronic database.
Dr. Cohen did not code or analyze the paper applicant data for the 45,000 applicants the EEOC identified during discovery. Instead, Dr. Cohen analyzed only the limited applicant data in Peoplemark’s e-records and two months of paper data from the Tennessee office.
According to Dr. Cohen, the e-records included applicant names, customers, and placements for the Grand Rapids, Orlando, and Livonia offices, but for Memphis and Owensboro offices, the database included only applicant names.
Although he had only a fraction of Peoplemark’s applicant information for the relevant time period and no felony data, Dr. Cohen “found no evidence that African Americans are hired at a lower rate than their representation in the labor force in the relevant occupational groups and geographic areas.” He also concluded that, “when a comparison is made of the percent of African American placements to the available labor pool, there is no evidence of disparity.”
To analyze the Memphis office, which was not included in the e-database, Dr. Cohen requested the race of applicant placements made in two random months. He had no information concerning felony convictions for the applicants. Based on this data alone, Dr. Cohen concluded that Peoplemark’s placements met or substantially exceeded the percentage of African Americans in the relevant labor market.
*614Similarly, the e-records did not contain information on the Kentucky office. For an unknown reason, Dr. Cohen did not request the same race placement information from the manager in the Kentucky office. Instead, he did not provide any indication of whether Peoplemark met the benchmark hiring requirement for the relevant labor market in Kentucky.
Although Dr. Cohen had database information for the remaining three offices, there is zero indication that he had any information concerning felony convictions of applicants. Furthermore, Dr. Cohen stated that because of the way that he coded jobs, some workers were counted multiple times, once for each unique EEO code that he assigned to a job.
Dr. Cohen had database information for the Livonia, Michigan office from 2004 to May 2006, when the office closed. Based on 747 employee applications and 937 placements, Dr. Cohen concluded that this office exceeded the benchmarks for the relevant labor market by between twenty to thirty percent.
For Grand Rapids, Michigan, Dr. Cohen analyzed 5,782 employee applications and 5,379 assignments. He concluded that Peoplemark’s hiring exceeded the relevant benchmarks by between nineteen and twenty-six percent in the relevant Michigan labor markets.
Dr. Cohen analyzed 6,563 applications and 8,239 assignments from the Orlando, Florida office. Based on this information he concluded that Peoplemark’s hiring exceed the relevant benchmarks by between twenty-two and twenty-nine percent.
Dr. Cohen did not consider the impact of felony convictions on Peoplemark’s placement rate for African American applicants. Moreover, the application materials he did review covered only a small fraction of the company’s total applicants.
H. Peoplemark’s Motion for Attorney’s Fees and Costs
Peoplemark moved for attorney’s fees, expert witness fees, sanctions, and costs in excess of $1,300,000. On March 31, 2011, the Magistrate Judge awarded People-mark $751,942.48. The Judge awarded attorney’s fees from October 1, 2009, onward and all expert fees incurred from the outset of litigation.
In explaining his fee award, the Magistrate Judge stated that an award of fees was justified in this case because the EEOC’s claim was meritless from the beginning and the EEOC conducted itself unreasonably during the course of pre-trial litigation. Specifically, the judge noted, inter alia:
This is one of those cases where the complaint turned out to be without foundation from the beginning. Once the EEOC became aware that its assertion that Peoplemark categorically refused to hire any person with a criminal record was not true, or once the EEOC should have known that, it was unreasonable for the EEOC to continue to litigate on the basis of that claim____ The report of Peoplemark’s expert shows that 22% of the 286 so-called victims of defendant’s purported policy had in fact been hired despite having felony records. Assuming this to be true, it means that the EEOC’s ‘blanket rejection of all felons’ claim upon which it chose to premise its disparate impact lawsuit was not well grounded. The question here is when the EEOC should have realized its claim of a blanket policy by Peoplemark had no foundation ...
It appears the EEOC was basing its whole case on some early statements made by an officer of the company coupled with the fact that Sherri Smith [sic] *615had not been hired, and did little independent follow-up. The court is left with the impression that plaintiffs counsel were content to simply place the matter in the hands of their experts and let them run with it.
There is no indication the EEOC conducted any investigation of the specific allegations of the 286 people it belatedly named as victims in this matter nor issued a reasonable cause determination as to those allegations prior to filing its complaint. However, the present case was litigated on other grounds. Had this case been dismissed on the basis that it was not properly investigated before it was brought, the attorney’s fees would be far greater. [T]he EEOC also tries to play down the significance of its claim in this lawsuit ... the EEOC argues that the very gravamen of its claim (that Peoplemark had a blanket policy) is irrelevant ... Obviously, whether People-mark had a blanket policy was not only relevant to the EEOC’s argument, it was crucial, and the EEOC never amended its complaint to assert otherwise. A good investigation would probably have shown that the EEOC could not make this case even prior to the filing of the lawsuit, but that certainly became evident when all of the evidence submitted by the other side was available even if the EEOC had not conducted its own independent investigation.
[R]egardless of the merits, it was unreasonable for the EEOC to continue to litigate (and drive up defendant’s costs) once it knew it could not produce an expert and thus could not prove its case. On this basis, plaintiff should have folded its tent no later than the end of 2009 when it was foreclosed from using its expert because she had not provided a timely report ... [T]he EEOC failed to adequately manage the prosecution of this case from the beginning. Notwithstanding that the EEOC knew from the time of its administrative investigation that it would need expert testimony to make its disparate impact case, ... it did not hire its expert until September 2009 ... Thereafter, having already received several continuances, the EEOC sought a further continuance of over one-third of a year to provide the expert’s report.
(emphasis added).
The District Judge affirmed the Magistrate Judge’s order granting fees in toto. The District Judge stated, inter alia:
[T]he EEOC unequivocally based its case on Peoplemark’s alleged categorical prohibition against the “hiring of any person with a criminal record.” ... Had the EEOC conducted a reasonable investigation in the years leading up to filing its lawsuit, or had it reviewed the evidence provided to it throughout the course of discovery, it would have quickly realized its theory of liability as pled was untenable ... the magistrate judge ... correctly concluded that the EEOC should have been aware of the fatal flaw in its case no later than October 1, 2009. See ... EEOC v. CRST Van Expedited, Inc., No. 07-cv-95, 2010 WL 520564 (N.D.Iowa Feb. 9, 2010) (holding EEOC lacked foundation to proceed based on failure to reasonably investigate ... ); EEOC v. Cintas Corp., Nos. 04-40132, 06-12311, 2011 WL 3359622 (E.D.Mich. Aug. 4, 2011) (holding failure to timely identify aggrieved parties and reasonably investigate such claims warranted attorney fee award against the EEOC ... ).
[A]s the magistrate judge aptly noted, “regardless of the merits, it was unreasonable for the EEOC to continue to litigate (and drive up defendant’s costs) *616once it knew it could not produce an expert and thus could not prove its case.”
[T]he EEOC filed a motion to extend [the case management deadlines] the day before it was required to name its experts.
Unsatisfied with [the court’s relief], the EEOC renewed its motion to extend deadlines ... [T]he EEOC indicated it had yet to identify or hire any expert, despite having litigated the case for ten months, investigated the case for several years, and knowing that it would need an expert from day one ... its expert never prepared a report, even though the court provided yet another deadline extension.
(Emphasis added).
II. Discussion:
In Christiansburg Garment Co. v. Commission, 434 U.S. 412, 416, 98 S.Ct. 694, 54 L.Ed.2d 648 (1978), the Supreme Court stated that because a plaintiff acts as a private attorney general vindicating public interests, a prevailing plaintiff ordinarily is to be awarded attorney’s fees in all but special circumstances. The Court in Christiansburg also stated that courts should not use the same standard to determine whether a prevailing defendant is entitled to attorney’s fees. Id. Rather, the Court found that a trial court should not award attorney’s fees to a defendant unless a it finds that the plaintiffs claim was “frivolous, unreasonable, or groundless, or that the plaintiff continued to litigate after it clearly became so.” Id. at 701.
The Court went on to state that trial courts must “resist the understandable temptation to engage in post hoc reasoning by concluding that, because a plaintiff did not ultimately prevail, his action must have been unreasonable or without foundation.” Id. at 700. “[T]he course of litigation is rarely predictable. Decisive facts may not emerge until discovery or trial ... Even when the law or the facts appear questionable or unfavorable at the outset, a party may have an entirely reasonable ground for bringing suit.” Id.
This court has further dissected the Christiansburg “frivolous, unreasonable, or groundless” standard into two separate analyses. To determine whether an award of fees is proper, this court looks to: 1) the substantive merits of a plaintiffs case and 2) the plaintiffs litigation tactics or method of prosecution. See e.g. Serrano, supra, 699 F.3d at 905 (6th Cir.2012) (determining that fees were not warranted because the legal issues raised were not “frivolous” or “groundless” and because losing motions, failing to respond properly to discovery requests, and withdrawal of claims did not constitute “egregious and unreasonable [litigation] conduct”).
In this case, the lower court and the majority find that the EEOC: 1) brought a case that was meritless as pled and 2) unreasonably continued to prosecute the case after it should have known its case was meritless.18 In contrast to the majority, I find a justification of fees on either basis clearly erroneous. As such, I will address both in turn.
*617A. Attorney’s Fees Are Not Warranted Because the EEOC’s Disparate Impact Claim Was Not Meritless From the Outset or At Any Later Point
1. Failure to Prove a Blanket No-Felon Policy
The majority disagrees with the lower court’s opinion that the EEOC’s case was “groundless from the outset.” The majority agrees, however, with the lower court’s opinion that the EEOC knew or should have known its case was meritless as pled when discovery revealed Peoplemark had placed some felons. According to the majority, for the EEOC’s case to have merit, the EEOC needed to be able to prove that the “pled, specific employment practice had an adverse impact on a protected group.” 19
I do not disagree that it was reasonable to expect the EEOC to know that People-mark hired some felons by October 1, 2009 — the date from which the lower court awarded fees. However, whether the EEOC knew that Peoplemark hired some felons is simply immaterial to the issue of whether the agency’s case had merit in fact.20
*618In reaching its conclusion that the EEOC should have known its case lacked merit as of October 1, 2009, the majority relied on flawed reasoning. The majority reasoned that the EEOC must match the employment practice it pled to the employment practice it ultimately would choose to allege in its prima facie case.
The majority relies on Johnson v. U.S. Dept. Of Health and Human Servs., 30 F.3d 45, 48 (6th Cir.1994), for the proposition that the EEOC was required to prove through statistical evidence that the pled, specific employment practice had an adverse impact on a protected group.
The EEOC has repeatedly asserted throughout the lower court proceedings and in its brief to this court that this view of the merits of a disparate impact claim is flawed. I agree.
The requirement that the EEOC match the employment policy in its pleading to the employment policy alleged in its prima facie case is flawed for two reasons: 1) it retroactively imposes the heightened pleading requirement that the Supreme Court rejected in Swierkiewicz, supra, and 2) it attempts to force the EEOC to prematurely meet its evidentiary burden at an arbitrary point after pleading, but before the close of discovery.
In Swierkiewicz, the Supreme Court decided “the question whether a complaint in an employment discrimination lawsuit must contain specific facts establishing a prima facie case of discrimination.” Swierkiewicz, 534 U.S. at 508, 122 S.Ct. 992 (2002). The Court held that “an employment discrimination complaint need not include such facts and instead must contain only a “short and plain statement of the claim showing that the pleader is entitled to relief.” Fed. Rule Civ. Proc. 8(a)(2).” Id. The Court went on to explain that a plaintiffs proof of a prima facie case is “an evidentiary standard, not a pleading requirement.” Id. at 510, 122 S.Ct. 992. Evidentiary standards concern the order and allocation of proof at trial or the dispositive motion stage, (emphasis in original). Id.
To measure a plaintiffs complaint against a particular formulation of the prima facie case at the pleading stage is inappropriate. Before discovery has unearthed relevant facts and evidence, it may be difficult to define the precise formulation of the required prima facie case in a particular case. Given that the prima facie case operates as a flexible evidentiary standard, it should not be transposed into a rigid pleading standard for discrimination cases.
Id. at 512, 122 S.Ct. 992. (internal citations omitted).
The Court in Swierkiewicz continued: This simplified notice pleading standard relies on liberal discovery rules and summary judgment motions to define disputed facts and issues and to dispose of unmeritorious claims. The provisions for discovery are so flexible and the provisions for pretrial procedure and summary judgment so effective, that attempted surprise in federal practice is aborted very easily, synthetic issues detected, and the gravamen of the dispute brought frankly into the open for the inspection of the court.
‡ ‡ ‡
If a pleading fails to specify the allegations in a manner that provides sufficient notice, a defendant can move for a more definite statement under Rule 12(e) before responding ... The Federal Rules reject the approach that pleading is a game of skill in which one misstep by counsel may be decisive to the outcome and accept the principle that the purpose of pleading is to facilitate a proper decision on the merits.
Id. at 512, 514, 122 S.Ct. 992.
Although the Supreme Court revisited the issue of notice pleading in Bell Atlantic *619Corp. v. Twombly, 550 U.S. 544, 127 S.Ct. 1955, 167 L.Ed.2d 929 (2007) and Ashcroft v. Iqbal, 556 U.S. 662, 129 S.Ct. 1937, 173 L.Ed.2d 868 (2009), the Court in Twombly stated clearly that its discussion of pleadings “does not run counter to Swierkiewicz.” 550 U.S. at 547, 127 S.Ct. 1955.
In Serrano, supra, this court recently confirmed that Swierkiewicz remains good law. The court, affirmatively stated that at the pleadings stage, a Title VII plaintiffs burden is slight. Id. at 894-95. “ ‘[A]ny Title VII plaintiff must carry the initial burden of offering evidence adequate to create an inference that an employment decision was based on a discriminatory criterion illegal under the Act ... [A] plaintiff has flexibility in how she meets that initial burden, and variance based on the facts of the case is expected.’ ” 699 F.3d at 894 (citing International Brotherhood of Teamsters v. United States, 431 U.S. 324, 358, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977)).
a. Requiring Amendment Retroactively Imposes a Heightened Pleading Standard on the EEOC
Instead of arguing that the EEOC’s disparate impact claim was meritless from the beginning, the majority and People-mark fault the EEOC for not amending its original complaint to state a different employment policy after discovering that Peoplemark had placed some felons during the relevant time period. This is problematic because it attempts to retroactively impose the heightened pleading requirement that the Supreme Court rejected in Swierkiewicz.
It is clear that, with the exception of a motion for a more definite statement, Title VII plaintiffs are not required to better elucidate their claim until the dispositive motion stage or at trial. At that point they have the evidentiary burden of coming forward with a prima facie case. Up until that point, Title VII plaintiffs are free to develop their theories of the case as they receive and process discovery, including, most importantly, the employer’s information. As succinctly stated in E.E.O.C. v. Scrub, Inc., 2009 WL 3458530, *2:
Iqbal and Twombly did not repudiate general notice-pleading, emphasizing, citing Swierkiewicz, that the manner of proof is distinct from pleading requirements. Furthermore, a complaint need not include evidence, nor allege all facts logically raised by the claim ... Litigants are entitled to discovery before being put to their proof; treating a complaint’s allegations as a statement of proof leads to ‘windy complaints’ and defeats Rule 8’s function.
Requiring a Title VII plaintiff to amend or, in the majority’s phrasing “reassess,” its complaint after filing but before discovery is complete should be rejected on this basis alone. To hold otherwise creates an end run around the Supreme Court’s holding in Swierkiewicz — it permits courts to indirectly force a Title VII plaintiff to plead evidence establishing its prima facie case in its complaint.
b. Defendants Bear the Burden of Requesting Clarification Because There is no Non-Arbitrary Point to Require Amendment
There is another reason to reject the majority’s reasoning. Namely, any point that a court picks during discovery to force a Title VII plaintiff to its proof will be completely arbitrary. This is why the Supreme Court stated in Swierkiewicz that defendants should use motions for more definite statement or summary judgment to weed out “synthetic issues,” narrow the dispute, and bring the gravamen of the claim “frankly into the open for the inspection of the court.” 534 U.S. at 512, 122 S.Ct. 992.
*620After a plaintiff satisfies the notice pleading requirement it is the defendant’s job to narrow or clarify what is at issue through the various procedural devices available to it. It makes little sense to require the plaintiff to alert the court or the other party each time its theory of the case changes. In fact, it is rather well established that the Federal Rules of Civil Procedure do not require a plaintiff to plead legal theories. See Federal Practice and Procedure, Wright, et al„ § 1219 (3d ed.2012). It is also unclear how a court would judge what newly uncovered facts require the plaintiff to amend his or her complaint.
If an employer believes that an alleged discriminatee cannot prove an element of his or her case, then the burden is on it to bring a motion to dismiss or move for summary judgment when this fact becomes apparent.
The lower court shifted the burden of clarification from Peoplemark to the EEOC. It did so, moreover, without allowing the agency sufficient opportunity to determine, on the basis of an expert analysis of comprehensive data, precisely how it should frame its theory.
In addition to improperly shifting the burden of clarification, the lower court ignored Peoplemark’s failure to affirmatively seek clarification of the agency’s factual basis and theories. Instead of seeking clarification, Peoplemark employed, and the lower court endorsed, discovery disputes that needlessly and fruitlessly ate up a substantial portion of the time the court had allotted to discovery.
Because Peoplemark failed to seek clarification of the agency’s theory of the case, the EEOC was left in the dark. Before the court’s ruling on the agency’s final motion to amend the case management order to permit an additional two weeks for production of Dr. Madden’s report, the agency had no notice that either People-mark or the court took issue with its ability to prove the first element of its prima facie case.
Although the EEOC was not required to amend or otherwise clarify its allegation it did so. When the agency realized that the court may have been rushing expert discovery because it did not see how the EEOC could prove a prima facie ease based on its original blanket policy allegation, the agency stated that it no longer maintained that Peoplemark had never hired a felon. The lower court and the majority claim that the EEOC’s clarification was too little too late, but in reality it was more than the EEOC was required to do. Its statement of its evolved theory of the case came before it had completed the process — which the court ultimately prevented it from completing — of analyzing the facts and Dr. Madden’s expert report.
The court was forcing its hand, and the EEOC put forward the best proof that it could regarding the first element of its prima facie case. The EEOC shed as much light as possible on the theory it intended to present at trial in light of what it knew as of October 23, 2009.21
*621The EEOC informed the district court that Peoplemark had “denied certain admissions pertaining to its practice and refused to respond to the interrogatory regarding its practices at each location until August 20, 2009.” The EEOC further stated that after reviewing the partially accessible e-records Peoplemark provided in July, 2009, it discovered:
Peoplemark’s Grand Rapids facility appeared to have hired a few individuals with criminal records; that its Memphis location did not hire individuals with criminal records; and that, with respect to the Florida and Kentucky facilities, some individuals with criminal records appear to have been hired.
In light of these discoveries, the EEOC stated that it no longer maintained that Peoplemark had a categorical policy. Instead of alleging a blanket policy, it intended to proceed on the theory that People-mark’s policy of considering felonies ór excluding felons for some positions had a disparate impact on African Americans. The EEOC explained that if it could show that any policy of considering felony convictions had a disparate impact then Peoplemark would have the burden of proving a business necessity.
On this record, I find no support for the lower court and the majority’s conclusion that fees are proper because the EEOC should have known its case was meritless as of October 1, 2009. The agency had a good faith basis for its original pleading alleging a blanket policy.22 It was not required to plead any element of its prima facie case. Although the agency’s theory of the case evolved, it was not required to inform the court or Peoplemark of this fact. Nonetheless, the EEOC informed the court that it appeared that the company was placing some felons. Given these facts, I believe the lower court committed plain error in finding that the EEOC did not have a good faith basis for continuing the case. At the very least, such finding was an abuse of discretion.
The practical consequence of the majority’s decision is to endorse lower court conduct that unreasonably impairs the ability of employment discrimination plaintiffs, including the EEOC, to discover and present crucial facts. Courts cannot restrict a litigant’s ability to obtain evidence and construct its case and then fault the litigant for failing to have proof of its allegations. Moreover, courts cannot compound the abuse by using the lack of definitive proof as a basis for awarding fees.
By not allowing the additional two weeks for Dr. Madden’s report, the district court — arbitrarily, in my view — precluded the EEOC’s opportunity to show whether it did or did not have a viable claim. The court then reasoned backward and found the agency’s case was without merit and awarded fees, in part, on that basis. The court’s backward looking reasoning is precisely the type of flawed merits analysis the Supreme Court rejected in Smerkiewicz, supra.
Permitting courts to arbitrarily determine the merit of cases in the manner that occurred here drastically undercuts the protection afforded unsuccessful plaintiffs by the Christiansburg decision and doctrine.
*622In finding that the EEOC’s case lacked merit, the lower court focused solely and erroneously on the fact that Peoplemark had placed some African Americans. The lower court completely ignored the agency’s sociological expert and the deposition testimony of Peoplemark’s managers and chief counsel, which supported the EEOC’s basic contention that Peoplemark maintained a felony policy that disparately impacted African Americans. As such, the lower court clearly erred when it found the agency’s case to be meritless.
B. Fees Are Not Warranted Because the EEOC Did Not Use Unreasonable Litigation Tactics
The lower court and the majority also found that the EEOC’s unreasonable prosecution of the case warranted an award of fees. The lower court cited the EEOC’s: 1) presuit investigation, 2) requests to amend the case management order, 3) failure to produce and expert report, and 4) failure to immediately dismiss. According to the lower court, the unreasonableness of the agency’s actions justify, in part, the fees that the lower court awarded. The majority does not take issue with the lower court’s analysis in this regard. I find the lower court abused its discretion in basing an award of fees, in part, on the agency’s prosecution of the case. I address each of the EEOC’s “unreasonable” litigation tactics in turn.
1. The EEOC’s Presuit Investigation
The lower court found that the agency’s presuit investigation was insufficient and led to filing of a complaint that it should have known was meritless. This focus on the putative insufficiency of the presuit investigation is contrary to established legal doctrine applicable to EEOC investigations.
In determining whether a claim is frivolous or groundless, “it is error for the district court to inquire into the sufficiency of the Commission’s investigation.” EEOC v. Keco Industries, Inc., 748 F.2d 1097, 1100 (6th Cir.1984). “The nature and extent of an EEOC investigation into a discrimination claim is a matter within the discretion of that agency.” Id. The purpose of the EEOC investigation is to investigate the individual charge and facilitate informal conciliation between the alleged discriminatee and the employer. Id. A district court must judge the facts and legal issues in a discrimination claim de novo. Id. at 1102.
The majority ignores the fact that the lower court based its award of fees, in part, on the EEOC’s “insufficient” presuit investigation. I believe it is necessary to address the lower court’s consideration of the agency investigation because it is at odds with this court’s decision in Keco.
The issue in Keco was whether the lower court had improperly relied on the presuit investigation to determine the sufficiency of the complaint. This court held that the lower court had improperly inquired into the presuit investigation in order to determine the sufficiency of the complaint.
Based on the same reasoning this court used in Keco, I find the lower court’s award of fees in this case based, in part, on the EEOC’s investigation to be an abuse of discretion.
Both the Magistrate Judge and the District Judge repeatedly faulted the EEOC for failing to conduct a more thorough presuit investigation. Although the Magistrate Judge did not explicitly base his fee recommendation on the ground that the EEOC failed to conduct an adequate presuit investigation, he did state: “[h]ad this case been dismissed on the basis that it was not properly investigated before it was brought, the attorney’s fees would be far greater.” Furthermore, both the Magistrate Judge’s recommendation of fees and the District Judge’s order approving fees *623faulted the EEOC for not conducting a more thorough investigation and for not conciliating on behalf of the 286 initially and tentatively identified alleged victims individually.
On reading both orders in their entirety, and in light of the record as a whole, it is clear that both the lower court judges held mistaken beliefs respecting the scope, nature, and purpose of the EEOC’s presuit investigation and misunderstood their power to determine the sufficiency of that investigation.23 Neither order explicitly stated that the EEOC’s presuit investigation was an independent basis for awarding fees. I believe, however, that the lower court’s misunderstanding of the agency’s investigation process materially contributed to its clearly erroneous imposition of fees.
2. The EEOC’s Pre-Trial Litigation Conduct
The next issue is whether the EEOC’s pre-trial litigation conduct justifies an award of fees. In assessing the unreasonableness of the EEOC’s prosecution of this case the lower court cited the agency’s: 1) failure to identify a class list during discovery; 2) three requests for deadline extensions; and 3) ultimate failure to produce an expert report.
The majority fails to discuss the lower court’s analysis in its fee award of the EEOC’s pre-trial litigation conduct.
This court has read expansively the Supreme Court’s opinion in Christiansburg. Consideration of those decisions, I believe, shows that the lower court abused its discretion in awarding fees, in part, on the basis of the EEOC’s pre-trial litigation conduct. Any negligence or untimeliness issues on the part of the EEOC do not rise to the level of that egregiousness of conduct, which the law requires for an award of attorney’s fees. See infra section c.
a. The EEOC’s Class Notice and Certification Issues
I. Sufficient Notice of the Class Claim
The first issue is whether the EEOC’s inability to identify a class list early in litigation was so unreasonable as to warrant an award of attorney’s fees. The Eighth Circuit recently held that the EEOC must identify, investigate, and attempt to conciliate each individual class member’s claims during its pre-suit investigation. EEOC v. CRST Van Expedited, Inc., 679 F.3d 657, 671, 676 (8th Cir.2012) (stating that the EEOC cannot use discovery as a fishing expedition to uncover more violations).
However, the Eighth Circuit’s holding is at odds with this Court’s decision in Keco, supra, and the law in at least six other courts.
In Keco, this Court stated that a “complaint filed by the EEOC is limited to the investigation reasonably expected to grow out of the initial charge of discrimination.” Keco, supra, 748 F.2d at 1101 (stating that complaints alleging class allegations “grow out of’ individual charges of discrimination). In determining the sufficiency of the evidentiary basis for a class allegation in a complaint, a court may only consider whether the EEOC attempted, as the agency did here, to conciliate on behalf of a class of individuals. Id. at 1102 (emphasis added). “The form and substance of *624those conciliations is within the discretion of the EEOC as the agency created to administer and enforce our employment discrimination laws and is beyond judicial review.” Id.
At least two other circuit courts and three district courts concur with this court’s holding in Keco. If the EEOC’s presuit investigation and communications with an employer put the employer on notice that the EEOC is investigating a class claim, then the employer has sufficient notice that the EEOC is alleging a class claim. See EEOC v. Rhone-Poulenc, Inc., 876 F.2d 16, 17 (3d Cir.1989) (stating that the EEOC is not required to provide documentation of attempts to conciliate on behalf of each potential claimant in an employment class action); Marshall v. Sun Oil, 605 F.2d 1331, 1335 (5th Cir.1979) (stating that “unnecessarily thorough investigations of individual eases would tend to transform the conciliation process into something contrary to congressional intent” because the process would become “too formalized and unduly prolonged” and “focus would be directed away from voluntary compliance with the law to endless and irreconcilable bickering”); Dinkins v. Charoen Pokphand USA, Inc., 133 F.Supp.2d 1237, 1245-46 (M.D.Ala.2001) (stating that “[wjhat matters is that the EEOC served [the employer] notice that it was investigating possible discrimination against a class of women” and that the EEOC need not “conciliate each individual’s Title VII claim separately”); EEOC v. Jillian’s of Indianapolis, 279 F.Supp.2d 974, 983 (S.D.Ind.2003) (stating that the EEOC could proceed on behalf of a local but not a nationwide class without naming individual members because the employer had sufficient notice that the EEOC was investigating a local class); EEOC v. Dillard’s Inc., 2011 WL 2784516, *6-8 (S.D.Cal.2011) (stating that the EEOC is not required to identify potential class members and limiting EEOC to a local class because the EEOC limited its presuit investigation efforts to one store location).
These cases make clear that the EEOC provides sufficient notice of a class claim if its investigation puts the employer on notice that it is considering bringing a class-wide claim. The EEOC need not know or disclose the identity of individual claimants during the presuit investigation.
In this case, although the EEOC significantly limited its document requests during its investigation, it made clear to Peoplemark that it was investigating the company’s no felony placement policy on a company-wide scale. It is evident that the Peoplemark understood the scope of the investigation because its Chief Counsel, in an effort to reduce the company’s document production burden during the investigation, stipulated to Peoplemark’s company-wide policies. Furthermore, Peoplemark itself states that it discussed a settlement for hundreds of alleged victims.
As such, there is clear evidence in the record demonstrating that Peoplemark had notice the EEOC was investigating a class claim. It was an abuse of discretion for the lower court to further inquire into the investigation and fault the EEOC for failing to identify and individually conciliate each claim during the investigation.
ii. Rule 23 of the Federal Rules of Civil Procedure does not govern EEOC Class Claims
Assuming the EEOC was not precluded from alleging a class-wide claim because of an insufficient presuit investigation, the next issue is whether the EEOC’s inability to identify the individual class members after two and a half months of discovery was unreasonable. Generally, Fed. R.Civ.P. 23 prescribes the time for identification of class members and when a court should determine whether to certify a *625class. However, in General Tel. Co. of the Northwest Inc., v. EEOC, 446 U.S. 318, 324, 100 S.Ct. 1698, 64 L.Ed.2d 319 (1980), the Supreme Court held that the “the nature of the EEOC’s enforcement action is such that it is not properly characterized as a “class action” subject to the procedural requirements of Rule 23.” Although in General Tel. Co. Of the Northwest Inc., the EEOC brought an enforcement action alleging a “pattern or practice” of discrimination, the Court’s reasoning applies with equal force to an enforcement action alleging disparate impact. The Court stated:
Given the clear purpose of Title VII, the EEOC’s jurisdiction over enforcement, and the remedies available, the EEOC need look no further than § 706 for its authority to bring suit in its own name for the purpose, among others, of securing relief for a group of aggrieved individuals. Its authority to bring such actions is in no way dependent upon Rule 23, and the Rule has no application to a § 706 suit.
Id. at 324, 100 S.Ct. 1698.
Section 706 gives the EEOC authority to prosecute and obtain remedial relief for a violation of any provision of the act, including a violation alleging disparate impact. 42 U.S.C. § 2000e-5(f)(l). Accordingly, the Court’s statement that EEOC need not follow Rule 23 should apply equally to enforcement proceedings alleging disparate impact.
The EEOC’s failure to identify and certify a class before the lower court’s discovery deadline cannot fairly be called an unreasonable litigation tactic, which justifies an award of attorney’s fees.
b. EEOC’s Motions to Extend, Failure to Submit an Expert Report, and Failure to Immediately Dismiss
In addition to citing the EEOC’s inability to identify a class of victims during the litigation, the lower court identified the EEOC’s additional unreasonable litigation conduct: 1) unsuccessful motions to extend, 2) failure to submit an expert report, and 3) failure to dismiss immediately after failing to produce an expert report.
In making its determination the lower court relied, in part, on EEOC v. Cintas Corp., 2011 WL 3359622 (E.D.Mich.2011), rev’d sub nom. Serrano v. Cintas Corp., 699 F.3d 884 (6th Cir.2012). In reversing Cintas, this court held in Serrano, that filing and losing over a dozen discovery and other motions, including motions to extend deadlines, failing to respond properly to requests to identify alleged discrimination victims, and identifying and later withdrawing claims on behalf of alleged discriminatees did not constitute egregious and unreasonable conduct justifying an award of fees. Serrano 699 F.3d at 905.
Consistent with its disregard of the other elements of the lower court’s unreasonable prosecution discussion, the majority fails to discuss the lower court’s reliance on Cintas.
This court in Serrano addressed EEOC conduct, which was almost identical to the conduct in this case and found the conduct was not so unreasonable as to justify an award of fees. See Serrano, 699 F.3d at 905. I find that this court’s holding in Serrano necessitates a finding that the lower court abused its discretion when it awarded fees based on the same reasoning that this court rejected in that case.
Moreover, this court has repeatedly held that the inconvenience caused by run-of-the-mill discovery disputes does not warrant an award of fees. “An award of attorney’s fees against a losing plaintiff in a civil rights action is an extreme sanction,” which “must be limited to truly egregious cases of misconduct.” Lowery v. Jefferson County Bd. Of Edu., 586 F.3d 427, 438, 439 (6th Cir.2009); Sensations, Inc. v. City of Grand Rapids, 526 F.3d 291, 300 (6th Cir.2008) (stating that “[t]he *626Sixth Circuit affirms awards of attorney-fees only when a plaintiff re-litigates already-settled legal matters, and we reverse the award of attorney fees when issues of law remain unresolved or when a plaintiff had an arguable basis for pursuing his or her claim”); Revis v. Meldrum, 489 F.3d 273, 293 (6th Cir.2007) (upholding fees only as to aggressive and unnecessary filings intended to harass defendants); Lisle v. Met. Gov’t of Nashville & Davidson County, 73 Fed.Appx. 782, 791 (6th Cir.2003) (stating fees appropriate only in cases “which not only should not have been brought, but which the plaintiff should have known not to bring”); Tahfs v. Proctor, 316 F.3d 584, 596 (6th Cir.2003); Dubuc v. Green Oak Twp., 312 F.3d 736, 755 (6th Cir.2002) (stating that fees not appropriate even though plaintiff was unsuccessful on similar claims in state court, which likely put him on notice of the groundlessness of his claim, because the claim preclusion issue was not clear cut); Riddle v. Egensperger, 266 F.3d 542, 553 (6th Cir.2001) (stating that reference to facts which “came to light during the extensive discovery” is exactly the type of post hoc reasoning prohibited by Christiansburg).
In my view, the lower court abused its discretion in finding that the EEOC unreasonably conducted itself during pre-trial litigation because it lost motions and failed to file its expert report. The lower court’s reasoning is directly at odd’s with this court’s decision in Serrano and cannot be squared with this court’s extensive case law prohibiting fee awards in all but the most egregious cases.
I. Motions to Extend and Failure to Submit an Expert Report
In faulting the agency for failing to timely submit Dr. Madden’s expert report, the court disregarded the EEOC’s reasonable explanations for its inability to meet the court’s deadlines. The court’s disregard of the circumstances necessitating the EEOC’s request for additional time to produce Dr. Madden’s report constituted an abuse of discretion.
Specifically, the court ignored the: 1) 250% increase in Peoplemark’s document production; 2) Dr. Madden’s expert opinion on the time necessary to complete a report; and 3) administrative constraints the agency faced.
Moreover, to the extent the lower court’s own view that Dr. Madden’s study took too long or included too much data influenced its opinion, the lower court abused its discretion. The lower court’s findings concerning Dr. Madden’s “failure to code in race” and its finding that both the document processor and Dr. Madden performed the same “coding” function were clearly erroneous.24
ii. Delay in Dismissal
In addition to taking issue with class certification and the untimeliness and ultimate failure to produce Dr. Madden’s expert report, the lower court held that the agency unreasonably delayed filing its dismissal order after the court had entered its summary judgment order in favor of Peoplemark. It found the agency’s efforts to settle the issue of costs and fees (and thus avoid an appeal) to be improper.
In my view, basing the fee award, in part, on the resulting brief period of delay was an abuse of discretion. It is not uncommon for parties to try to reach finality by compromise even after one of them has decisively prevailed either on motion or at trial. There is no indication that, in trying *627to avoid sanctions and further litigation, the EEOC was seeking to harass People-mark. At most, its unsuccessful attempt to settle was negligent, and negligence is not a basis for a fee award. Salkil, supra, 458 at 532.
As with the EEOC’s initial suggestion that the parties conduct fact discovery pri- or to expert discovery and its requests for additional time to produce Dr. Madden’s expert report, I believe the EEOC’s settlement attempt shows the agency’s intent to litigate in the most organized and efficient manner possible. I also think that an amount of deference, rather than criticism, is owed to the EEOC when it submits such requests. It is the agency we trust to enforce the discrimination laws of this country. I do not think it unreasonable to assume that the agency’s experience in employment discrimination litigation places them in the best position to suggest efficient discovery methods, reasonable timetables, or possible settlement options.
c. The EEOC’s Prosecution of the Case was Not Reckless or Intentional; Mere Negligence is Insufficient
Even considering the EEOC’s actions in the aggregate, I do not believe the agency conduct can be deemed so egregious and unreasonable as to warrant an award of fees. See Salkil, supra, 458 at 532 (reversing under same standard of review for 28 U.S.C. § 1927) (stating “simple inadvertence or negligence” is insufficient and requiring conduct that “trial judges, applying [the] collective wisdom of their experience on the bench could agree falls short of the obligations owed by a member of the bar to the court”). None of the agency’s actions involved re-litigation of already settled legal matters, which is the most common circumstance justifying fee awards in favor of prevailing defendants in civil rights cases. See Sensations, Inc., supra, 526 F.3d 291 (6th Cir.2008); Dubuc, supra, 312 F.3d 736; Carrion v. Yeshiva University, 535 F.2d 722, 728 (2d Cir.1976); Badillo v. Central Steel & Wire Co., 717 F.2d 1160, 1163 (7th Cir.1983).
Moreover, the procedural record in this case simply does not support a finding that the EEOC recklessly disregarded or intentionally thwarted the court’s deadlines. To the contrary, the EEOC provided well-documented requests to the court each time it sought amendments to the case management order. The EEOC clearly explained why it was unable to meet the court’s deadlines. It cited discovery disputes, federal regulations, and a substantial and unforeseeable increase in document production. The agency also provided the court with multiple affidavits from experts, individuals within the agency, and vendors in an effort to document its diligence. Despite these showings, the lower court chose to believe that the agency could have acted with more diligence in identifying a class of claimants, hiring and approving its experts, and, ultimately, in producing Dr. Madden’s expert report. In light of agency’s explanations and the totality of the evidence in the record, the court’s findings amount to no more than disbelief and speculation.
III. Conclusion
The Supreme Court intentionally set the bar high in Christiansburg. It recognized that defendant’s do not serve the public interest in defending employment discrimination suits. Thus, its only goal in awarding fees to defendants, in extremely rare instances, was to prevent frivolous suits or reckless and abusive ligation tactics. I believe that the lower court’s interpretation of the merits of this case is clearly erroneous and rests on a misinterpretation of the role of the prima facie case in employment discrimination law. If I am correct, then the EEOC’s case was not meritless at any point during the litigation *628and the court abused its discretion in awarding fees, in part, on the that basis. The remaining issues that the lower court found with the agency’s prosecution of the case, in my view, amount to no more than run-of-the-mill discovery disputes, which are simply insufficient as a matter of law to support an award of fees.25 For the foregoing reasons, I respectfully dissent.
. Most simply put, my disagreement with the majority’s analysis and result arises from my different reading, as explicated in this dissent, of the record. Where the majority sees misfocus and dilatoriness on the part of the EEOC, I see an effort to gain information to refocus and reassess the defendant’s conduct and practices, and, as importantly, obstructive tactics on the defendant’s part that needlessly, but successfully, ate up much of the time the court allocated for discovery. That success had a domino effect, causing delay in the EEOC's acquisition and assessment of crucial information (which was vastly more extensive than the EEOC had anticipated). That, in turn, plus unanticipated delays in creating the database essential to the expert’s analysis, made production of her expert report within the final deadline impossible. Had the court given the expert the additional six weeks to finish her work and submit the report, this situation would not have arisen. Without the expert's report, it is easy to say by hindsight that the EEOC case was a complete misfire. Because the lower court did not grant the request for a very modest amount of time to finish the work on the report, neither it nor we can say that in fact there was “no there there” to the agency’s claim of discrimination. While her report may have confirmed that assessment, it may also have provided the sine qua non evidence of discrimination in placement sufficient to enable the EEOC to prevail.
I remain convinced that the lower court abused its discretion in not allowing the very modest bit of additional time needed to complete the expert’s report, which was crucial to its case. In doing so, the court disregarded the cumulative effects of the defendant's obstructionist tactics during earlier stages of the *596case. The time the lower court took to respond to the defense generated discovery disputes enhance the effectiveness of those tactics.
. Contrary to Peoplemark’s repeated assertions that Osten made this statement only during the investigation phase of this case, Osten asserted the same belief in his December 2009, deposition. At that time, long after this suit had been pending, he again stated, that Peoplemark maintained a “generally well-known” practice of not hiring felons.
. The cover sheets covered only a one year period and indicated only the name of an applicant and whether he or she disclosed a felony conviction. Based on Peoplemark’s representation that it hired no felons during this period, the EEOC assumed that, at a minimum, those applicants who had disclosed a conviction were not placed in temporary positions. From there the EEOC used statistical formulas to estimate the percentage of those individuals who were black. It used this methodology to attempt to conciliate the case.
. Unlike other federal agencies, the EEOC does not use its investigations to gather the maximum amount of possible information to prosecute cases within the agency administrative hearing process. The EEOC aims only to gather enough information to move the parties toward an informal agreement settling charges of discrimination. See discussion infra. The EEOC alleges that during the conciliation it used a formulaic approach to propose an amount of backpay for a group of hundreds of African Americans. It calculated the proposed settlement amount based on the average wage and duration of a temporary assignment. According to the EEOC, it never claimed to have identified particular individual victims or amounts owed to them during the investigation.
. At the Rule 16 conference the EEOC asked the court to postpone experts until the parties completed fact discovery. The court rejected the EEOC’s request and shortened the discovery period the EEOC proposed from fifteen to twelve months. See discussion, infra.
. In my view, the Magistrate Judge’s order constituted plain error. It ordered the EEOC to identify class membership without a basis for being able to do so. The information Peoplemark had provided was barren of the basic data — race of applicants, whether they had felony records, and whether they had, despite a felony record, been placed — that the EEOC needed to comply with the interrogatory. The Magistrate Judge compounded the error and abused his discretion when he imposed sanctions.
. Ultimately, emails between Peoplemark and the EEOC show that the EEOC only received one year of information sheets.
. Although the sample interrogatory in the record is complex, its segmentation appears to be an effort to help, not hinder People-mark.
. The details of how the interrogatory dispute proceeded are: In its April 7, 2009, response to the EEOC, Peoplemark answered only interrogatory No. 3(b). Peoplemark refused to produce any information responsive to the EEOC’s requests for application documents. On April 24, 2009, April 30, 2009, and May 7, 2009, the parties discussed the numerosity issue. On May 7, 2009, Peoplemark answered the EEOC's interrogatories through No. 6.c.(2), asserting the same numerosity argument. On May 18, 2009, the EEOC contacted the court's case manager to clarify the local rule on subparts. The case manager advised that the court permitted subparts and the parties should construe the Case Management Order’s twenty five question limit accordingly. The EEOC immediately conveyed this information to Peoplemark. On May 19, 2009, Peoplemark emailed the EEOC and stated that it was inappropriate for the EEOC to contact the court independently on the subpart issue. On May 20, 2009, Peoplemark offered to answer the remaining interrogatories if the EEOC agreed that it would not serve additional interrogatories without seeking the court’s leave. The EEOC rejected the proposal and stated that Peoplemark needed to answer the remaining interrogatories within ten days to avoid a motion to compel. Peoplemark answered the remaining interrogatories, but it still refused to produce the applications and information sheets the EEOC requested in its first set of interrogatories and request for production of documents.
. At various times throughout discovery, the EEOC's statistical/labor economist expert, Dr. Janice Madden, filed affidavits in support of the EEOC’s requests to extend deadlines. She explained that data processing took place in two parts and was the most time consuming phase of statistical analysis. First, a document processor needs to create a uniform e-database of applicant information from the variety of raw applicant/employee materials. Although a statistic’s expert could do this step as well, Madden explained that it is cheaper and more efficient to outsource the first task to a data processing company. According to Dr. Madden, after the e-database is complete, a team of experts must numerically code the descriptive data listed in the e-database so that it can be plugged into statistical formulas on a computer. Dr. Madden explained the coding task: *603By using the terms "coding the data,” I am referring to the process by which the electronic entry of alphameric information is transformed into a variable format that a statistical program can interpret. For example, the data entry of a hypothetical item such as “graduated Central high school” could be coded as “12” for years of education or as "1” for a variable indicating whether the applicant is a high school graduate (or with a “0” for someone who has not graduated).
. The EEOC requested an additional week for each deadline to take into account the Christmas holiday.
. In my view, the Magistrate Judge’s failure to acknowledge the more than 250% increase in the volume of material that the agency and Dr. Madden had to work through constituted plain error.
. It is unclear how the court concluded that December 31, 2009, was five months from August 24, 2009. A date of January 24, 2010, would have been five months from August 24, 2009. The February 11, 2010 date finally requested by the EEOC was extended slightly in order to accomodate the Christmas holiday.
. It is true that the EEOC received CDs of Peoplemark's e-records prior to receiving CDs of scanned copies of Peoplemark's paper records. However, it is highly unlikely that sending the e-records to the document processor prior to August would have saved any time. Peoplemark's expert relied on the company’s electronic application materials. The limited nature of his expert report shows the extremely scant data actually contained in Peoplemark’s e-records. For example, the report does not analyze felony convictions, suggesting the e-records did not contain felony information. Furthermore, the database contained zero placement information for either the Kentucky office or the Tennessee office. For the Grand Rapids, Livonia, and Orlando offices, the database contained only 13,079 “employee applications.” Thus, although, as the court stated, the e-database "represents the bulk of the personnel information electronically stored on CD’s that Peoplemark had,” it did not represent the bulk of personnel information held by Peoplemark generally-
. There is no indication in the record that Peoplemark contended that Dr. Madden declaration of the time she needed for her work was exaggerated or otherwise misstated. In other words, the court sua sponte, and without a basis other than its subjective view, rejected Dr. Madden's unequivocal, uncontested, and detailed representations.
. At the outset, in its initial pleading, the EEOC was only required to raise an inference of discrimination. Serrano v. Cintas Corp., 699 F.3d 884, 898 (6th Cir.2012). A plaintiff in an employment discrimination case is not required to state facts sufficient to establish a prima facie case in its pleading. Swierkiewicz v. Sorema N.A., 534 U.S. 506, 508, 122 S.Ct. 992, 152 L.Ed.2d 1 (2002). See supra discussion of merits. However, to survive a motion for summary judgement the EEOC needed to be able to prove a prima facie case of discrimination. In a disparate impact case, a plaintiff must show: 1) a specific employment practice; 2) statistical proof that the practice adversely affects a protected group; and 3) a causal link between the disparate impact and the challenged employment practice. Johnson v. U.S. Dep’t of Health & Human Servs., *61330 F.3d 45, 48 (6th Cir.1994); Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 655-58, 109 S.Ct. 2115, 104 L.Ed.2d 733 (1989).
. The EEOC was not idle from March 11-March 24. During this time, in addition to trying to negotiate a settlement, the EEOC was appealing the court’s gránt of People-mark's motion to compel items contained in its privilege log.
. The lower court found the EEOC's complaint of a blanket no-felon hiring policy was meritless from the outset and clearly meritless as of October 1, 2005. According to the lower court, the EEOC unreasonably prosecuted the case because it: 1) failed to conduct a more thorough presuit investigation, 2) failed to identify a class list; 3) failed to file an expert report, and 4) filed and lost motions to amend the case management order.
The majority states that when discovery revealed facts that were inconsistent with Peoplemark’s General Counsel’s statements about a blanket policy, the EEOC should have “reassessed its claim.”
. The majority argument boils down: 1) a plaintiff in a disparate impact case must properly identify the employer’s specific employment practice, the first element of its prima facie case, in its pleading; 2) the employment practice alleged in the pleading must match the employment practice alleged in the plaintiff’s prima facie case during a motion for summary judgement or at trial; and 3) if a plaintiff misidentifies the employer’s specific employment practice in its pleading and fails to amend the complaint prior to a motion for summary judgment, or trial then the plaintiff must necessarily lose her case for failing to correctly identify the employment practice. Both the lower court and the majority express that it is proper to require a plaintiff to match the employment practice it alleges in a pleading to the employment practice it alleges as the first prong of its prima facie case at summary judgment or trial. This proposition is contrary to the Supreme Court’s holding in Swierkiewicz that a plaintiff need not allege facts sufficient to support a prima facie case at the pleading stage. See, infra, discussion.
. I use the word "some” because the number of felons hired is not known. People-mark’s expert did not report any felony data. The EEOC was in the process of making this, among numerous other calculations, when the court declined to give it an additional two weeks to complete its report. A report showing that Peoplemark hired a minimal amount of felons would actually be consistent with the EEOC's original assertion that Peoplemark, Inc. had a blanket policy of not hiring felons, which was imperfectly implemented at its four offices throughout the country. This view finds support in the December, 2009 depositions of Peoplemark’s employees and Chief Counsel, near the end of the discovery period. One manager testified that sometimes felons "slip through the cracks.” Another manager testified that she had been told verbally not to hire felons. A third testified to her personal practice at her office location of doing individual assessments of convictions. Most importantly, in December, 2009, the company’s Chief Counsel continued to maintain that when he made his original statements from 2005-2007, it was “well-known” that Peoplemark had a policy of not hiring felons.
Given the conflicting testimonies regarding companywide practices and practices at the individual locations, it was not unreasonable for the EEOC to continue to assert that, as a matter of overall company policy, Peoplemark did not place felons for employment with its clients. That the policy may have been randomly ignored or disregarded did not undercut its existence.
Moreover, and most importantly, by aborting the process that would have led to Dr. Madden’s report, the lower court — and the majority — has no basis for determining whether the placement of felons was so infrequent and random that the existence of the policy, which the witnesses acknowledged existed, had the discriminatory effect the EEOC alleged. Both the lower court and the majority implicitly assume that Dr. Madden’s report on People-mark's complete records would not have shown that Peoplemark in fact had and implemented a felony policy that disparately impacted placement of African-Americans.
. At this point, the EEOC had not yet had the opportunity to depose the managers or the Chief Counsel. Had the EEOC conducted the depositions, it would have been able to tell the court that it intended to present multiple alternative theories at trial — one blanket verbal policy, another policy of considering felonies as a factor in the application process, and perhaps even a third policy of using a blanket rejection of felons for certain jobs. The plaintiff in a Title VII case, especially in a case of disparate impact, can only form a complete prima facie case after it has discovered all of the evidence in the hands of the employer and has had an opportunity to analyze the information. Both courts and academics have noted the informational asymmetry in these cases. See Swanson v. Citibank, N.A., 614 F.3d 400, 412 (2010) (Posner, J., Dissenting) (citing Suzette M. Malveaux, Front Loading and Heavy Lifting: How Pre-Dismissal Discovery Can Address the Detrimental Effect of *621Iqbal on Civil Rights Cases, 14 Lewis & Clark L.Rev. 65 (2010)).
. From the outset until nearly the last moments of discovery allowed by the lower court, Peoplemark's Chief Counsel continued to state Peoplemark had a "well-known” blanket no-felon policy. In my view, that statement from that official sufficed, without more, to show the agency had a good faith basis to continue litigating. Peoplemark’s lead lawyer (and Vice President of the company) was acknowledging the factual accuracy of the EEOC’s principal contention.
. As noted supra, the EEOC’s process is different from many federal agencies. It does not use the administrative investigation process to prepare for an agency trial before an administrative law judge. Instead the agency, through its investigation, only aims to gather enough information to help marshal the parties toward an informal settlement. The EEOC's principal goal is conciliation, not preparation for suit.
. The Supreme Court has affirmed repeatedly the type of analysis Dr. Madden intended to employ. See Anthony Boardman & Aidan Vining, The Role of Probative Statistics in Employment Discrimination Cases, 46 Law and Contemporary Problems, (1983).
. I agree with the majority that a plain reading of the statute confirms that expert fees are included within, or a subset of attorneys fees. Accordingly, I also find the lower court abused its discretion in awarding expert fees in this case.