The Americans with Disabilities Act (ADA) requires employers to reasonably accommodate their disabled employees; it does not endow all disabled persons with a job — or job schedule — of their choosing. Jane Harris, a Ford Motor Company employee with irritable bowel syndrome, sought a job schedule of her choosing: to work from home on an as-needed basis, up to four days per week. Ford denied her request, deeming regular and predictable on-site attendance essential to Harris’s highly interactive job. Ford’s papers and practices — -and Harris’s three past telecommuting failures — backed up its business judgment.
Nevertheless, the federal Equal Employment Opportunity Commission (EEOC) sued Ford under the ADA. It alleged that Ford failed to reasonably accommodate Harris by denying her telecommuting request and retaliated against *758her for bringing the issue to the EEOC’s attention. The district court granted summary judgment to Ford on both claims. We affirm.
I
The Ford Motor Company employs about 224,000 employees worldwide. True to its founder’s vision, Ford uses its employees in assembly lines to perform independent yet interconnected tasks. Resale buyers of steel come early on the lines— before any assembling begins. They purchase raw steel from steel suppliers and then, as their name suggests, resell the steel to parts manufacturers known as “stampers.” The stampers then supply the steel parts to the vehicle assemblers, who put together the vehicles.
As an intermediary between steel and parts suppliers, the resale buyer’s job is highly interactive. Some of the interactions occur by email and telephone. But many require good, old-fashioned interpersonal skills. During core business hours, for example, resale buyers meet with suppliers at their sites and with Ford employees and stampers at Ford’s site — meetings that Ford says are most effectively performed face to face. And Ford’s practice aligns with its preaching: It requires resale buyers to work in the same building as stampers so they can meet on a moment’s notice. This high level of interactivity and teamwork is why, in Ford’s judgment, “a resale buyer’s regular and predictable attendance in the workplace” is “essential to being a fully functioning member of the resale team.” R. 60-2 at ¶ 11.
A former Ford resale buyer with irritable bowel syndrome takes center stage in this case: Jane Harris. Her job performance was, on the whole, subpar. Early on in her six-plus year tenure, she won a few awards, and Ford recognized her for her “strong commodity knowledge” and “diligent[ ]” work effort. R. 66-2 at 2; R. 60-14 at 6. But over time, the awards and compliments morphed into low ratings and criticisms. Harris placed in the bottom 22% of her peer group in her fourth full year (2007) and in the bottom 10% in her fifth year (2008). It got worse. By her last year (2009), Harris “was not performing the basic functions of her position.” R. 60-2 at ¶ 14. Ford said she lacked interpersonal skills, delivered work late, didn’t show a concern for quality, and failed to properly communicate with the suppliers. She again ranked in the bottom 10% of her peers.
In addition to performing poorly while at work, she repeatedly missed work entirely. In 2008, she missed an average of 1.5 work days per week; in 2009, she was absent more than she was present. And when she didn’t miss work, she would often come in late and leave early. As her coworkers and supervisors put it, Harris worked on a “sporadic and unpredictable basis,” R. 60-8 at ¶ 4, and had “chronic attendance issues,” R. 60-2 at ¶ 8; R. 60-4 at ¶ 3.
Harris’s poor performance and high absenteeism harmed those around her. When she missed work, her teammates had to pick up the slack, including by taking on the functions that Harris could not perform at home. Her supervisors also had to assume her job responsibilities. Her absences caused the resale-buyer team “stress and frustration,” R. 60-8 at ¶¶ 4-5, further compounded Harris’s mistakes, and frustrated suppliers.
Harris’s irritable bowel syndrome of course contributed to the situation. It gave her uncontrollable diarrhea and fecal incontinence, sometimes so bad that “it” could “start[] pouring out of [her]” at work. R. 41 — 4 at 1. She occasionally *759couldn’t even make the one-hour drive to work without having an accident. The vicious cycle continued, as her symptoms increased her stress, and the increased stress worsened her symptoms — making her less likely to come to work.
Ford tried to help. Harris’s first supervisor, Dawn Gontko, for example, adjusted Harris’s schedule to help her establish regular and predictable attendance. Most significantly, Gontko allowed Harris two opportunities to “telecommute on an ad hoc basis” in an “Alternative Work Schedule.” R. 60-3 at ¶ 3. Under this schedule, Harris worked four 10-hour days (known as flex time) and could telecommute as needed on her work days. Each trial lasted one to two months. But neither succeeded: Despite the ad hoe telecommuting and flexible schedules, Harris “was unable to establish regular and consistent work hours” and failed “to perform the core objectives of the job.” Id.; R. 60-7 at 2.
Ford next tried its “Workplace Guidelines” — a reporting tool specially designed to help employees with attendance issues tied to illnesses. These also failed to improve Harris’s attendance or illness. So did the efforts of Harris’s next supervisor, John Gordon, which included allowing Harris to telecommute both during and after core business hours. R. 60-2 at ¶ 8. When this third telecommuting attempt failed, the act repeated itself: The new supervisor, like the old, employed the “Workplace Guidelines,” and the guidelines again failed to remedy Harris’s attendance problems or illness.
Undeterred by these three failed telecommuting attempts, Harris requested leave “to work up to four days per week from home.” R. 60-10 at 1. Gontko had told her, after all, that her job would be appropriate for telecommuting. Ford’s telecommuting policy generally said the same thing. And several of her coworkers telecommuted. So why couldn’t Harris?
Ford’s practice and policy limited telecommuting for resale buyers. In practice, Ford’s buyers telecommuted, at most, on one set day per week. That aligned with its policy, which makes clear that those jobs that require “face-to-face contact”— and those individuals who were not “strong performers” and who had poor time-management skills — were among those not “appropriate for telecommuting.” R. 60-11 at 4.
Before making a decision on the request, two of Ford’s human-resources representatives and Gordon met -with Harris. In the meeting, Gordon went through Harris’s ten main job responsibilities and asked Harris to comment on how she could perform those tasks from home. Of the ten tasks, Harris admitted that she could not perform four of them from home, including meetings with suppliers, making price quotes to stampers, and attending some required internal meetings. Harris added, however, that she did not envision needing to stay home four days per week, only that she wanted the freedom of “up to 4 days.” R. 66-10 at 3 (emphasis added). Harris’s higher-ups told her that they would get back to her about her request.
Ford determined that Harris’s proposed accommodation was unreasonable. Management met with Harris to inform her of the decision. Gordon again listed Harris’s ten job responsibilities: four that could not be performed at home; four that could not effectively be performed from home; and two that were “not significant enough to support telecommuting].” Id. at 4-5. Gordon explained the circumstances under which telecommuting could work: on a, predictable schedule where the strong-performing employee agrees to come to the worksite as needed even on days set for telecommuting. Harris’s coworkers who *760telecommuted fit that bill. But Harris didn’t, and neither did her proposed schedule.
Even though Ford did not grant her requested telecommuting schedule, management told Harris that they could accommodate her in other ways, such as moving her closer to the restroom or looking for jobs better suited for telecommuting. Harris turned down each alternative accommodation. The second meeting ended as Ford informed Harris that it would “talk with her again if she identifie[d] another accommodation.” R. 66-10 at 6. Harris never did. Rather, she sent an email one week later claiming that the denial of her request violated the ADA. And she filed a charge of discrimination with the EEOC a day after that.
The rest of Harris’s time at Ford did not go well. She felt threatened by Gordon in their weekly meetings scheduled to improve her attendance and. performance. And in July of 2009, she ranked in the bottom 10% of her peers for the second evaluation in a row. She disputed the evaluation, claiming that it represented retaliation by Ford for her filing of the discrimination charge. Though asked to elaborate, Harris never did. She instead began a Performance Enhancement Plan. Designed so that the tasks could be “easily” completed within a 30-day deadline, R. 60-15 at ¶ 6, Harris had to complete a one-page spreadsheet, resolve “material claims,” develop a plan to complete work that had been outstanding since the previous year, and the like. R. 60-2 at ¶¶ 20-22. Harris did not satisfy the requirements of the Plan, failing to complete the tasks either entirely or on time. After several years of subpar performance and high absences, this was apparently the last straw: Mike Kane (the Senior Purchasing Manager for Raw Materials) and Lisa King (his manager) decided to terminate Harris on September 10, 2009.
Almost two years later, on August 25, 2011, the EEOC sued Ford under the ADA. It alleged that Ford failed to reasonably accommodate Harris’s disability (violating 42 U.S.C. § 12112(a), (b)(5)(A)), and that it discharged her in retaliation for filing her charge (violating 42 U.S.C. § 12203(a)). On June 29, 2012, Ford moved for summary judgment.
The district court granted Ford’s motion on September 10, 2012. It concluded that “working from home up to four days per week is not [a] reasonable” accommodation under the ADA and that “the evidence [did] not cast doubt on Ford’s stated reason for terminating Harris’s employment: poor performance.” EEOC v. Ford Motor Co., No. 11-13742, 2012 WL 3945540, at *7-*8 (E.D.Mich. Sept. 10, 2012). The EEOC appealed, and a divided panel of this court reversed on both claims. EEOC v. Ford Motor Co., 752 F.3d 634 (6th Cir. 2014).
We granted en banc review, thereby vacating the panel’s decision. Giving fresh review to the district court’s summary-judgment decision and drawing reasonable inferences in the EEOC’s favor, we must determine whether there exists a “genuine dispute as to any material fact” on either issue: failure to accommodate or retaliation. Fed. Rule Civ. Proc. 56(a). At the summary-judgment stage, we view the facts “in the light most favorable to the nonmoving party” (usually by adopting the plaintiffs version of the facts) “only if there is a ‘genuine’ dispute as to those facts.” Scott v. Harris, 550 U.S. 372, 380, 127 S.Ct. 1769, 167 L.Ed.2d 686 (2007) (emphasis added). A “genuine” dispute exists when the plaintiff presents “significant probative evidence” “on which a reasonable jury could return a verdict for her.” Chappell v. City of Cleveland, 585 *761F.3d 901, 913 (6th Cir.2009); see Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 247-48, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986). Determining whether a genuine dispute exists of course requires a “fact-intensive, case-by-case” analysis. Dissent Op. at 770-71, 775-76, 785-86. But it equally requires looking to case law for guidance and addressing all the facts in the record—including those that uniformly cut against the plaintiff. Undertaking this analysis, we hold that there is no genuine dispute of material fact on this record: A reasonable jury could not return a verdict for the EEOC on either claim.
II
Many disabled individuals re- • quire accommodations to perform their jobs. The ADA addresses this reality by requiring companies like Ford to make “reasonable accommodations to the known ... limitations of an otherwise qualified individual with a disability” where such an accommodation does not cause the employer “undue hardship.” 42 U.S.C. § 12112(b)(5). To comply with the ADA, then, Ford must “reasonably] accommodat[e]” Harris (undisputedly a disabled individual for purposes of this appeal) if she is “qualified.” §§ 12112(a), (b)(5) (emphasis added); see Smith v. Ameritech, 129 F.3d 857, 866 (6th Cir.1997).
To be “qualified” under the ADA, Harris must be able to “perform the essential functions of [a resale buyer]” “with or without reasonable accommodation.” 42 U.S.C. § 12111(8). A “reasonable accommodation” may include “job restructuring [and] part-time or modified work schedules.” Id. at § 12111(9)(B). But it does not include removing an “essential function[ ]” from the position, for that is per se unreasonable. Brickers v. Cleveland Bd. of Educ., 145 F.3d 846, 850 (6th Cir.1998); see Sch. Bd. of Nassau Cnty. v. Arline, 480 U.S. 273, 287 n. 17, 107 S.Ct. 1123, 94 L.Ed.2d 307 (1987). The district court held that Harris was not qualified because her excessive absences prevented her from performing the essential functions of a resale buyer. We agree.
A
Is regular and predictable on-site job attendance an essential function (and a prerequisite to perform other essential functions) of Harris’s resale-buyer job? We hold that it is.
1
We do not write on a clean slate. Much ink has been spilled establishing a general rule that, with few exceptions, “an employee who does not come to work cannot perform any of his job functions, essential or otherwise.” EEOC v. Yellow Freight Sys., Inc., 253 F.3d 943, 948 (7th Cir.2001) (en banc) (quoting Tyndall v. Nat’l Educ. Ctrs., 31 F.3d 209, 213 (4th Cir.1994) (internal quotation marks omitted)). We will save the reader a skim by omitting a long string-cite of opinions that agree, but they do. E.g., Samper v. Providence St. Vincent Med. Ctr., 675 F.3d 1233, 1237-38 (9th Cir.2012) (collecting cases); Mason v. Avaya Commc’ns, Inc., 357 F.3d 1114, 1122-24 (10th Cir.2004) (same). Our Circuit has not bucked the trend. E.g., Ameritech, 129 F.3d at 867. And for good reason: “most jobs require the kind of teamwork, personal interaction, and supervision that simply cannot be had in a home office situation.” Rauen v. U.S. Tobacco Mfg. L.P., 319 F.3d 891, 896 (7th Cir.2003).
That general rule—that regularly attending work on-site is essential to most jobs, especially the interactive ones— aligns with the text of the ADA. Essential functions generally are those that the employer’s “judgment” and “written [job] de*762scription” prior to litigation deem essential. See 42 U.S.C. § 12111(8). And in most jobs, especially those involving teamwork and a high level of interaction, the employer will require regular and predictable on-site attendance from all employees (as evidenced by its words, policies, and practices).
The same goes for the EEOC’s regulations. They define essential functions as those that are “fundamental” (as opposed to “marginal”), 29 C.F.R. § 1630.2(n)(1), so that a job is “fundamentally alter[ed]” if an essential function is removed. 29 C.F.R. § Pt. 1630(n), App. at 394. To guide the essential-function inquiry, the regulations speak in factors — seven of them. The first two restate the statutory considerations. 29 C.F.R. § 1630.2(n)(3)(i)-(ii). The remaining five add other considerations. 29 C.F.R. § 1630,2(n)(3)(iii)-(vii). In many jobs, especially the interactive ones, all seven point toward finding regular and predictable on-site attendance essential. Take the amount of time performing that function, for example, § 1630.2(n)(3)(iii): Most of one’s work time is spent at work, and many interactive functions simply cannot be performed off site. Or take the consequences of failing to show up for work, § 1630.2(n)(3)(iv): They can be severe. See Equal Employment Advisory Council Supp. Br. 9. Ditto for the terms of the collective bargaining agreement, § 1630.2(n)(3)(v): They certainly won’t typically exempt regular attendance. Other employees’ work practices are no different, § 1630.2(n)(3)(vi)-(vii): Other employees usually attend work at the worksite. And so on, such that most jobs would be fundamentally altered if regular and predictable on-site attendance is removed.
The EEOC’s informal guidance on the matter cuts in the same direction. An employer may refuse a telecommuting request when, among other things, the job requires “face-to-face interaction and coordination of work with other employees,” “in-person interaction with outside colleagues, clients, or customers,” and “immediate access to documents or other information located only in the workplace.” EEOC Fact Sheet, Work At Home/Telework as a Reasonable Accommodation (Oct. 27, 2005), http://www.eeoc.gov/facts/ telework.html; cf EEOC, Employer Best Practices for Workers with Caregiving Responsibilities, http://www.eeoc.gov/policy/ docs/caregiver-best-practiees.html (Jan. 19, 2011) (explaining that “impromptu team meetings” are a valid factor for denying an employee the privilege to work in a flexible work schedule). That is because, as the EEOC elsewhere explains, “the inquiry into essential functions is not intended to second guess an employer’s business judgment with regard to production standards.” 29 C.F.R. § Pt. 1630(n), App. at 395. Nor is it meant “to require employers to lower such standards.” Id. But that’s what would happen in many jobs if regular, in-person attendance was not required.
A sometimes-forgotten guide likewise supports the general rule: common sense. Waggoner v. Olin Corp., 169 F.3d 481, 482-84 (7th Cir.1999). Non-lawyers would readily understand that regular on-site attendance is required for interactive jobs. Perhaps they would view it as “the basic, most fundamental” “activity” of their job. Webster’s Third New International Dictionary 777, 920 (1986) (defining “essential” and “function”). But equipped with a 1400-or-so page record, standards of review, burdens of proof, and a seven-factor balancing test, the answer may seem more difficult. Better to follow the commonsense notion that non-judges (and, to be fair to judges, our sister circuits) hold: Regular, in-person attendance is an *763essential function—and a prerequisite to essential functions—of most jobs, especially the interactive ones. That’s the same rule that case law from around the country, the statute’s language, its regulations, and the EEOC’s guidance all point toward. And it’s the controlling one here.
2
That rule has straightforward application here: Regular and predictable on-site attendance was essential for Harris’s position, and Harris’s repeated absences made her unable to perform the essential functions of a resale buyer. The required teamwork, meetings with suppliers and stampers, and on-site “availability to participate in ... face-to-face interactions,” R. 60-2 at ¶ 11, all necessitate a resale buyer’s regular and predictable attendance. For years Ford has required resale buyers to work in the same building as stampers, further evidencing its judgment that on-site attendance is essential. And the practice has been consistent with the policy: all other resale buyers regularly and predictably attend work on site. Indeed, even those who telecommute do so only one set day per week and agree in advance to come into work if needed. Sealing the deal are Harris’s experiences and admissions. Her excessive absences caused her to make mistakes and caused strife in those around her. And she agreed that four of her ten primary duties could not be performed from home. R. 66-10 at 2. On this record, the EEOC cannot show that regularly attending work was merely incidental to Harris’s job; it was essential to her job.
It follows that Harris’s up-to-four-days telecommuting proposal-which removed that essential function of her job— was unreasonable. Brickers, 145 F.3d at 850; Mason, 357 F.3d at 1124. The employee bears the burden of proposing an accommodation that will permit her to effectively perform the essential functions of her job. Jakubowski v. Christ Hosp., Inc., 627 F.3d 195, 202 (6th Cir.2010); accord Dissent Op. at 779. Harris proposed only one accommodation—one that would exempt her regular and predictable attendance from her resale-buyer job. In failure-to-accommodate claims where the employee requests an “accommodation that exempts her from an essential function,” “the essential functions and reasonable accommodation analyses [] run together.” Samper, 675 F.3d at 1240. One conclusion (the function is essential) leads to the other (the accommodation is not reasonable). That’s this case. Harris’s proposed accommodation was unreasonable.
Nor could Harris perform the essential functions of her job with Ford’s past reasonable accommodations. ' Three times Ford allowed Harris to telecommute on an as-needed basis (on flex time, no less). And three times Ford developed plans to improve her attendance. But all six efforts failed because Harris proved unable “to establish regular and consistent work hours” or “perform the core objectives of the job.” R. 60-3 at ¶ 3. The ADA does not give her a seventh try. Harris is not a “qualified individual” as a matter of law. 42 U.S.C. § 12111(8).
B
The EEOC sees it differently. It argues that three sources—(1) Harris’s own testimony, (2) other resale buyers’ telecommuting practices, and (3) technology— create a genuine dispute of fact as to whether regular on-site attendance is essential. But none does.
(1) Harris’s testimony. An employee’s unsupported testimony that she could perform her job functions from home *764does not preclude summary judgment, for it does not create a genuine dispute of fact. Neither the statute nor regulations nor EEOC guidance instructs courts to credit the employee’s opinion about what functions are essential. That’s because we do not “allow employees to define the essential functions of their positions based solely on their personal viewpoint and experience.” Mason, 357 F.3d at 1122. And for good reason: If we did, every failure-to-accommodate claim involving essential functions would go to trial because all employees who request their employer to exempt an essential function think they can work without that essential function.
In any event, Harris’s testimony does not add much. Harris testified that she used conference-call capabilities to perform a “vast majority” of her otherwise face-to-face interactions. R. 66-3 at ¶¶ 3-8. But she does not say that she could perform the vast majority of her work as effectively off-site, and the essential-job-function inquiry does not require employers to lower their standards by altering a job’s essential functions. See 29 C.F.R. § Pt. 1630, App. at 395-96 (portion titled “Section 1630.2(n) Essential Functions”). Harris’s testimony thus does not contradict the uniform record evidence that a resale buyer could not work from home on an unpredictable basis without lowering production standards. See id. Nor does Harris say that she could perform all of her duties from home; she indeed admits that four of her ten main duties had to be done at the worksite. R. 66-10' at 2. And Harris’s past failed telecommuting experiences put to rest any doubts as to whether she could effectively work from home — she couldn’t. R. 603 at ¶ 3; R. 60-7 at 2. The EEOC needs more to reach a jury.
(2) Other employees’ telecommuting schedules. The evidence of other buyers’ schedules likewise doesn’t do the trick. Unlike an employee’s own testimony, though, this consideration has support in the regulations, 29 C.F.R. § 1630.2(n)(3)(vii), and in our case law, Rorrer v. City of Stow, 743 F.3d 1025,1042 (6th Cir.2014). And unlike an employee’s own testimony, it makes sense to look at this kind of evidence: It reflects the employer’s judgment — which is not just what the employer says but also what the employer does. Picking up on this, the EEOC argues that because Ford allowed several other resale buyers to telecommute, working from the worksite must not have been essential.
On this record, we disagree. This argument might work if the other employees’ schedules were materially similar (say, unpredictably telecommuting three days per week). But Harris’s coworkers worked from home on materially different schedules: on one set day per week — no more, and sometimes less. The most any employee was even authorized to work from home was two days per week, and that employee actually telecommuted only one day per week. And critically, every telecommuter agreed in advance to come into work on their set telecommuting day if needed at the worksite. That’s a far cry from Harris, who: (i) requested up to four days per week; (ii) would not schedule the days in advance; and (iii) refused to come on-site if needed. None of these other employees’ more predictable and more limited schedules removed regular on-site attendance from the resale buyer’s job. They thus do not create a genuine issue of fact.
In addition to being legally and factually unsupported, the EEOC’s view here would cause practical harm to private employers. The ADA encourages — indeed, requires— employers to make reasonable accommodations for its employees, including allowing telecommuting under the proper cir*765cumstances. 42 U.S.C. § 12111(9)(B). But if the EEOC’s position carries the day, once an employer allows one person the ability to telecommute on a limited basis, it must allow all people with a disability the right to telecommute on an unpredictable basis up to 80% of the week (or else face trial). That’s 180-degrees backward. It encourages — indeed, requires — employers to shut down predictable and limited telecommuting as an accommodation for any employee. A “good deed would effectively ratchet up liability,” which “would undermine Congress’ stated purpose of eradicating discrimination against disabled persons.” Ameritech, 129 F.3d at 868 (citation omitted). The practical effect? Companies would tighten telecommuting policies to avoid liability, and countless employees who benefit from currently generous telecommuting policies would suffer. A protective tool becomes a weapon if used unwisely; and telecommuting should not become a weapon.
(3) Technology. Despite its commonsense charm, the EEOC’s appeal to technology ultimately fails to create a genuine fact issue. It is “self-evident,” the EEOC declares without citation to the record or any case law, that “technology has advanced” enough for employees to perform “at least some essential job functions” at home. Reply Br. 4; accord Dissent Op. at 776. In the abstract, no doubt, this is precisely right. E.g., Vande Zande v. Wis. Dep’t of Admin., 44 F.3d 538, 544 (7th Cir.1995) (recognizing as much). But technology changing in the abstract is not technology changing on this record. Our review of a district court’s summary-judgment ruling is confined to the record. And no record evidence — none—shows that a great technological shift has made this highly interactive job one that can be effectively performed at home. The proper case to credit advances in technology is one where the record evinces that advancement. There is no such evidence here.
In fact, the evidence here shows the opposite: technology has not changed so as to make regular in-person attendance marginal for this job. Ford uses “fairly limited” video conferencing and “tend[s] more towards audio conferencing.” R. 60-5 at 44-48. Harris also testified that she used email and her computer. These technologies — email, computers, telephone, and limited video conferencing — were .equally available when courts around the country uniformly held that on-site attendance is essential for interactive jobs. The extra-record changes in technology, like Harris’s testimony and her coworkers’ practice before it, therefore do not create a genuine issue of fact as to the essential nature of regularly and predictably attending work on-site. Summary judgment remains proper.
One more point, for clarification. None of this is to say that whatever the employer says is essential necessarily becomes essential. Contra Dissent Op. at 773-74; 775-76. Suppose, for instance, that a fire department regularly allows certain firefighters to refrain from driving fire trucks. But then the department denies the same accommodation to a firefighter with a known disability that prevents her from driving the trucks. A genuine fact issue might exist as to whether driving a fire truck is actually essential — it is contradicted by materially similar job practices. Cf. Rorrer, 743 F.3d at 1042; see also Solomon v. Vilsack, 763 F.3d 1, 12 (D.C.Cir.2014). Our ruling does not, in other words, require blind deference to the employer’s stated judgment. But it does require granting summary judgment where an employer’s judgment as to essential job functions— evidenced by the employer’s words, poli*766des, and practices and taking into account all relevant factors — is “job-related, uniformly-enforced, and consistent with business necessity.” Tate v. Farmland Indus., Inc., 268 F.3d 989, 993 (10th Cir. 2001). That aptly describes Ford’s judgment regarding regular and predictable on-site attendance for resale buyers. The district court accordingly properly granted summary judgment.
C
Our conclusion that Harris was unqualified for her position makes it unnecessary to consider whether Ford showed bad faith in the discussions to work out a reasonable accommodation while Harris was still employed. Even if Ford did not put sufficient effort into the “interactive process” of finding an accommodation, 29 C.F.R. § 1630.2(o )(3), “that failure is actionable only if it prevents identification of an appropriate accommodation for a qualified individual. ” Basden v. Prof l Transp., Inc., 714 F.3d 1034, 1039 (7th Cir.2013) (emphasis added); see Mason, 357 F.3d at 1124 n. 4. Courts thus need not consider this form of non-independent liability “if the employee fails to present evidence sufficient to reach the jury on the question of whether she was able to perform the essential functions of her job with an accommodation.” Basden, 714 F.3d at 1039. It suffices here to hold that any failure by Ford does not create liability because, as we just concluded, the EEOC did not produce such evidence.
But one more word on this: The record, in any event, uniformly shows that Ford did act in good faith “to initiate”— and maintain — “an informal, interactive process” with Harris. 29 C.F.R. § 1630.2(o )(3). It met with Harris to engage in an “interactive discussion, dialogue^] and opportunity to review various options that would meet both the needs of the business as well as [Harris’s] personal needs.” R. 66-10 at 2. It sought clarification on Harris’s telecommuting request (to which Harris reiterated that she was asking for the unpredictable “up to [four] days per week”). Id. at 3. It twice met with Harris and identified two alternative accommodations — moving Harris closer to the restroom and changing Harris to a position with more telecommuting opportunities. — even though it was not legally required to counteroffer, Jakubowski, 627 F.3d at 202-03. Contra Dissent Op. at 774-75, 778-81. And even after Harris rejected both counteroffers, Ford persisted that it was willing to “talk with [Harris] again if she identified] another accommodation” because Ford wanted “her to remain in the workplace.” R. 66-10 at 6. It was Harris’s turn to propose a reasonable accommodation to Ford, and she never did. Having failed to do so, she doesn’t get the chance to try again before a jury.
To sum up, the EEOC must prove that Harris is a “qualified individual,” which means she can perform the essential functions of a resale buyer with a reasonable accommodation. The record shows that Harris cannot regularly and predictably attend the workplace — an essential function, and a prerequisite to other essential functions — even with the past reasonable accommodations of telecommuting trials and specialized plans to improve her attendance. And Harris’s proposed unpredictable, ad hoc telecommuting schedule was not reasonable because it would have removed at least one essential function from her job. Harris is unqualified as- a matter of law, and the district court correctly granted summary judgment on this claim.
*767III
That conclusion goes some way to answering the next question: Did Ford retaliate against Harris for making a charge of discrimination? We hold that it did not.
The ADA separately prohibits companies like Ford from “discriminating] against any individual because such individual has ... made a charge ... under this chapter.” 42 U.S.C. § 12203(a). Discrimination here means retaliation — that “but for” an employee’s statutorily protected activity the employer would not have taken the “adverse employment action.” Univ. of Texas Sw. Med. Ctr. v. Nassar, — U.S. -, 133 S.Ct. 2517, 2533, 2535, 186 L.Ed.2d 503 (2013); see Lewis v. Humboldt Acquisition Coty., 681 F.3d 312, 318-19 (6th Cir.2012) (en banc). To assess these claims, we use the familiar McDonnell-Douglas burden-shifting framework. See Penny v. United Parcel Serv., 128 F.3d 408, 417 (6th Cir.1997). The plaintiff “must first establish, by a preponderance of the evidence, [her] ‘prima facie’ case.” St. Mary’s Honor Ctr. v. Hicks, 509 U.S. 502, 506, 113 S.Ct. 2742, 125 L.Ed.2d 407 (1993). If the plaintiff does so, the defendant has a burden of production to articulate a nondiscriminatory reason for its action. Id. at 507, 113 S.Ct. 2742. If the defendant meets its burden, the plaintiff must prove the given reason is pretext for retaliation. Id. at 515, 113 S.Ct. 2742.
Assume for now that the EEOC has met its prima facie case (but more on this later). The burden shifts to Ford, which has met it by producing evidence that it fired Harris because she was a poor performer. It offered undisputed evidence of back-to-back-to-back poor performance reviews, Harris’s lacking interpersonal skills, and Harris’s many absences, which in turn caused mistakes. And it offered evidence that Harris failed three specialized attendance plans before it terminated her. The burden shifts back to the EEOC to show pretext to prevail on its retaliation claim.
To demonstrate pretext, a plaintiff must show both that the employer’s proffered reason was not the real reason for its action, and that the employer’s real reason was unlawful. Hicks, 509 U.S. at 515, 113 S.Ct. 2742; see Reeves v. Sanderson Plumbing Products, Inc., 530 U.S. 133, 148, 120 S.Ct. 2097, 147 L.Ed.2d 105 (2000). To avoid summary judgment, then, the EEOC must present evidence from which a reasonable jury could find that poor performance was not the real reason that Ford terminated Harris, and that unlawful retaliation in fact was.
It trips over the first hurdle: No reasonable jury could find that Ford terminated Harris for a reason other than poor performance. Harris’s performance and interpersonal issues have been well documented. The EEOC indeed admits they existed. Suffice it here to say that, among other problems, Harris failed to update spreadsheets, complete her paperwork, schedule her training sessions, price items correctly, and finish her work on time. Her performance issues are why she ranked in the bottom 10% of her peer group before she made her charge.
The EEOC offers other evidence that, in its view, shows that Ford fired Harris because she filed a charge with the EEOC, not because of these performance issues. Timing is on the EEOC’s side: The mere four months between Harris’s charge and her discharge seems suspicious. But while this “gives us pause,” “temporal proximity cannot be the sole basis for finding pretext.” Donald v. Sybra, Inc., 667 F.3d 757, 763 (6th Cir.2012). So the EEOC needs more to reach a jury. It relies on three facts or inferences to create a genuine issue of material fact: *768(1) the meetings between Harris and her supervisor Gordon where Harris felt threatened; (2) the post-charge negative performance review; and (3) the alleged design of the post-charge performance-enhancing plan. Even when coupled with the timing, none suffices.
(1) Harris-Gordon Meetings. The meetings between Harris and Gordon do not create a genuine fact issue. To start, we doubt a reasonable jury could view these meetings — which Ford says were meant to help Harris, a worker with a long history of attendance and performance problems — as meant to hurt her. We “look at the facts as they appear to the person making the decision to terminate [the employee],” not at “the employee’s subjective [beliefs].” Kendrick v. Penske Transp. Servs., Inc., 220 F.3d 1220, 1231 (10th Cir.2000). Harris’s unexpressed “subjective skepticism regarding the truth of’ whether Gordon was actually trying to help her does not alone “raise a triable issue as to pretext.” Hedrick v. W. Resewe Care Sys., 355 F.3d 444, 462 (6th Cir.2004) (citing Mitchell v. Toledo Hosp., 964 F.2d 577, 584 (6th Cir.1992)). Plus, these kinds of meetings do “not constitute harassment simply because they cause the employee distress.” Keever v. City of Middletown, 145 F.3d 809, 813 (6th Cir. 1998).
But putting that aside, an even more fundamental point resolves this issue: The meetings involved only Gordon, a non decisionmaker. Actions by nondecisionmakers cannot alone prove pretext. Bush v. Dictaphone Corp., 161 F.3d 363, 369 (6th Cir.1998). Neither can decision-makers’ statements or actions outside of the decisionmaking process. Id.; see Rowan v. Lockheed Martin Energy Sys., Inc., 360 F.3d 544, 550 (6th Cir.2004). Both principles apply to Gordon. When Ford decided to terminate Harris, Gordon was on vacation. R. 60-2 at ¶ 26. And critically (and undisputedly), no one at Ford consulted with him or received a recommendation from him before making its termination decision. Id.; R. 60-15 at ¶ 8. So by definition, Gordon was a nondecisionmaker outside of this decisionmaking process. As Harris’s direct supervisor, he of course had an effect on her termination: He oversaw her overall poor performance and reported her failures during the performance-enhancing plan to his supervisors. See R. 66-23 at 26-28; R. 60-5 at 67-68. But we do not define “decision-maker” at such a high level of generality. The record uniformly shows that Gordon had no direct relation to the actual termination decision, and thus his allegedly harassing conduct cannot be imputed to Ford.
Nor can Gordon’s conduct in these meetings be imputed to Ford through the so-called “cat’s paw” theory. That theory would hold Ford liable if Gordon, motivated by retaliatory animus, intended to cause Harris’s termination and proximately caused the actual deeisionmakerg to terminate her. Staub v. Proctor Hosp., 562 U.S. 411, 131 S.Ct. 1186, 1194, 179 L.Ed.2d 144 (2011). In its five appellate briefs and in its brief below, the EEOC never so much as hinted that this theory might apply, which doubly forfeited the argument. United States v. Huntington Nat’l Bank, 574 F.3d 329, 332 (6th Cir.2009) (“[C]onclusory allegations and perfunctory statements, unaccompanied by citations or some effort at legal argument, do not meet th[e] standard” for raising an argument on appeal.); Estate of Quirk v. Comm’r, 928 F.2d 751, 757-58 (6th Cir. 1991) (“It is well-settled that, absent exceptional circumstances, a court of appeals will not consider an argument by an appellant that was not presented to or considered by the trial court.”). That was a wise *769move by the EEOC, for, among other reasons, no evidence shows a “direct relation between the injury asserted [termination] and the injurious conduct alleged [Gordon’s intimidation].” Staub, 131 S.Ct. at 1192; see Romans v. Michigan Dep’t of Human Servs., 668 F.3d 826, 836-37 (6th Cir.2012). The dissent would nevertheless apply this theory. Dissent Op. at 785-86. But that contravenes the rule that the parties (not judges) raise the arguments. And it expands this theory — fattens the cat, so to speak — far too much. This argument was forfeited, and, in any event, Gordon was no monkey, and Ford, not his cat. Staub, 131 S.Ct. at 1190 n. 1.
(2) Performance Review. The 2009 post-charge negative performance review fares no better. At first glance, this looks bad for Ford. Harris received her first “lower achiever” rating post-charge, and she received only “excellent plus” ratings before her charge. The EEOC stops there. But digging deeper — and looking at the whole record — reveals two reasons why no reasonable jury could find this low rating proof of pretext. For one, 2009 was the only year that Harris could have received the lower-achiever rating. Ford overhauled its ratings system that year for all employees, ditching the default “excellent plus” category (which 80% of workers received) in favor of a more accurate description of a worker’s performance. In Harris’s case, that meant “lower achiever” — the first and only time she could receive that rating. For two, the change in name did not change Harris’s low numerical ranking. In her only performance review after the charge, she ranked in the same percentile range as she did immediately before the charge: the bottom 10%. That’s not evidence of retaliation; that’s just poor performance — both before and after the charge.
(3) Performance-Enhancing Plan. Harris’s testimony that Ford designed the post-charge performance-enhancing plan to ensure her failure does not create a genuine dispute either. Any negative inference from this testimony is unreasonable because it comes unaccompanied by facts in the record, save Harris’s own speculation. And we do not accept the plaintiffs speculation where, as here, it does not create a “genuine” dispute of fact— that is, when it is “blatantly contradicted by the record.” Scott, 550 U.S. at 380, 127 S.Ct. 1769. The record shows that Harris failed two prior plans to improve her performance and attendance, similar to this one — -and both before she filed her charge. The record also shows that Harris failed to achieve any of the objectives identified in post-charge plan, R. 60-2 at ¶¶ 22-25 — not just the objective the EEOC says is evidence of retaliation (eliminating her backlog of paperwork, see Dissent Op. at 782-84.). And the record shows that ■ Ford used similar performance-enhancing plans for other employees who, like Harris, performed poorly. See, e.g., R. 60-4 at ¶ 17; R. 60-15 at ¶ 5. Harris’s testimony thus fails to create a genuine dispute of fact because it is “so utterly discredited by the record that no reasonable jury” could believe it. Scott, 550 U.S. at 380, 127'S.Ct. 1769.
The EEOC has failed to present evidence from which a reasonable jury could find that the real reason Ford terminated Harris was unlawful retaliation and not poor performance. Ford is entitled to judgment as a matter of law because the EEOC created at most “a weak issue of fact as to whether the employer’s reason was untrue and there [is] abundant and uncontroverted independent evidence that no discrimination had occurred.” Reeves, 530 U.S. at 148, 120 S.Ct. 2097. Lacking evidence that creates a genuine dispute of fact, the EEOC’s retaliation claim fails as *770a matter of law. The district court correctly granted summary judgment in Ford’s favor.
Now briefly back to the EEOC’s prima facie case, for it provides an alternate ground on which to grant summary judgment: The EEOC cannot establish but-for causation. To prevail on a retaliation claim, a plaintiff must “establish that his or her protected activity was a but-for cause of the alleged adverse action by the employer.” Nassar, 133 S.Ct. at 2534. Here that means that the EEOC must present evidence from which a reasonable jury could And that Ford would not have fired Harris if she had not made her charge.
For many of the same reasons discussed above, no reasonable jury could have found such causation here. In addition to Harris’s past failings, she admitted that she would hot be able to attend work on-site in a regular and predictable manner in the future. And this attendance was an essential element of her job. No reasonable jury could find that Ford — a for-profit corporation — would continue to pay an employee who failed to do her job well in the past, and who, by her own admission, could not perform the essential elements of her job in the future. The EEOC thus cannot demonstrate that Harris’s charge was the but-for cause of Ford’s decision to fire her, which means that Ford was entitled to summary judgment for that reason as well.
IV
Nearly thirty years later, it’s worth repeating: To overcome a well-supported motion for summary judgment, the non-moving party “must do more than simply show that there is some metaphysical doubt as to the material facts.” Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 586, 106 S.Ct. 1348, 89 L.Ed.2d 538 (1986). The EEOC has not done so here. We affirm.