OPINION
BRUGGINK, Judge.This is an action brought under both the Fair Labor Standards Act, 29 U.S.C. §§ 201-219 (2000) (“FLSA”), and the Federal Employee Pay Act, 5 U.S.C. §§ 5541-5550a (2000) (“FEPA”). The action was transferred to this judge on September 16, 2004. In dispute is the amount of overtime pay, if any, due plaintiffs arising out of work done to complete the 2000 census. The court earlier granted in part and denied in part the government’s motion for summary judgment. Christofferson v. United States, 64 Fed.Cl. 316 (2005).1 Although the suit is brought on behalf of former Census employees across the United States, the present opinion deals exclusively with the Concord, California Local Census Office and, more particularly, the overtime eligibility of Field Operations Supervisors (“FOS,” used hereinafter for both singular and plural forms) within that office.2 The matter is pending after trial held in San Francisco from May 9-13, 2005. For the reasons explained below, we hold that the FOS positions were improperly designated as exempt under the FLSA.
BACKGROUND
The Bureau of the Census has numerous on-going functions. It is best known, however, for the work it does conducting the constitutionally-mandated decennial census. The effort is monumental, but episodic. The Bureau employs a permanent workforce of ap*69proximately 7,000 employees. The permanent workforce, however, is inadequate to perform the Bureau’s data gathering function during each decennial census. Vast numbers of temporary employees are hired to complete this task. During the entire course of the 2000 census effort, for example, more than 960,000 temporary employees were hired. The maximum temporary field staff employment at any one time peaked at over 500,000 workers. Plaintiffs are among these former temporary employees of the Census Bureau hired to collect information for the 2000 census.
The organizational structure put in place to collect data during each decennial census is also temporary. The 2000 census commenced with the activation of the Bureau’s existing Regional Census Centers (“RCC,” hereinafter used for both the singular and plural forms). Each RCC became responsible for a number of temporary Local Census Offices (“LCO,” used hereinafter for both singular and plural forms). Initially, in 1998, 130 LCO were opened. In 1999, 390 additional LCO were added. Among those opened in 1999 was the Concord, California LCO, which became the subject of the trial. It was one of 38 LCO assigned to the Seattle RCC. It opened on July 1, 1999, and closed on September 30, 2000. The Concord LCO was responsible for collecting census data in Contra Costa County, located in the eastern San Francisco Bay area.
The 2000 census, at least insofar as it engaged the Bureau’s regional structure, was conducted in three distinct phases, beginning in 1998 and ending in 2000. The first phase, called the “Address List Development” phase, consisted of pre-census preparatory operations, primarily address verification. The second phase, which took place over two months in the spring of 2000, was the “Before Census Day Enumeration” phase. It consisted of twelve operations, primarily intended to ensure that census forms were distributed.3
The third and final phase, “After Census Day Enumeration and Coverage Improvement,” was the most labor intensive in terms of the use of temporary employees. It took place between April and August 2000. During this phase, the LCO were responsible for filling in the numerous gaps in the responses to the census forms, as well as verifying data. As such, it required significant direct contact with residents.
The third phase consisted of four operations. Two involved the use of FOS — “Non-Response Followup” (“NRFU”) and “Coverage Improvement Followup” (“CIFU”). NRFU was a labor-intensive operation scheduled to be completed within nine weeks. Its purpose was to collect census data from households that did not return completed census forms. Data collection during this phase was done primarily by “enumerating” each household — sending an employee called an Enumerator to interview someone in the household. During the NRFU operation, the Concord LCO employed 2,522 individuals in temporary positions. After NRFU was completed, CIFU commenced. During CIFU, a select number of households were enumerated, including those that had not been identified previously or were identified as vacant during NRFU. CIFU was a smaller operation than NRFU. It lasted approximately two weeks. The FOS in the Concord LCO worked mostly during these two operations.
The government presented a number of permanent Census Bureau employee witnesses. Gail Leithauser works in the Census headquarters office in Washington, D.C., as Assistant Division Chief for Geography and Data Collection Activities. Mark Holdredge is Chief of the Human Resources Division. Michael Weiler is Special Assistant to the Associate Director for Field Operations. They explained in general the Bureau’s approach to creating a decentralized temporary field management structure for each decennial census and the typical organizational structure within the various LCO, including Concord.
The Concord LCO management team also testified for the government. The head of each LCO was the Local Census Office Manager. In Concord, this was Delores Brooks, *70who testified as a rebuttal witness. Reporting to her were three Assistant Managers: one for recruitment of workers, one for administration, and one for field operations (“AMFO”). The AMFO supervised the FOS and other field personnel. In Concord, this position was held by Timi Tumbaga, who also testified at trial. The management team in the LCO consisted of these four individuals, along with an automation technician. Each position within the management team was a “time limited appointment,” meaning it lasted throughout the decennial census effort, was not in the excepted service, and could last for more than one year. These officials were all treated as exempt from FLSA overtime eligibility.
Below the management team, the other employees in the LCO were hired as-needed under “temporary” appointments. This temporary designation meant that these individuals were hired for less than one year. There were three types of such employees who fell under the supervision of the AMFO. The lowest paid and most numerous were the Enumerators. On the Enumerators fell the brunt of the collection work. During NRFU they filled out census questionnaires based on interviews with individuals at non-responsive households — households that had not yet completed and returned census forms. The “enumeration” typically took place when most residents were likely to be home, ie., in the early morning or early evening hours on weekdays and all day on the weekends. The workday began when the Enumerator left home to begin enumerating and ended when he or she returned home. Some Enumerators were full-time employees, but most worked on a part-time basis. This position was designated non-exempt by the Office of Personnel Management (“OPM”), meaning the employees were eligible for FLSA overtime pay.
Enumerators reported to a single Crew Leader (“CL,” used hereinafter for both singular and plural forms). CL supervised a group of Enumerators within an assigned geographic area called a Crew Leader District. Although a CL (with minor exceptions) always had a Crew Leader District, the configuration of the district itself varied between operations, and the same CL would not necessarily be assigned to the same district, nor be assigned the same Enumerators.
During NRFU, CL were responsible for supervising the work of Enumerators. This consisted primarily of ensuring that non-responding households were contacted and interviewed. The CL were given assignments in terms of books unique to a geographic area showing which households had not mailed in a census form. The CL met daily with their assigned Enumerators, usually in the morning or evening, to distribute work assignments, to review and collect completed questionnaires and time sheets, and to supply more questionnaires and supplies as needed. The CL also attempted to answer questions and resolve problems. CL were expected to keep their teams (ranging from 10 to 19 Enumerators) directed, supplied, and motivated. Most CL had an assistant, an Enumerator assigned to them and given the title Crew Leader Assistant (“CLA,” used hereinafter for both singular and plural forms), to help in these duties. Both CL and CLA were treated as FLSA non-exempt.
CL, in turn, reported to a FOS. A single FOS might supervise a number of CL and, indirectly, their respective Enumerators. The function of the FOS, according to the position description, was to “oversee! ] the activities of the field personnel.” They were classified as FLSA exempt under the executive exemption and thus not entitled to overtime pay. Although they were not considered part of the LCO management team, Mr. Weiler referred to the FOS as an “on-the-ground manager.” He summarized their role as follows:
We needed to have a manager covering a defined geographic area for a defined operation, to make sure that whatever the particular operation was, was being carried out according to procedures, that it was staying on schedule, that we were staying fully staffed, and that we were providing the workers!,]first line of supervisors, the crew leaders!,] with technical support, personnel support, and other operational support.
Tr. at 38.
It was contemplated, in other words, that the key functions of a FOS — developing a *71group of trainees, selecting CL, monitoring operations, preparing reports, and scheduling — could be performed for more than one type of operation, with varying groups of employees, and in different locations. As Mr. Weiler explained, the FOS position was therefore critical to a decentralized management structure. Although the various elements of work, location, and personnel changed, the constant was a single FOS over a single district, at least with respect to the operations which called for FOS.
Consistent with this vision, the number of FOS positions at the Concord office varied during the census from zero to eight, depending on the operation. During one of the phase two operations, “Undeliverable as Addressed” (“UAA”), a single FOS was utilized. She had CL and Enumerators reporting to her. The two operations with the greatest need for FOS were NRFU and CIFU. Some operations did not require significant field work, meaning there was no need for FOS.
Because of the large number of temporary employees needed for the operations, the use of the term “hired” or “selected” does not fit well in connection with Enumerators, CL, or FOS. The process, as characterized by former FOS Dale Kirkland, was closer to requisitioning. As Mr. Holdredge explained, the Bureau attempted to identify a huge pool of persons who met minimum qualifications. The nationwide recruiting goal was to have 3,000,000 persons apply for slots. These persons would then be interviewed by telephone or in person to inquire about their availability and to further determine their qualifications. Those who satisfied these tests were invited to training sessions. Applicants were paid during training. Before training commenced, applicants were administered an oath and put on the Bureau’s payroll, at a minimum for the duration of training. Neither the completion of training nor the execution of a form SF-50-B, “Notification of Personnel Action,” although essential steps in the hiring process, meant that a person was actually employed in the field. It simply meant that a person could report for work, if summoned. The SF-50-B showed a “not to exceed” (“NTE”) date, typically of six-week duration. It was not uncommon for plaintiffs to have their NTE dates extended on multiple SF-50-Bs, some of which might also include a promotion from an Enumerator or CL slot to a FOS slot. As Mr. Weiler explained, this requisitioning and re-staffing process occurred over and over again. It would be prompted by loss of personnel or by the shift to a new operation, as the short NTE dates meant a whole new crew of individuals would need to be activated. In Concord, the biggest push to identify a pool of qualified individuals occurred immediately leading up to the NFRU operation. An effort was made to obtain a stable of qualified and trained individuals who lived in all parts of the Concord census area because the Bureau preferred to have workers, particularly Enumerators, conduct their activity in their own neighborhoods.
Training typically occurred immediately prior to commencement of an operation. It was normally conducted by the immediate supervisor. For example, AMFO Tumbaga conducted FOS training. FOS, in turn, trained their CL, and CL trained their Enumerators. The most basic training was for the position of Enumerator. Once trained for a given slot, persons could be activated on an as-needed basis. Training was also “cascading,” meaning CL were trained not only on the CL function, but on the Enumerator function, and FOS received training for all three positions. Typically, when FOS trained their CL, the LCO assigned for training a group of persons slightly larger than the number whom it was anticipated were needed. The FOS was then expected, during training, to identify the seven or eight persons she wanted to use as CL. Not designating someone as a CL in effect meant that, if they wished to continue during the operation itself, their pay would drop to that of an Enumerator. Some of the Enumerators were designated as CLA, however, and, as there was attrition, they were often moved up to the CL slot. In the midst of an operation, FOS were given authority to make such promotions to the CL slot by “battlefield promotions” — simply by designation and minimal paperwork.
The training methods employed understandably left little to chance. There were *72training manuals for all the field positions. The term repeatedly used to characterize all training was “verbatim” — the training instructor literally read every word of the appropriate training manual to students. Questions to test comprehension were also scripted. If students did not understand a point, it would be covered again from the text.
To consider the process leading up to the NRFU operation, immediately after a group of Enumerators had been trained, the central office matched them with certain geographic areas within the LCO. The basic geographic building block was the census block. It is not entirely clear, even after trial, how or when a block was identified, although the system was borrowed from the 1990 census. (It was not, however, based on zip codes.) Groups of these census blocks were, in turn, assembled into larger temporary groupings, unique to each operation, called assignment areas. The number of census blocks contained within an assignment area was based upon the projected ability of a single Enumerator to complete the enumeration of a set number of households within a week. Urban assignment areas might, because of the greater density of its population, consist of a single block of apartment buildings; rural assignment areas might consist of many square miles of territory. There were more households to be enumerated in NRFU than in CIFU. Because the size of assignment areas was dependant on the workload contained within their boundaries, they were unique to each operation.
These assignment areas were then grouped into Crew Leader Districts and assigned to a single CL. The CL Districts were, in turn, grouped into FOS Districts, each with an assigned FOS. CL Districts and FOS Districts, like assignment areas, were also the product of the workload contained within them, or, more accurately, the number of assignment areas and thus the number of subordinates needing supervision. For each operation the Bureau had a standard ratio regarding the number of subordinates to a single CL and FOS. Generally, a CL supervised 12 to 19 Enumerators. Each FOS, in turn, supervised approximately eight CL. The result, at least during the NRFU operation, was that one FOS could be a first or second line supervisor over as many as 150 people.
Because the workload between operations changed, the number and geography of assignment areas varied between operations. The new assignment areas were then grouped into new CL and FOS Districts for each operation. This generally resulted in different numbers of CL and FOS Districts between operations. For example, during the UAA operation of phase two, there was only one FOS District covering the entire LCO area. The number of FOS Districts changed in Concord from eight during the NRFU operation to five during the CIFU operation.
FOS, therefore, were only hired as-needed for each operation. For each operation where FOS were needed, the Bureau hired a number of FOS equal to the number of FOS Districts it created for the operation. Each FOS was appointed with a definite expiration date corresponding with the end of the operation. If the succeeding operation did not call for a FOS, or fewer FOS, the unneeded FOS were released. If the following operation did call for FOS, the Bureau could extend the NTE date beyond the original operation’s end date. This happened from NRFU to CIFU.
If a FOS’ NTE date was extended, however, the FOS was assigned a new FOS District made up of a different agglomeration of census blocks. CL were also given a new group of blocks. The CL and Enumerator personnel would also change between the various operations. Thus, the only permanent feature of this geographic pyramid was the census block. All other components — assignment areas, CL Districts, and FOS Districts — were unique to each operation and fluctuated between them.
None of these interim employees had a regularly set schedule. They were paid weekly for actual hours worked. They did not earn leave, and their work hours could be altered to accommodate peak workloads. CL were expected to meet daily with their Enumerators. During NRFU, the LCO had identified non-responding households by *73block and by address. Typically, Enumerators were assigned a book containing assignment area information. At a morning meeting, CL would collect the Enumerators’ work, resupply the them, and answer any questions they might have. There were target quotas. An Enumerator was expected, by whatever reasonable means possible, to meet a householder so that the census form could be completed. It was not an easy task in many cases. Householders could be uncooperative; they might be hard to find; addresses might be stale; neighborhoods might be unsafe; there might be guard dogs protecting the home. Enumerators generally took pride in their work. They felt a strong sense of ownership regarding their books and were territorial about completing the assignments in them.
The CL, in addition to handing out assignments and returning completed forms to the LCO, were to keep Enumerators motivated and directed, to troubleshoot, and to maintain quality control. CL were expected to help Enumerators through access problems. For example, to facilitate entry into gated communities, CL would approach law enforcement, apartment, or community association managers. Because Enumerators often had other*, non-Census jobs, they typically worked mornings or evenings, or both. That was also the best time to catch working householders. The result was that CL had to make time available to meet with Enumerators, individually or as a group, as best they could. In addition, they needed to be accessible throughout the day or evening to answer questions by telephone.
The relationship between FOS and CL was similar to that of CL and Enumerator. FOS kept up the assignment flow to the various CL teams and returned forms to the LCO. FOS were also responsible for motivating and directing CL, with whom they tried to meet frequently. FOS would also troubleshoot for CL and maintain quality control.
The court heard from a number of persons who served as Enumerators, CL, or FOS regarding their work. Some of them served in more than one of these positions. Their testimony focused on the extent of the independence and discretion of FOS and their ability to recommend personnel actions. As a group, they were a fascinating shoe of America. They came from all walks of life and brought with them a wide array of experience, age, and skills.
Doug Lewis, a farmer in Contra Costa County, testified for the government. He worked as both a CL and a FOS during NRFU. He explained that he felt he had the authority to take work away from a CL with whom he was not happy, but did not have authority to fire. On occasion he had to improvise ways to get census data under unusual circumstances, such as from residents of offshore islands. He estimated that he spent three quarters or more of his time on supervision, primarily in planning work for his teams.
Mr. Lewis’ testimony is consistent with that of Michael Burns, the Special Assistant to the Regional Director in Seattle. Mr. Burns helped prepare for the 2000 census by working on a “dress rehearsal,” which was intended to test the procedures in place. At that time he was Assistant Regional Census Manager, but he was later promoted to Deputy Regional Director, in charge of the census. He visited the Concord LCO twice during the 2000 census. He explained that FOS were expected to exercise discretion in implementing their duties because of the unpredictability inherent in tracking down respondents.4 He also testified that, during the census, authority for appointment and corrective action with respect to temporary employees rested with FOS. Appointment authority did not extend to deciding that more personnel were needed and making the decision to add staff, however. It merely meant that FOS could select from training pools the individuals they wanted to work with. There was no documentation of this delegation offered to the court, and Burns’ knowledge of termination authority was indirect. As most FOS testified, they did not understand that *74they actually had final authority to terminate.5
Mr. Burns was one of several witnesses who referred to the concept of battlefield promotions. The shortness of the operations often did not leave sufficient time to go through separate training and selection in the event a CL quit or was released. Battlefield promotions referenced the authority of a FOS to promote Enumerators to an open CL position without having to go through the typical requisition process. The agency’s administrative manuals confirm this practice, in which a FOS would simply designate an Enumerator from one of the existing crews to replace a CL, thereby entitling that person to higher pay.
Mr. Bums also explained a practice called a “blitz,” in which resources were concentrated across CL Districts, but within a given FOS District. As we consider below, there was conflicting testimony about whether the FOS had an active role in deciding to go into “blitz” mode.
Sandra Wilson served as an Enumerator, CL, and FOS. She became a FOS in time to work on CIFU and NRFU. She picked her own CIFU crews but inherited her NRFU crews. Her description of responsibilities suggested that she viewed her role as involving a degree of discretion and autonomy. She fired one Enumerator for misconduct and approved the decision of a CL to withhold work from an Enumerator. “Basically, any decisions they made, I had to approve.” Tr. at 310. This is somewhat at odds with the characterizations of others, including that of Mr. Weiler, who testified that a CL could terminate an Enumerator, needing only input from the FOS.6 She confirmed, however, the change in geographic boundaries in the transition from NRFU to CIFU, as well as the change in CL from one operation to another.
Other witnesses also testified for the government. Valerie Vergara has a degree in genetics from the University of California, Davis. She served briefly with the pay status of a CL, although her work primarily was that of CLA to Tom Sheridan. Her FOS was Luita Lynch. Ms. Vergara testified regarding the day-to-day interactions between FOS, CL, CLA, and Enumerators. Robert Liu served as a CL for four months in 2000. He worked under FOS Kenneth Owens. He testified that he had the authority to select his own Enumerators and CLA. On more than one occasion Mr. Liu recommended to Mr. Owens that an Enumerator be fired.
Testifying for plaintiffs was Conrad Dandridge. He is a supervisor with the Transportation Safety Administration. Mr. Dandridge began as an Enumerator during the first phase operation of distributing forms to problem addresses. He quickly became a CLA and, during NRFU and CIFU, served as a FOS. He also stayed on during closing activities in August 2000, being paid as an Enumerator. Kenneth Owens, who worked “off and on” during 1999 and 2000, was trained as an auditor. He served as an Enumerator, CL, FOS, and as a SPOS. A SPOS functioned in much the same way as a FOS, overseeing CL and their teams in “special places,” such as homeless shelters, marinas, group homes, and soup kitchens. Mr. Owens worked as a FOS during the NRFU and CIFU operations through July 2000. He closed out his employment with a two-week appointment in the middle of August 2000, as an Enumerator, helping with clean-up. Kathleen Lander served as a CL from the beginning of April 2000 until the end of July 2000. She graduated with honors from the University of Southern California with a degree in journalism. Ms. Lander has had a career as a writer and has led community volunteer organizations. As a CL, she had 10 to 12 employees. Kenneth Owens was her FOS.
*75Elizabeth Bernstrom also testified. She has a bachelor’s degree in liberal studies and has done graduate studies in English and social work. Like most of the CL and FOS witnesses, she has also been in supervisory positions in the past. Her Census employment began in June 2000 when she was hired to work for the Assistant Director for Recruitment. She later served as a FOS during most of the NRFU operation. She was not re-hired for the CIFU operation. Ms. Bernstrom stated that most problems encountered as a FOS were covered in the manuals given to them by the Bureau. Those that weren’t, such as understaffing, were handled by Ms. Tumbaga.
James MacDonald served as a FOS during the NRFU operation in the 2000 census. He has owned a print shop and served as an ambulance driver and a ski patrolman. Because he owned a boat at a local marina, he served as a SPOS for a brief effort to provide form coverage to marinas in the area. His employment ended prior to the end of NRFU. Mr. MacDonald believes he was fired for trying to generate a pool of people to apply to be Census workers.7 He felt he had no authority to hire, promote, or fire, and was, in his view, too busy to do so. He simply avoided performance problems by not assigning work to problem employees.
Gail Roche held a number of positions during the 2000 census. Like other witnesses, she had a diverse and interesting background, having been trained as a geographic information systems specialist and having worked in the insurance business. She was initially hired as an Enumerator and then served as a CLA, a CL, and a FOS. She was present during the UAA operation, as well as during NRFU and CIFU. Her geographic assignments shifted between operations. Ms. Roche felt that 90% of her time was supervisory, but that she had virtually no discretion in performing her work. The only area in which she felt she could exercise some independent judgment was in moving an Enumerator to an assistant CL position. She could not think of a single problem that had not been anticipated to some degree by the operations manuals. She did not interview employees; she was simply sent personnel. Anything unusual had to be discussed with Ms. Tumbaga. On the occasion in which her teams participated in a blitz, it was decreed by the LCO.
One of the government’s rebuttal witnesses, Kenneth Holmes, served as a CL under Doug Lewis. He testified that he recommended the termination of two Enumerators, one for inefficiency and another for falsifying forms. He consulted Mr. Lewis, who agreed that the two should be terminated. As to the process of termination, Mr. Holmes was unable to recall with certainty whether he employed the standard forms included in the supervisory handbook. He recalls that he took work away from both individuals and did not permit them to claim further hours.
DISCUSSION
The FLSA provides a framework for calculating overtime pay for federal employees. 29 U.S.C. § 201; 5 U.S.C. § 5542. The relevant regulations appear at 5 C.F.R. § 551.201-.205 (2005). The parties disagree as to whether FOS were properly classified under FLSA as exempt executive employees. The government has the burden of proving such an exemption. Corning Glass Works v. Brennan, 417 U.S. 188, 196-97, 94 S.Ct. 2223, 41 L.Ed.2d 1 (1974); 5 C.F.R. § 551.202(c). Further, the “exemption criteria must be narrowly construed to apply to only those employees who are clearly within the terms and spirit of the exemption.” 5 C.F.R. § 551.202(b).8
OPM regulations provide standards to determine whether an employee qualifies for exemption as an executive employee: “An executive employee is a supervisor or manager who manages a Federal agency or any subdivision thereof (including the lowest recognized organization unit with a continuing function) and customarily and regularly di*76rects the work of subordinate employees and [whose work] meets [the primary duty test].” 5 C.F.R. § 551.205.
The primary duty test is met if the employee:
(1) Has authority to make personnel changes that include, but are not limited to, selecting, removing, advancing in pay, or promoting subordinate employees, or has authority to suggest or recommend such actions with particular consideration given to these suggestions and recommendations; and
(2) Customarily and regularly exercises discretion and independent judgment in such activities as work planning and organization; work assignment, direction, review, and evaluation; and other aspects of management of subordinates, including personnel administration.
5 C.F.R. § 551.205(a). These characteristics of the job must also constitute the primary aspect of the position. “[A]n employee’s primary duty may be the one he or she engages in during a majority of his or her work time; or, it may be the duty which is most important or significant, even though it occupies less than half of the employee’s work time.” Adams v. United States, 27 Fed.Cl. 5, 13 (1992), rev’d on other grounds, 178 F.3d 1306 (Fed.Cir.1998); see also 5 C.F.R. § 551.104 (2005). Accordingly, defendant must show both that the FOS managed recognized organizational units with a continuing function and that their work satisfied the two-part primary duty test.
Lowest Recognized Organization Unit with a Continuing Function
The threshold inquiry is whether FOS managed a “recognized organizational unit with a continuing function.” See 5 C.F.R. § 551.205. A snapshot taken of the organizational chart of an LCO at the height of the NRFU process shows the FOS District as a distinct box at the lowest end of the hierarchy. The government contends that this district meets the requirement for a recognized organizational unit with a continuing function because the FOS District’s function continues for the duration of a particular operation. Persons holding the FOS position, therefore, according to the government, satisfy this element of the executive exemption. Plaintiffs disagree.
In Adams v. United States, 350 F.3d 1216 (Fed.Cir.2003), the Federal Circuit explained that a recognized organizational unit is “an established and defined organizational entity with regularly assigned employees.” 350 F.3d at 1223 (quoting OPM Letter 551-7 § B(l)(b) (1975)); see also 5 C.F.R. § 551.104 (codifying this standard). Drawing on the OPM standard, it further explained:
This requirement distinguishes supervisors who are responsible for planning and accomplishing a continuing workload from “leaders” who head temporary groups formed to perform a special assignment of limited duration, or who direct the work of other employees assigned to a project but do not exercise full supervision over such employees. Leaders of this nature do not qualify for exemption as executive employees.
Adams, 350 F.3d at 1223; see also 5 C.F.R. § 551.104.
At the trial level in Adams, we were faced with identifying the lowest recognized organizational unit with respect to certain operations of the United States Border Patrol. Adams v. United States, 44 Fed.Cl. 772 (1999). We came to the conclusion that the shift supervised by the “Watch Commander,” also known as a FOS, constituted the lowest recognized organizational unit. We rejected the plaintiffs’ argument that the cutoff point was at the higher organizational level of the Border Patrol Station. We also rejected the government’s argument that the relevant cutoff point was below the FOS level, at the point at which Senior Border Patrol Agents supervised groups of border patrol agents.
There are obvious similarities between the FOS shift in Adams and the FOS District in the case at bar other than the similarity of the title of the supervisory figure. There, FOS also supervised groups of field agents on something less than a permanent basis, although the same supervisors and employees continued throughout the shift. 44 Fed. Cl. at 778. It is not insignificant, however, that the shifts in Adams were longer and *77more uniform in nature. They were of a six-month to one-year duration, whereas here the districts lasted three weeks to two months at the most. A close examination of the FOS position here, therefore, is required before concluding that it is comparable to the FOS position in Adams.
Gail Leithauser explained that the FOS District was created as the “third lowest” level of geographic organization9 and was to be a “constant geographic entity to help manage and control operations for a census field operation.” Tr. at 136. While this observation might be correct with respect to one of the operations which involved supervision of numerous personnel in the field, in fact the vast majority of the Concord LCO operations did not involve activation of the FOS District.
Mark Holdrege testified concerning the creation of the temporary regional census office structures, including the development of position descriptions and classification for FLSA exemption status. He participated in the headquarters process that led to decisions in these respects. He defended the agency’s decision to treat the FOS position as FLSA exempt. While recognizing that the transition from operation to operation could involve changes to personnel, geography, and activity, he explained that, from his perspective, what was important was that the organization chart looked the same from operation to operation. Once again, this was only literally true during those few operations in which FOS were utilized. During the bulk of operations the structure existed only on paper.
What Mr. Holdrege meant was that the FOS position, whether for NRFU or CIFU, for example, was in charge of a FOS District, which in turn was a geographic assemblage of CL Districts. His focus, in short, was on the similarity of “structure of supervision and management” across the different operations. Tr. at 467. When asked whether the lowest recognized unit varied over time or between operations, he said, “no.” Id. at 471. In his view, the FOS District was the lowest recognized unit irrespective of the fact that the particular district might cease to exist, the CL might change, the Enumerators might change, the assemblage of CL Districts might change, and the particular function would change. For Mr. Holdrege, it was the continuing similarity in structure that mattered: “[I]n terms of what their supervisory managerial role would be, I’m seeing that [as] fairly stable____” Tr. at 472.
We disagree with Mr. Holdrege’s conclusion that there was meaningful stability in the role of the FOS. His focus on a theoretical unit as drawn on an organization chart ignores the reality of alterations in function, geography, and personnel. It also ignores the fact that some of the FOS operations involved no CL Districts at all. In effect, the agency concluded these positions met the lowest unit test because, when the FOS District/CL District structure was mobilized, it involved a FOS in a supervisory position over CL.
The FOS Districts were not permanent in any sense. They were unique to a particular operation with a lifespan only as long as the operation itself. There were approximately 22 field operations for the 2000 Census. Approximately 12 of these were conducted by the LCO. While a total of eight operations utilized FOS, only four of the Concord LCO’s operations utilized FOS — Update/Leave (“U/ L”), UAA, NRFU, and CIFU. These operations lasted from three weeks to two months. Only NRFU and CIFU were consecutive. All four operations involved a realignment of the function, geography, and manpower of the district. The U/L and UAA operations each involved only one FOS, whose district spanned the entire LCO. NRFU, which began approximately one month after UAA ended, employed eight FOS, each with their own district. CIFU only employed five FOS, with districts different than NRFU. In short, the FOS District morphed according to the need of the moment, if it was needed at all.
Nor is it appropriate to ignore the fact that the FOS here were not performing the same activities when they moved from operation to operation. In Adams, the function of the shift — to patrol the station’s assigned area— persisted between assignments. Even when new shifts were formed, their function and *78geography did not change, merely the personnel did. Here, however, the shift to a different operation meant a new function and identity. In addition to changes in personnel, the shape and number of FOS Districts changed with each operation, if they were utilized at all. They had no continuing identity. In short, we are persuaded that the differences between the role of the FOS in Adams and that of the FOS here are not simply ones of degree, but also are differences in kind.10
Instead, it is our view that the distinction drawn by OPM in 5 C.F.R. § 551.104 between, on the one hand, units with truly ongoing functions, and on the other, those which are temporary in nature, could have been crafted with these Census positions in mind. In no sense did the FOS District remain “an established and defined organizational entity,” see 5 C.F.R. § 551.104, nor was it staffed with “regularly assigned employees,” see id., nor did it have a “continuing function,” see 5 C.F.R. § 551.205. The task changed; the leadership changed; the geography changed; and the employees changed. The FOS therefore were “‘leaders’ who head[ed] temporary groups formed to perform a special assignment of limited duration,” within the contemplation of the Adams court. 350 F.3d at 1223 (quoting 5 C.F.R. § 551.104). This deficiency is sufficient to make the FOS non-exempt. Other reasons exist, however, as we discuss below.
Primary Duty Test — Ability to Take Personnel Actions
The first element of the primary duty test requires that an executive employee have actual authority to make certain personnel changes or, at a minimum, the authority to suggest or recommend those actions with particular consideration given to those suggestions and recommendations. The types of personnel changes the employee must have authority to make or recommend “include, but are not limited to, selecting, removing, advancing in pay, or promoting subordinate employees.” 5 C.F.R. § 551.205(a)(1). Despite the use of the word “or,” this court has held that an exempt employee’s authority to make or recommend personnel actions must pertain to both the selection or removal and advancing in pay or promoting of subordinate employees. Angelo v. United States, 57 Fed.Cl. 100, 114-15 (2003) (recognizing that OPM’s interpretation of FLSA must be consistent with the Department of Labor’s interpretation absent some justification for treating federal employees differently from non-federal employees).
Regarding selection or removal, the government makes much of the FOS’ ability to select their CL and to recommend discharge of CL or Enumerators. The position description reflects that FOS “make[] recommendations for selection and hiring of experienced and qualified crew leaders and Enumerators.” It is true that FOS had a role in selecting their CL. This was typically done as part of the CL training process. In addition, FOS had a role in recommending retention of CL for later operations.
If the FOS position meets the letter of this element of the primary duty test, it does not meet its spirit for several reasons. As former FOS Dale Kirkland, involved in recruiting, stated: “[Y]ou didn’t hire anyone in the Census Bureau; you requisitioned someone.” Tr. at 699. The pool of individuals from which FOS “selected” their CL was in fact chosen by the central office. If there were more applicants for the CL slot than available positions, then the role of the FOS was more accurately to de-select one or more trainees and persuade them to stay on as Enumerators or CLA. The process involved virtually no discrimination or opportunity for *79assessment. According to former CL Valerie Vergara, a common tactic at CL training was for FOS to tell prospective CL that the position was a lot of work so that some would voluntarily remove themselves from consideration. It was no different for Enumerator hiring. As Nance Rogers, a CL during NRFU, explained, at the end of Enumerator training she simply explained the hours required of an Enumerator and asked who was still interested in the position, and because the number of “yes’s” typically equaled the available slots, she swore them in.
Subsequent hiring to fill gaps was even less intentional, driven instead by the tyranny of necessity. The central office, particularly toward the end of the NRFU operation, was taking dramatic measures to keep up a pool of willing field workers. FOS could ask for names and phone numbers of applicants in their territory. The pool of applicants, moreover, was driven more by geographic concerns than anything else. FOS or others would simply start calling people at the top of the list and work their way down until they obtained the needed reinforcements. Perhaps Mr. Kirkland put it best: “[The Bureau] ramped up from zero to half a million people in a very short period of time. You did not have the luxury of testing and selecting the people overall for a period of time and experience with them.” Tr. at 702.
The same can be said of the FOS’ removal authority. It is true, as most former FOS testified, that they felt they could, and in fact did on rare occasions, end the employment of unsatisfactory workers. They did that, however, by taking the assignment books away from the CL or Enumerator and not assigning further work to them. If the employee did not work, he or she did not get paid. The evidence that formal means of termination were used was sketchy at best. What generally occurred, instead, was that the employee’s appointment was allowed to expire. It is apparent from the conflicting descriptions of the government witnesses that there was confusion about where the ultimate responsibility for firing lay. Even Ms. Brooks, the LCO Manager, explained that she did not have authority to terminate employees for cause. That required approval by the Seattle RCC, although the recommendation of the FOS supervising the employee was very important.
FOS also could elevate an Enumerator or CLA to a CL position with virtually no involvement by the LCO using the so-called “battlefield promotion.” If a CL left, a CLA was asked to move into the slot. At no point in the process was there the luxury of interviews, testing, or careful deliberation over applicants, however. The need to get the job done using whoever was available dictated hiring and promotion considerations.
Overall, it was apparent from the testimony as a whole that any discretion the FOS had in making selection, removal, and promotion recommendations was insignificant. This was not because their roles were insignificant, but because the apparatus for putting FOS in charge of their crews and maintaining them at the required strength simply did not leave room for deliberation or meaningful discretion. We conclude, therefore, that the first element of the primary duty test is not met.
Primary Duty Test — Independent Judgment and Discretion
A second element of the primary duty test is that the employee “customarily and regularly exercises discretion and independent judgment in such activities as work planning and organization; work assignment, direction, review, and evaluation; and other aspects of management of subordinates, including personnel administration.” 5 C.F.R. § 551.205(a)(2). “Discretion and independent judgment” is defined at 5 C.F.R. § 551.104, where the following limitations are expressed:
The work must be sufficiently complex and varied so as to customarily and regularly require discretion and independent judgment in determining the approaches and techniques to be used, and in evaluating results. This precludes exempting an employee who performs work primarily requiring skill in applying standardized techniques or knowledge of established procedures, precedents, or other guidelines which specifically govern the employee’s action.
*80Once again, the caution at the end of this paragraph could have been drafted with the FOS position in mind. Two manuals were used by FOS throughout the 2000 census. The first was a manual titled the “Field Supervisory Census Employee Handbook” (“Handbook”) which was used for both the FOS and CL positions. It was handed out at training and contained all the information necessary for FOS and CL to understand their function and operation. It was distinct from the operational manuals, which were specific to a particular census field activity. The Handbook contained such things as personnel instructions, safety considerations, payroll information, and other administrative guidance.11
In addition to the Handbook, the Bureau prepared a single “Field Operations Manual” (“Manual”), hundreds of pages in length, which had specific chapters instructing the AMFO, the FOS, and the CL on how to conduct particular operations, such as NRFU, CIFU, and special places operations, as well as myriad instructions of more general application. The level of detail is prodigious. The expectation was that anyone employed in field operations should be able to instruct themselves or subordinates fully on tasks and approved methods from the manuals.
The Handbook and Manual attempted to anticipate virtually any conceivable contingency. As Mr. Holmes, a government witness testified, all administrative and personnel aspects of the FOS and CL jobs were anticipated in the Handbook and other manuals. Problems, cautions, and potential solutions were presented in an outlined, step-by-step fashion. For example, there are ten pages in the Handbook dealing with safety issues and a separate page of instructions in the event an Enumerator is bitten by a dog.12 The Handbook also re-affirmed that the training sessions to be given by supervisors were scripted: “You will not have to create or ad-lib any parts of the training.” Handbook at 8-4. Mr. Weiler agreed that the FOS “didn’t have any discretion with respect to how they delivered ... training.” Tr. at 123.
The FOS who testified made it clear that they felt virtually no sense of discretion or independent judgment. Elizabeth Bernstrom explained that the Handbook and Manual answered most problems, but that to the extent problems arose that were not addressed there, they were simply not solved. She felt no discretion to come up with creative solutions. Instead, creativity came from the LCO. She was ordered to operate in a “blitz” mode, for example, even when she felt it was counterproductive. There was constant pressure to complete a huge amount of work in a limited time, while subject to the instruction that no overtime could be authorized. Rather than trying to come up with creative solutions, she simply followed the LCO’s instructions to the letter, even if it meant deviating from the Handbook and Manual. As she expressed it, she “dared not” exercise independence or discretion. Tr. at 783. She characterized herself as a glorified gopher. She felt she was a “little puppet for ... the AMFO.” Id.
Their subordinates also had this impression of FOS. One CL, Nance Rogers, testified, “We would go to the book,” referring to the CL manual for each operation. Tr. at 733. She did not recall problems that could not be answered by reference to the manual. The person who served as her FOS during most of the NRFU operation was of little help to her. Another CL, Elizabethe Walton, testified to the same effect. Her FOS, Doug Lewis, would give her advice and suggestions for maintaining productivity, but she got the impression he did not exercise independent judgment or discretion:
At the times that I talked to my supervisor, he frequently referred to his guidelines and his supervisors, and I did not get *81the impression that he had a lot of jurisdiction. He would tell me that things were told to him by other people that he had to do.
Tr. at 710.
Although the government argues that FOS maintained control over the techniques to use in completing the work, in fact, the discretion was severely restricted. The evidence, for example, that FOS could direct “blitzes”— special concentrations of workers within the FOS District — was conflicted at best. Ms. Brooks, the Concord LCO Manager, made it clear that the decision to engage in a district-wide blitz was made by her and the AMFO in consultation with the Seattle RCC. This was because sending CL and Enumerators across CL District lines implicated budget concerns, which had to be approved by the RCC. Inevitably blitzes meant a loss of productivity in other areas and an increase in mileage claims. We are persuaded that FOS had no independent authority either to call a “blitz” or allocate manpower across CL District lines.
There was also evidence that FOS could not independently order “pairing” — the assignment of two Enumerators to one interview as a safety measure. Ms. Walton, a CL, on one occasion asked for permission to address a safety issue by having two Enumerators handle an assignment together. Doug Lewis, her FOS, told her he was not authorized to permit the practice.
Finally, the evidence strongly points to the conclusion that FOS could not authorize overtime. Ms. Tumbaga, the AMFO, testified that FOS or CL could, in emergency circumstances, approve the use of overtime. This is directly contrary to the Handbook, however, which states, “Employees are not allowed to work more than 8 hours in a day or 40 hours in a week without receiving advance written or verbal approval from the LCO Manager.” Handbook at 8-8. It is also inconsistent with testimony of virtually every other witness, including Ms. Brooks. Though Ms. Brooks is cited by the Handbook as the approving authority for overtime, she testified that she had to seek approval from the RCC. The government offered no documentary evidence that overtime was ever approved, much less by a FOS. We believe the practice to have been that the Seattle RCC kept complete control over the approval of overtime, and in fact all costs.13 Use of overtime was not within the discretion of FOS.14
There is a fundamental irony in the government’s contention that FOS exercised discretion and independent judgment. The Bureau’s plan for the field operation aspects of the census was designed from top to bottom to minimize discretion. The training, Handbook, Manual, and command structure were all intended to anticipate every conceivable eventuality and create uniformity of execution. The need to approach data gathering in this fashion is more than understandable, but it is completely at odds with the notion that the supervisors were expected to improvise on the fly. Their discretion was intentionally limited so that the results would be uniform, accurate, and fast.
To the extent FOS were expected to exercise discretion or independent judgment, it was virtually exclusive to the details of handing out assignments or placement of workers. Although some locations created special problems (such as gated communities, homeless shelters, or group homes), even these anomalies were anticipated in the manuals. Without denigrating in any sense the need for leadership or for organizational and motivational skills in the FOS position, it is accurate to say that innovation was discouraged. Unforeseen situations had to be raised to the AMFO level for advice, sometimes even higher. The need for uniformity in the collection of data simply precluded any meaningful amount of discretion or independent judg*82ment in FOS. In sum, the second element of the primary duty test is also not met.
CONCLUSION
The “exemption criteria must be narrowly construed to apply to only those employees who are clearly within the terms and spirit of the exemption,” 5 C.F.R. § 551.202(b). The facts are clear that FOS did not manage a recognized organizational unit with a continuing function, did not have the authority to take or recommend personnel actions, and did not customarily and regularly exercise independent judgment and discretion. Any one of these findings would be sufficient to hold that they were improperly classified as exempt.
We thus conclude that the government has not met its burden of proving that FOS in the Concord LCO were exempt from overtime pay. The parties are directed to consult and prepare a joint status report by September 9, 2005, proposing further post-trial proceedings.
. In that opinion, the court dismissed plaintiffs' FLSA claims for overtime pay for hours worked in excess of eight per day. Plaintiffs’ request for equitable tolling of the FLSA’s limitations period was also denied. Plaintiffs’ FLSA claims for hours worked in excess of forty in a single workweek were preserved, as were their claims regarding whether Field Operations Supervisors and Special Places Operations Supervisors ("SPOS”) were properly classified as non-exempt under the FLSA.
. At trial, plaintiffs’ counsel informed the court that the prior ruling on the statute of limitations barred the claims of SPOS Kenneth Owens, the only plaintiff employed as a SPOS in the Concord LCO. This opinion, therefore, does not discuss the exemption status of the SPOS position.
. "Census Day” was April 1, 2000. The data collected by the Census Bureau was supposed to give an accurate count, or "snapshot,” of the population on this date.
. This was supported by Ms. Tumbaga, who testified that she expected FOS to exercise independent judgment about such things as access, staffing, overcoming hostility, and working through community-based groups.
. Ms. Tumbaga testified that she had to sign off on any terminations proposed by the FOS, but she relied on their recommendations. One exception related to a proposed termination of a CL at a time when she was exploring the termination of a FOS. In fact, she testified that "We didn't terminate. Appointments expired.” Tr. at 370.
. Gail Leithauser clarified the seeming inconsistency by explaining that CL could, de facto, terminate an Enumerator by simply not giving any more work.
. Consistent with Ms. Tumbaga's testimony, however, his personnel file shows that his appointment was simply allowed to expire. Pis.' Ex. 5.
. "If there is reasonable doubt as to whether an employee meets the criteria for exemption, the employee should be designated FLSA nonexempt.” 5 C.F.R. § 551.202(d).
. The CL District and assignment area would be the lower two levels.
. In its post-trial brief, for the first time, the government also argues that the portion of 5 C.F.R. § 551.104 dealing with temporary groups necessarily refers only to groups that do not carry out primary functions of the employing entity. It alleges that a bona fide temporary group requires no separate supervision, because supervisors already exist within the permanent structure of the organization: "[W]hen the specified mission or project is complete, the group disappears as a group, leaving the established structure intact." Def.'s Post-trial Br. at 13-14. We find no such distinction in the language of the regulation. Furthermore, it is odd that the government would make this argument with respect to the Census Bureau. By that standard, the lowest permanent structure within the Bureau is the RCC. The LCO themselves were only temporary.
. Other subjects included personal property damage claims, appointing staff, releasing staff, recommending staff to other positions, and completion of reporting forms.
. A typical bit of advice appears on page 4-5: "If you are approached by a peculiar-looking animal, cross over to the other side of the street. Stay away from the animal. If you are directly approached or unable to get away from a suspicious dog, avoid eye-contact. Dogs see eye-contact as threatening or even a challenge.”
. According to Ms. Brooks, the LCO did not have a set pool of money to use as needed. As a result, any decision that would have increased costs had to be approved by the RCC. This severely hampered the use of blitzes and the approval of overtime, as either would have required RCC approval.
. Nance Rogers testified that she "had lied on every one of my [time sheets for Enumerators]” to the extent she did not reflect overtime for her crew. Tr. at 740.