ORDER

A great many Social Security disability cases turn on the question whether the individual seeking benefits retains the residual capacity to perform a job or jobs that exist in significant numbers in the national economy. Commonly, though not invariably, a vocational expert offers testimony about the existence of jobs that a person with limitations like those of the claimant can perform. Here, the administrative law judge heard no such evidence; he concluded on his own that claimant Daniel A. Borski, while impaired, could still do sedentary and light unskilled work. The Appeals Council of the Social Security Administration and the district court affirmed. We find, however, that on this particular record the ALJ erred in his decisions about which evidence could be used to evaluate Borski’s residual functional capacity, and which was to be excluded. In spite of the deferential standard of review that is appropriate for these cases, we conclude that this one must be remanded to the agency for further proceedings.

I

Borski suffers from a wide range of limitations. He filed his applications for *222Disability Insurance Benefits and Supplemental Security Income in September 1996, at the age of 44. As of that time, he had worked for twelve years as a bartender in his father’s tavern; before that, he had been an unskilled laborer. He quit the latter job because of debilitating pain in his back and knees—pain that was documented in medical records that spanned more than twenty-five years. Among other things, he had back surgery in 1978, surgery on a fractured knee in 1982, and an effusion to his left knee in 1994. Medical examinations in 1996 and 1997 revealed articular spurring in his knees and thoracic kyphosis (a deformity of the spine). According to Dr. Daniel Romond, Borski’s treating orthopedist, these conditions made it impossible for Borski to do heavy work, but they did not exclude light work as long as it did not require “full time standing and walking or sitting.”

Other doctors came to similar conclusions. Dr. Stephen Alt oversaw a functional capacity assessment of Borski in April 1997, in conjunction with a group called ERGOS Work Recovery, Inc. Dr. Alt’s evaluation report concluded that Borski could perform light work with frequent position changes. The evaluation report indicated that Borski had trouble with prolonged periods of standing or walking, but that he could tolerate two-hour periods with occasional breaks. The report also noted that Borski attempted a stooping activity during the evaluation, but that he declined to try others because he was afraid he could not get up again from a stooped position. The report reflected this problem in a section labeled “Stooping/Bending” in which the box for “occasional” was checked and the word “Bending” was circled. It is unclear whether this notation meant to indicate that Borski could “Stoop” all of the time, or none of the time.

Dr. A. Neil Johnson, an internist who examined Borski in December 1996, found moderate crepitus (cracking sounds at the joints) in Borski’s neck, back, and knees. He too noted that Borski had trouble getting on and off the examining table, walking heel-to-toe, and raising his legs without pain; on that occasion Borski could not squat or hop. Looking at matters from another perspective, Dr. Charles Barnes, a psychologist, evaluated Borski’s cognitive abilities in May 1997. He concluded that despite Borski’s completion of the 11th grade, he had only 4th grade spelling skills, 5th grade math skills, 6th grade reading skills, and a full scale I.Q. of 78. Dr. Barnes concluded that Borski was not a good candidate for jobs requiring significant retraining or classroom time.

Borski himself testified that he could not remain on his feet or sit with his legs down for more than 20 or 25 minutes, and that he could no longer lift more than 25 pounds. He said that he could not, crouch, bend, or stoop. Friends or relatives had to help him with every-day activities such as yard work and lifting heavy bags, but he still shopped for groceries, fished once or twice a month, ran errands, talked on the telephone, and played cards. Borski’s neighbor, Lisa Nelson, corroborated his account, noting that Borski often complained of pain, walked to the grocery store with difficulty, and had trouble bending for groceries and carrying heavy bags. Nelson also reported that Borski could not sit through an entire bingo or card game without having to move around or to leave and go to bed.

Borski also proffered evidence from Stephen Weigert, whom Borski offered as a vocational expert, but the ALJ rejected Weigert’s opinion. Weigert’s report, which was based on the medical evidence and a personal interview with Borski, concluded that Borski needed to alternate be*223tween standing and sitting positions more frequently than was possible in typical jobs in either the “light work” category or the “sedentary” category. Weigert concluded that in light of these positional needs, coupled with Borski’s age, limited education, and low intelligence, there was not a significant number of either sedentary or light work unskilled jobs in the national economy that Borski could perform. Noting that Weigert had mistakenly thought that Borski could not read, the ALJ rejected this opinion. The ALJ also thought that Weigert had overestimated Borski’s need for positional changes.

The ALJ also rejected Borski’s own account of his physical limitations. The ALJ believed that Borski’s every-day activities suggested greater mobility; he attached weight to the fact that Borski controlled his pain with over-the-counter medications only; he noted that Borski had not required physical therapy since 1994 (when he could sit for two to three hours and stand for eight to nine); and he found that Borski’s job as a bartender had been far more demanding than Borski later claimed. The ALJ did not call an alternative vocational expert to consider whether a person with the physical limitations the ALJ had chosen to credit would be able to perform work in the national economy; instead, the ALJ consulted the Medical-Vocational Guidelines (known as the “grid”) in 20 C.F.R. Part 404, Subpart P, Appendix 2, and decided on his own that Borski could work. The Commissioner’s final decision, after the Appeals Council chose not to disturb the ALJ’s conclusion, was to deny Borski’s two claims.

II

Our review of this decision is deferential: we ask only if the Commissioner’s decision was supported by substantial evidence. 42 U.S.C. § 405(g); Richardson v. Perales, 402 U.S. 389, 399-401, 91 S.Ct. 1420, 28 L.Ed.2d 842 (1971). Nevertheless, we do not function as a rubber-stamp for the Commissioner, and if the ALJ failed to consider all relevant evidence, the result cannot stand. Clifford v. Apfel, 227 F.3d 863, 871 (7th Cir. 2000). Here, the ALJ was required to follow the familiar five-step process for determining whether Borski was disabled for either purpose, asking (1) whether the claimant was not presently employed or engaged in substantial gainful activity, (2) whether his impairment or combination of impairments was severe, (3) whether he met any of the impairments on the “list” in 20 C.F.R. Part 404, Subpart P, Appendix 1 (ie. was he entitled to a conclusive presumption of disability), (4) whether he was unable to perform his past relevant work, and (5) whether he was unable to perform any other work within the economy. 20 C.F.R. § 416.920; see Stevenson v. Chater, 105 F.3d 1151, 1154 (7th Cir. 1997).

Only the fifth of those steps is at issue here; the Commissioner concedes that Borski meets the preceding four. At step five, the Commissioner bears the burden of proving that there are sufficient jobs in the national economy for a hypothetical person with the same impairments that the claimant has. Knight v. Chater, 55 F.3d 309, 313 (7th Cir. 1995); Haddock v. Apfel, 196 F.3d 1084, 1088 (10th Cir. 1999). There are two ways to do that. The regulations provide a “grid,” which the ALJ is permitted to use if the applicant’s exertional capacity, age, education, and past work experience fit the requirements of a rule within the grid. See Haddock, 196 F.3d at 1088. But the grid may be used only if the claimant’s individual characteristics fit precisely within the criteria; in that subset of cases, it offers a convenient short-cut for the ALJ. Otherwise, however, more particularized evidence about the jobs that *224would be available for the applicant is necessary. Without such evidence, the record is incomplete, and it is impossible for this court to know one way or the other if the ALJ’s decision on step five might have been supported by substantial evidence. Compare Zurawski v. Halter, 245 F.3d 881, 889 (7th Cir. 2001) (ALJ must be able to refer to evidence in the record that leads logically to her conclusion).

Here, the ALJ used the grid, and Borski challenges the propriety of his doing so. Borski points to three limitations that were not adequately reflected in the grid: (1) his complete inability to stoop, (2) his need for frequent positional changes, and (3) his low intelligence. Before considering these, we must highlight one ambiguity in the record. Much turns here on what it means to be able to do something “frequently,” or “occasionally,” and what the circle around the word “bending” meant. Neither party discusses who should bear the burden of these uncertainties: the Commissioner, or Borski. Perhaps because the Commissioner bears the burden in general at step five, both parties have agreed here that she has this burden as well. This may be correct; on the other hand, the Commissioner may have arguments to the contrary. If she did, however, she did not make them, and we thus evaluate the case on the assumption that ambiguities in these respects go against the Commissioner. With that in mind, we turn to Borski’s arguments.

Ability to stoop and to change position is required for both sedentary and light work positions, which speak of “occasional” stooping (meaning stooping for less than one-third of the workday). See SSR 83-10; Lauer v. Apfel, 169 F.3d 489, 492 (7th Cir. 1999). It is very hard to find any affirmative evidence in this record, however, that Borski could stoop “occasionally.” He himself testified that he could not stoop at all, but we put that to one side because the ALJ made a negative credibility determination with respect to his testimony. Dr. Romond said that Borski should “avoid” stooping, which is a far cry from saying that he was able to stoop up to a third of the time. Dr. Alt’s evidence is similarly inconclusive: on the portion of the evaluation overview labeled “Stooping/Bending,” the box for “occasional” is checked and the word “Bending” is circled, seemingly to avoid including the word “Stooping” in the recommendation. This makes sense if one recalls that Borski refused during the examination to perform activities in low height postures, after he attempted one activity requiring stooping. In our view, a decision to disbelieve Borski, plus inferences from the limitations in the two doctors’ reports, simply does not add up to substantial evidence that Borski could perform at the level required for these jobs.

Borski’s need for positional changes (moving from standing to sitting and back again) was also insufficiently addressed. There is a total lack of evidence to support the proposition that Borski could remain in a sitting or a standing position long enough to perform either sedentary or light work. Sedentary work requires more or less the same exertional level as light work. The difference between the two relates to position: sedentary work requires “frequent” (ie. up to two-thirds of the workday) sitting and “occasional” standing, while light work requires “occasional” sitting and “frequent” standing. The Social Security Administration has recognized that someone who needs to go regularly from one position to the other cannot do either kind of work unless the need to change positions can be “accommodated by scheduled breaks and a lunch period.” SSR 96-9p. Where scheduled breaks are not enough, the applicant *225needs a so-called “sit/stand option.” That takes the case out of the grid and requires the input of a vocational expert. Id.; see SSR 83-12; see also Peterson v. Chater, 96 F.3d 1015, 1016 (7th Cir. 1996); Jesurum v. See’y of Health & Human Servs., 48 F.3d 114, 120 (3d Cir. 1995) (collecting cases).

The ALJ concluded that Borski did not require a “sit/stand option” and thus turned to the grid. For support, he relied on two pieces of evidence. First, he looked at the notes from Borski’s 1994 physical therapy sessions, which indicated that Borski could sit for two to three hours a day and stand for eight or nine. But that was old evidence, and the ALJ did not indicate how (if at all) he was taking into account the degenerative nature of Borski’s condition. See, e.g., Clifford, 227 F.3d at 870-71 (ALJ should have considered whether increasing pain over a sixteen-month period implied a worsening condition). Second, he looked at Dr. Alt’s 1997 work evaluation, which observed that Borski could tolerate two hours of standing and walking “with occasional brief non-work breaks.” But this hardly supports the proposition that Borski could work without those breaks. It is not even clear what a “non-work break” is, but it is fair to assume that a break from standing would entail sitting or lying down. Importantly, we do not know what the ALJ thought it meant, and thus we have no way of evaluating whether Dr. Alt’s report can serve as evidence in support of the ALJ’s conclusion.

Also troublesome is the ALJ’s failure to discuss an apparently contradictory finding in the 1997 work evaluation that Borski could sit and stand occasionally with “frequent” position changes. The terms “frequent” and “occasional” have a special meaning in the context of work capacity, and Dr. Alt and the evaluators from ER-GOS Work Recovery, Inc. presumably knew this, as the evaluation report included those special definitions. It was incumbent upon the ALJ to discuss evidence that was contrary to the Commissioner’s position, see Godbey v. Apfel, 238 F.3d 803, 808 (7th Cir. 2000), and he did not do so here.

These flaws are enough to require a remand to the ALJ for further proceedings. We thus do not need to consider whether the ALJ took adequate account of Borski’s intellectual capacity, though we note that the ALJ was aware of this point and concluded that this was not a severe impairment. We also decline Borski’s invitation to second-guess the ALJ’s credibility determination with respect to Borski’s own testimony, even though some of the support the ALJ cited for his ultimate conclusion is questionable. (For example, the ALJ thought that Borski was taking only over-the-counter pain medication, whereas the record shows that he was taking Ibuprofen 800, which requires a prescription. It is also unclear how Borski’s day-to-day activities undermine his claimed impairments.) Finally, we do not regard the ALJ’s failure to discuss Nelson’s testimony as a significant problem; as the Commissioner points out, an ALJ does not need to provide a written evaluation of every piece of evidence in the case. See Diaz v. Chater, 55 F.3d 300, 308 (7th Cir. 1995).

For the reasons given, we Vacate the district court’s judgment and Remand this case to that court with instructions to remand it to the agency for further proceedings consistent with this order.