ORDER
William Banks applied for Disability Insurance and Supplemental Security Income benefits at the age of fifty-one, alleging that he could no longer work because *930of debilitating back and neck pain stemming from a 1995 car accident. An administrative law judge (ALJ) concluded that Banks was not “disabled” within the meaning of the Social Security Act because he could perform his past job as a beer salesperson — a job modified by his former employer to accommodate his back injury. The Appeals Council denied review, and Banks sought review in the district court pursuant to 42 U.S.C. § 405(g). A magistrate judge, handling the case with the parties’ consent, granted summary judgment in favor of the Commissioner. Because we are unable to conclude from the ALJ’s cursory analysis that substantial evidence supports the Commissioner’s decision, we vacate and remand for further proceedings.
In 1986 Banks began working as a salesperson for a Chicago beer distributor. His duties included stocking heavy beer cases and kegs at retail liquor stores and sporting venues, setting up promotional displays, writing orders, and supervising delivery personnel. Nine years later, in May 1995, Banks was involved in a car accident. A few days after the accident he began experiencing severe lower back and neck pain. Magnetic resonance imaging (“MRI”) tests in July 1995 revealed that Banks had “severe advanced degenerative” changes in his lumbar spine (lower back) and cervical spine (neck), “completely collapsed” spaces between several vertebrae, and narrowed spaces between others. Banks received physical therapy, heat treatment, and nerve-block injections in his neck.
Banks returned to his beer distributor job in November 1995 but after two weeks reported to his doctor, neurologist Dr. Jeffrey Kramer, that he was having a “terrible” time working because of his back pain. Dr. Kramer advised Banks to stop working and referred him to an orthopedic surgeon, Dr. F. Todd Wetzel. Dr. Wetzel confirmed the diagnosis of degenerative disc disease and prescribed pain medication and more physical therapy. The following month Dr. Wetzel reported that Banks was “progressing slowly” and could return to work four hours per day, so long as he refrained from lifting more than ten pounds.
Shortly thereafter Banks returned to work part-time, servicing customer accounts but no longer stocking their stores. By mid-April, Dr. Wetzel reported that Banks had begun working eight hours per day, although he still experienced “constant” back and neck pain. Dr. Wetzel believed that Banks was able to continue working at a light to medium level while undergoing physical therapy. The following month, however, Banks returned to Dr. Wetzel complaining that he had aggravated his back at work while reaching for a telephone. Dr. Wetzel expressed concern that if Banks was as “unstable regarding a mechanical problem” as it appeared, he “may have difficulty returning to work.” In August 1996, at Dr. Wetzel’s urging, Banks underwent a three-level discography, which turned out to be negative. Dr. Wetzel concluded that “[a]t this point we have effectively cleared a structural cause” for Banks’ pain symptoms.
Banks continued working full-time until January 6, 1997, when he was laid off along with eleven other employees. He has been unemployed ever since. A few months after his layoff, Banks began a new round of physical therapy at a rehabilitation clinic. The following month he told his treating physician, Dr. Stacy Harris, that his lower back pain had improved. But in late July 1997 Banks returned to Dr. Harris complaining of neck pain and weakness in his left arm. Dr. Harris found that Banks was unable to raise his right arm above 60 degrees, which he said *931“raises concern for cervical nerve root compression with neurological compromise.” He referred Banks back to Dr. Wetzel who, after an examination in August 1997, suspected a massive rotator cuff lesion. The record does not reflect what, if any, treatment Banks received after seeing Dr. Wetzel. At the hearing Banks explained that, without health insurance, he could not afford further treatment.
Banks filed for disability benefits on August 14, 1997, alleging an onset date of January 6, 1997, the date that he was laid off from his sales job. Agency doctors examined Banks in connection with his application. Dr. Biele, a general practitioner, reported that Banks exhibited a full range of motion in his cervical spine with tenderness in the paravertebral muscle. The range of motion in Banks’ lumbar spine was limited and his straight-leg test was positive. All motion was painful. He was able to squat and walk on his heels and toes with pain. Dr. Biele noted that Banks’ fine motor skills were somewhat impaired, e.g., Banks had trouble unbuttoning a button and picking up a pencil, but that his gross motor strength was intact. New x-rays were taken which showed “moderately advanced osteoarthritis” and disc degeneration.
The agency initially, and upon reconsideration, denied Banks’ application for benefits. Banks then requested and was granted a hearing before an ALJ, during which he testified that as a result of his 1995 car accident, he suffered from constant pain in his lower back and neck that radiated down his left arm. He said that he could not carry more than ten or fifteen pounds, walk more than three blocks, or sit for more than thirty minutes without a break, though after five minutes he experienced discomfort. As for his daily activities, Banks testified that he did not do an “awful lot.” He said that he was able to drive only short distances and that household chores such as sweeping and vacuuming were “very uncomfortable” for him. At least once a day when his back pain became unbearable, Banks said that he had to lay on the floor for 15 to 20 minutes until the pain subsided.
Dr. Giovanna Raimondi, a neurologist, testified at the ALJ’s request. Based upon her review of Banks’ medical records, Dr. Raimondi opined that although Banks had degenerative disc disease, his records revealed no nerve compression, decreased muscle strength, or sensation loss. She stated that Banks’ limitations primarily concerned “discomfort of the spine” that affected “[h]is ability to get up and down when needed, maybe for two minutes, once every hour or every two hours ... [a]nd limitation of lifting weight.” She concluded that Banks could do light work without prolonged standing or sitting.
Following the hearing, the ALJ requested that another neurologist, Dr. Richard Brannegan, evaluate Banks. Dr. Branne-gan performed an electromyogram to determine whether Banks’ back injury was affecting his lower extremities. The elec-tromyogram did not demonstrate any “lumbrosacral nerve injury.” Like Banks’ other doctors, Dr. Brannegan opined that his pain was caused by degenerative disc disease. Although “no abnormal findings are demonstrable in the lower extremities,” the doctor reported that “most any movement that requires prolonged sitting, standing, bending, etc. will produce or accentuate the low back pain.” Specifically, he estimated that Banks could stand or walk without interruption for thirty minutes but for two to three hours at most in an eight-hour day; could sit for only five to ten minutes at a time, though the total amount of sitting depended upon the type of chair; could occasionally stoop, crouch, *932or kneel; and could not reach with his left arm.
Applying the familiar five-step analysis, see 20 C.F.R. § 416.920, the ALJ concluded that Banks was not disabled. The ALJ found that Banks satisfied the first two steps because he suffered from severe back and neck impairments and had not engaged in substantial gainful activity since January 1997. But the ALJ concluded that his severe impairments did not meet or equal any impairment listed in 20 C.F.R. Pt. 404, Subpt. P, App. 1, and thus Banks did not automatically qualify for benefits under the third step of the inquiry. At the fourth step, the ALJ found that Banks had the residual functional capacity to do light and sedentary work “provided that he does not have to perform either activity for extended periods of time,” which, the ALJ concluded, enabled Banks to return to his previous beer sales job, “as that job was modified to require no lifting.” In so concluding, the ALJ found incredible Banks’ allegations that he was unable to work. The ALJ did not go on to determine whether, at the fifth step, a significant number of other jobs existed that Banks could perform in the national economy.
We will affirm a decision denying benefits so long as it is supported by substantial evidence, a deferential but not entirely uncritical standard. Steele v. Barnhart, 290 F.3d 936, 940 (7th Cir.2002). Where the Commissioner denies benefits, at a minimum the ALJ must “build an accurate and logical bridge from the evidence to her conclusion.” Dixon v. Massanari, 270 F.3d 1171, 1176 (7th Cir.2001). The decision cannot stand if it lacks evidentiary support or “is so poorly articulated as to prevent meaningful review.” Steele, 290 F.3d at 940.
On appeal Banks challenges the ALJ’s decision at step four, primarily arguing that the decision lacks the detail needed to permit meaningful review. Banks first argues that the ALJ provided no explanation for discrediting his testimony regarding the effect of his symptoms on his ability to work. We generally will not overturn an ALJ’s credibility determination unless it was “patently wrong.” Zurawski v. Halter, 245 F.3d 881, 887 (7th Cir.2001). But we will reverse where the ALJ provides no explanation for the determination. Id. at 887; Brindisi v. Barnhart, 315 F.3d 783, 787-88 (7th Cir.2003); Steele, 290 F.3d at 941-42. In evaluating the credibility of statements supporting a Social Security application, an ALJ must comply with Social Security Ruling 96-7p. Brindisi, 315 F.3d at 787. According to that ruling, an ALJ’s evaluation of a claimant’s credibility must contain “specific reasons” and “must be sufficiently specific to make clear to the individual and to any subsequent reviewers the weight the adjudicator gave to the individual’s statements and the reasons for that weight.” SSR 96-7p. The ruling specifically cautions that “[i]t is not sufficient to make a conclusory statement that ‘the individual’s allegations have been considered’ or that ‘the allegations are (or are not) credible.’ ” SSR 96-7p.
Yet that is precisely what the ALJ did here. The ALJ discounted Banks’ allegations in a single, conclusory statement: “The claimant’s allegations of inability to work were not credible because of significant inconsistencies on the record as a whole.” Nowhere in the decision did the ALJ identify these “significant” inconsistencies or explain how they influenced his conclusion.
Social Security Ruling 96-7p also requires that an ALJ carefully consider “the intensity, persistence, and functionally limiting effects of the [claimant’s] symptoms,” when evaluating the credibility of a claimant’s complaints of disabling pain. The *933ALJ’s decision does not contain any such analysis. The ALJ merely stated, without elaboration, that because no nerve damage had been detected, Banks’ pain was “local” and “insufficient” to preclude him from working:
Dr. Raimondi, a medical expert who testified at the hearing, felt that, based on an MRI, the claimant’s injuries did not involve impingement or pressure upon the spinal nerves. Hence, the claimant’s pain is local and insufficient to exclude him from light and sedentary work provided that he does not have to perform either activity for extended periods of time.
Notably, Dr. Raimondi’s testimony does not suggest that Banks’ complaints of severe pain, although local, were incredible. Dr. Raimondi testified that although she was “puzzled” about the cause of Banks’ pain, she did not “doubt the pain that he’s describing, that it is in there.” And, even though the medical evidence revealed no nerve impingement, it was undisputed that Banks suffered from a serious degenerative condition in his spine. In any event, ALJs may not limit themselves to an observation that the severity of the pain cannot be demonstrated by medical evidence. As we have noted, an ALJ must consider “all avenues presented that relate to pain,” including the “claimant’s prior work record information and observations by treating physicians, examining physicians, and third parties, ... the nature and intensity of claimant’s pain, precipitation and aggravating factors, dosage and effectiveness of any pain medications, other treatment for the relief of pain, functional restrictions, and the claimant’s daily activities.” Zurawski, 245 F.3d at 887; see also SSR 96-7p. Here, the ALJ did not recite the requisite factors, much less identify and explain the inconsistencies he perceived between Banks’ daily activities, his complaints of pain, and the medical evidence.
The Commissioner defends the ALJ’s perfunctory determination by arguing that the ALJ implicitly considered the factors for evaluating symptoms set forth in 96-7p. But, as we recently reiterated, “nothing in Social Security Ruling 96-7p suggests that the reasons for a credibility finding may be implied,” rather, the ALJ must “specify the reasons for his finding so that the applicant and subsequent reviewers will have a fair sense of the weight given to the applicant’s testimony.” Golembiewski v. Barnhart, 322 F.3d 912, 916 (7th Cir.2003) (per curiam) (emphasis in original); see also Steele, 290 F.3d at 941 (“[Rjegardless whether there is enough evidence in the record to support the ALJ’s decision, principles of administrative law require the ALJ to rationally articulate the grounds for her decision and confine our review to the reasons supplied by the ALJ.”). Nor are we persuaded that the evidence identified by the Commissioner supports the ALJ’s credibility determination, even if the ALJ had implicitly considered it. The Commissioner first contends that the ALJ’s mention that Banks’ daily activities-driving and walking short distances, light cooking and cleaning, and going to the movies-supported the adverse credibility finding. But such minimal daily activities do not necessarily undermine or contradict a claim of disabling pain, particularly where, as here, the claimant also testified that his activities were punctuated by frequent rest periods. See Zurawski, 245 F.3d at 887.
The Commissioner next notes that in the body of the decision the ALJ observed that in December 1997 Banks was taking only over-the-counter medication for his pain, that physical therapy had not been particularly helpful in relieving the pain, and that epidural therapy had helped but only temporarily. The ALJ, however, merely referred to those facts in passing; he pro*934vided no explanation for why they might be inconsistent with Banks’ complaints of disabling pain. To the contrary, the fact that several rounds of physical therapy did not alleviate Banks’ pain, and other methods were only temporarily successful, seem to support, not contradict, the complaints of chronic pain. Further, the record reveals that Banks repeatedly complained to his doctors of severe back and neck pain, for which his doctors prescribed at least nine different pain medications. At the hearing Banks explained that he had stopped taking prescription medication because the drugs made him extremely drowsy and, although initially helpful, their effectiveness had diminished. In addition, Banks testified he could no longer afford medical treatment after losing his health insurance when he lost his job in January 1997. In any case, without some reasoned explanation by the ALJ for rejecting Banks’ allegations, we cannot sustain his adverse credibility determination.
The ALJ’s inadequate credibility determination is not inconsequential here. At the fourth step the ALJ found that Banks retained the ability to perform his past work as a beer salesperson, as that job was modified to require no lifting. But crediting Banks testimony would establish that he was unable to satisfactorily perform that job after the accident because of his severe back and neck pain. Banks testified that he failed to reach his sales quota because of his inability to set up displays and handle the product for his customers. Because of his meager sales, Banks said he received, for first time in his 10-year career at the company, a poor performance evaluation. In addition, Banks testified to significant difficulties doing his job even as modified. Banks missed work initially because of pain but after being warned that future absences 'Would result in termination, Banks said that “Regardless of what it took to get there, I would get there.” According to Banks, he had to lay down and rest between each sales call, which may have further hampered his productivity. The record also reflects that in mid-1996 Banks aggravated his back at work, prompting his orthopedic surgeon to report that if Banks was as “unstable regarding a mechanical problem” as it appeared, he “may have difficulty returning to work.” Nothing in the record indicates that Banks’ condition improved since then; if anything, his degenerative condition worsened. He now has pain and weakness in his left arm in addition to his back and neck ailments.
The ALJ, however, did not discuss any of this evidence; rather, he seems to have relied solely on the fact that Banks worked in his modified position “for a considerable period of time” before being laid off “under circumstances that are not clear but which apparently had nothing to do with his ability to perform the work as modified.” But even if Banks lost his job due to downsizing, his functional difficulties and less than satisfactory performance are evidence of an inability to perform his job. The ALJ could not simply ignore this evidence. See Zurawski, 245 F.3d at 888 (it is incumbent upon an ALJ to discuss evidence contrary to the Commissioner’s position). If the ALJ thought Banks’ testimony was incredible, he should have explained why; the fact that Banks continued to work despite his pain is not alone determinative. We have emphasized that “employment is not proof positive of ability to work, since disabled people, if desperate (or employed by an altruist), can often hold a job.” Wilder v. Apfel, 153 F.3d 799, 801 (7th Cir.1998). Indeed, “[sjome disabled people manage to work for months, if not years, only as a result of superhuman effort, which cannot be sustained.” Perlman v. Swiss Bank Corp. *935Comprehensive Disability Prot. Plan, 195 F.3d 975, 983 (7th Cir.1999).
In addition to the insufficient credibility finding, Banks argues that the ALJ’s assessment of his residual functional capacity (RFC) falls below the mark. We agree. The ALJ determined that Banks retained the functional capacity to perform light and sedentary work “provided that he does not have to perform either activity for extended periods of time.” The ALJ does not identify what specific sedentary or light activities Banks cannot perform for extended periods, and it is unclear what he means by “extended,” particularly given that the evidence concerning how long Banks could sit, stand, or walk was conflicting. Because the ALJ’s RFC finding itself is unclear — and is unaccompanied by an analysis of the relevant evidence — we lack sufficient basis to conclude whether or not substantial evidence supports the finding, or whether the RFC comports with the actual functional demands and duties of Banks’ past work.
Not only is this finding unclear, but it is internally inconsistent. If the ALJ meant that Banks is incapable of prolonged sitting, standing, and walking, as the evidence showed, those limitations conflict with his determination that Banks could perform fight and sedentary work. See Peterson v. Chater, 96 F.3d 1015, 1016 (7th Cir.1996). According to Social Security Ruling 83-12 an individual who “may be able to sit for a time, but must then get up and stand or walk for a while [sic] before returning to sitting ... is not functionally capable of doing either the prolonged sitting contemplated in the definition of sedentary work ... or the prolonged standing or walking contemplated for most fight work.” Id. at 1016 (quoting SSR 83-12). The ruling goes on to say that there are jobs that allow a person to “sit or stand with a degree of choice,” id., and Banks’ job as a beer salesperson may have had that option. But Dr. Brannegan opined that Banks could stand or walk for at most two or three hours in an eight-hour work day regardless of such an option. If Banks could not stand or walk for six hours out of an eight-hour day, then he could not perform his beer salesperson job, since the job even as modified still required him to be on his feet for most of a work day that often exceeded eight hours. The Commissioner argues that Dr. Bran-negan’s opinion was not credible, but the ALJ made no such finding. The ALJ did not address Dr. Brannegan’s opinion at all. He simply concluded that Banks could perform his past work, “subject to the above residual functional capacity.” The ALJ, however, should have compared the specific, physical demands of Banks’ past job with his present capabilities. See Nolen v. Sullivan, 939 F.2d 516, 518-19 (7th Cir.1991); Strittmatter v. Schweiker, 729 F.2d 507, 509 (7th Cir.1984); SSR 82-62. His failure to do so, combined with the other flaws we have mentioned, renders the RFC finding woefully deficient.
None of this is to say that a conclusion that Banks is not disabled would be indefensible. But the ALJ’s failure to articulate the basis for discrediting Banks’ testimony and other evidence contrary to his conclusion, prevents us from finding that substantial evidence supports his determination at step four of the disability analysis. We therefore remand this case to the Commissioner for a thorough step-four evaluation, which should include reasoned assessments of Banks’ credibility and his residual functional capacity. If on remand the ALJ finds that Banks can no longer perform the duties of his past job, the ALJ should move to step five and consider, with the assistance of a vocational expert, whether Banks can do other work in the national economy.
*936For the reasons set forth above, the judgment of the district court is VACATED, and the case REMANDED to the district court with instructions to remand the case to the Commissioner for further proceedings consistent with this order.