[157]*157OPINION OF THE COURT
AMBRO, Circuit Judge.
Carol Post believes that she is entitled to long term disability benefits under her former employer’s disability plan. Her treating physicians maintain that she is disabled. On the other hand, Hartford Insurance Company, the plan administrator (who also happens to fund the plan), has hired reviewing physicians who maintain that Post is not disabled. In other words, the central issue in this case— whether Post is disabled — is a “battle, of the experts.”
“Battle-of-the-experts” cases are often easy for a reviewing court. If the trial court’s standard of review is arbitrary and capricious, then Hartford usually wins when it has produced sufficient evidence supporting its position. It cannot be said to have acted arbitrarily, and summary judgment in its favor is appropriate. On the other hand, if the standard is de novo, then summary judgment for either party must be vacated because there is credible evidence on both sides of the key fact question.
But this case, a claim that ERISA benefits were improperly denied, is anything but easy, for the trial court’s standard of review is neither arbitrary and capricious (at least in its traditional form) nor de novo. In these cases, district courts must select a standard of review that accords with the extent to which the plan administrator operates under a conflict of interest. Here we conclude that the District Court did not select the proper standard of review, and so we vacate and remand for consideration under the standard we deem to apply.
We affirm, however, the Court’s grant of summary judgment on Post’s claim for breach of fiduciary duty because it is barred by res judicata.
I. Facts and Procedural History
Carol Post was in a serious car accident in November 1993, just a few days after having major dental surgery. At the time, she was employed as a dentist by Overlook Hospital in Summit, New Jersey. She sustained a whiplash injury in the accident, but she nonetheless attempted to return to work soon afterward. After six days of working, she was forced to stop because of intractable pain. Overlook, however, offered for her to try working as a pharmacist for a while (as she has both dentistry and pharmacy degrees), and she accepted. She returned to work in December 1993, but was forced to take nearly a day off each week because of pain. After nine months of off-and-on working, she resigned due to pain in September 1994. During this period, she tried numerous physical therapy treatments, none of which significantly improved her condition. She returned to work again in January 1995, but resigned four months later because of continuing pain. She has not worked since.
Post’s medical record is voluminous. Between 1993 and 2003, she visited 14 doctors. Her pain management regimens ranged from traditional treatments like prescription drug combinations, trigger-point injections, and various forms of physical therapy, to more exotic treatments like acupuncture and biofeedback. She reports that none has given her significant relief. Her primary treating physician is currently Dr. Carolyn Britton, a professor of neurology at Columbia University. According to Dr. Britton, Post suffers from chronic post-traumatic pain syndrome characterized by severe myofacial pain; regular, debilitating headaches accompanied by sensitivity to light, nausea, and vomiting; irritable bowel syndrome; and insomnia. Dr. Britton believes that this syndrome is [158]*158directly attributable to Post’s car accident and that it renders her disabled from any sustained employment.
In keeping with Dr. Britton’s determination, Post’s view of the record is that it indicates that she sustained a traumatic whiplash injury that sensitized her central nervous system, thus triggering the development of chronic pain syndrome. This is Dr. Britton’s diagnosis, and it is supported by a number of other evaluations in the record.
Hartford, on the other hand, believes that the record indicates that Post suffered no more than a whiplash injury that has now healed. While it concedes that Post continues to report pain, it contends that the record contains no reliable diagnosis of a recognized debilitating condition. In support of its view, Hartford primarily relies on the reports of Dr. Ekaterina Mal-ievskaia, its reviewing physician, and Dr. Christopher Lynch, who performed an independent medical examination. Hartford also cites the opinions of Drs. Michael John Fiore and Joel Harris,1 who evaluated Post in 1994 and 1996, respectively.
This ease is governed by the Employee Retirement Income Security Act (“ERISA”), 29 U.S.C. §§ 1001-1461, because Overlook Hospital’s disability plan (the “Plan”) is an “employee welfare benefit plan” as defined by 29 U.S.C. § 1002(1). Post filed a disability claim with Hartford, Overlook’s disability carrier, soon after she ceased working in 1995. Hartford approved her claim, subject to periodic renewal. To be considered “totally disabled” under the Plan after December 6, 1997, she had to be “prevented by [disability from doing any occupation or work for which [she was] or could become qualified.”
From 1995 until 2002, Hartford paid out benefits. In August 1998, the Social Security Administration approved Post’s application for disability benefits, citing intractable cervical pain, chronic pain syndrome, and fibromyalgia2 as the relevant diag[159]*159noses. Soon after Post was approved for Social Security benefits, Hartford asked her to submit a copy of the administrative decision so that it could offset her benefits. She responded through counsel that 'Hartford was not entitled to an offset under the plain language of the Plan, but she did provide Hartford with a copy of the decision. Hartford eventually relented and accepted Post’s reading of the Plan.
For reasons not apparent from the record, sometime in late 1999 Hartford took a renewed interest in Post’s claim. The company surveilled her and reported in its claim notes that surveillance was unsuccessful, as she was not seen leaving her house. Hartford also began requesting copies of Post’s tax records, ostensibly to take a non-Social Security income offset, as the Plan allowed. It provides that “Hartford has the right to require, as part of Proof of Loss: (1) your [Post’s] signed statement identifying all Other Income Benefits, and (2) [satisfactory proof to the Hartford that you and your Dependents have duly applied for all Other Income Benefits which are available. The Hartford reserves the right to determine if proof of loss is satisfactory.” Hartford contends that the “proof ... that you ... have duly applied for all Other Income Benefits” language gives it the right to demand tax returns, though it is not clear how a tax return would reflect whether Post had applied for other income benefits. The plain language of this provision does not authorize the review of tax returns. (Incidentally, the tax returns confirm that Post was not receiving any income during the disputed period.)
In June 2001, Hartford determined that Post should submit to an independent functional capacity evaluation to confirm her disability. This was permissible under the Plan. Hartford hired a third-party service to notify Post of its request and to set up the evaluation. Because Post had requested that all communication go through counsel, the service’s operator phoned her attorney to schedule the evaluation. Here, the confusion began. As Hartford’s counsel explained at oral argument, apparently the service’s operator told Post’s attorney that Post had requested that he be phoned to schedule the evaluation — meaning simply that Post had requested that all communication go through him. Post’s attorney took the statement to mean that Post had requested the evaluation; thus, when he spoke with Post and found that she knew nothing about it, he relayed to the service that she had not requested it. It then reported to Hartford that Post had refused an evaluation in violation of the Plan. No written request was ever made.
In lieu of a functional capacity evaluation, Hartford referred Post’s file to its medical director, Dr. Malievskaia. She conducted a paper review and concluded that Post was not disabled because of a [160]*160lack of objective findings, specifically the absence of 11 of 18 potential trigger points that would support a diagnosis of fibro-myalgia.
In January 2002, Hartford terminated Post’s benefits. In its termination letter, Hartford quoted the Plan’s termination triggers, putting the following in bold font: “the date you refuse to be examined, if The Hartford requires an examination.” The letter went on to cite as the bases for termination Post’s alleged failure to submit to an evaluation at Hartford’s request and Dr. Malievskaia’s conclusion that Post was not disabled. The letter also invited Post to file an appeal within 60 days and to send any documents that she believed relevant. In March 2002, Hartford denied Post’s appeal. Hartford, however, recognized the confusion over scheduling the evaluation and offered to revisit its decision if she agreed to one. In the meantime, Post had sued Hartford for wrongful denial of benefits, and undergoing an evaluation became part of a settlement agreement. The settlement fully resolved that lawsuit.
Because Post’s treating physicians refused to write a prescription for a full-scale functional capacity evaluation, citing the damage it might cause given Post’s condition, Hartford agreed to a less strenuous examination. To perform the exam, Hartford hired Dr. Christopher Lynch. The record does not reflect any board certifications or specialties, only that he is a physician. His examination consisted primarily of testing Post for the 18 trigger points for fibromyalgia. Finding tenderness but no definite trigger points, Dr. Lynch concluded that she did not have fibromyalgia or any other disabling condition. After he submitted his report, Hartford issued a final denial of Post’s claim. Hartford specifically directed Dr. Lynch not to submit his report to Post, so she had no opportunity to respond to it.
Post then filed this suit in the District Court. In it, she claims that Hartford violated 29 U.S.C. § 1132(a)(1) and (2). Subparagraph 1132(a)(1)(B) allows an ERISA plan beneficiary to sue “to recover benefits due to him under the terms of his plan, to enforce his rights under the terms of the plan, or to clarify his rights to future benefits under the terms of the plan.” Paragraph 1132(a)(2) allows a beneficiary to sue for breaches of fiduciary duties that cause losses to the plan.
The District Court granted summary judgment in Hartford’s favor on the § 1132(a)(1)(B) claim, ruling that Post could not establish that Hartford acted arbitrarily and capriciously in denying her benefits. The Court also granted Hartford summary judgment on the § 1132(a)(2) claim on the ground that it was barred by res judicata. Specifically, the Court noted that it had dismissed that claim with prejudice in Post’s previous suit, and so she could not revive it in this suit. Post appeals both rulings.3
II. Deciding § 1132(a)(1)(B) Claims
A. The Sliding Scale Standard of Review
ERISA does not specify the standard of review that a trial court should apply in an action for wrongful denial of benefits. In Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101, 113, 109 S.Ct. 948, 103 L.Ed.2d 80 (1989), the Supreme Court held that the default standard of review in all § 1132(a)(1)(B) cases is de novo. The Court noted in a dictum that when a plan by its terms gives the administrator dis[161]*161cretion, which the plan at issue in Firestone did not, the administrator’s decisions are upheld unless they abuse that discretion. Id. at 115, 109 S.Ct. 948. On the issue of conflicts of interest, the Court noted that “if a benefit plan gives discretion to an administrator or fiduciary who is operating under a conflict of interest, that conflict must be weighed as a ‘facto[r] in determining whether there is an abuse of discretion.’ ” Id. (quoting Restatement (Second) of Trusts § 187 cmt. d (1959)).
Addressing conflicts of interest in the post-Firestone era, most courts of appeals have adopted a “sliding scale” standard of review. This approach grants the administrator deference in accordance with the level of conflict. Thus, if the level of conflict is slight, most of the administrator’s deference remains intact, and the court applies something similar to traditional arbitrary and capricious review; conversely, if the level of conflict is high, then most of its discretion is stripped away. Doe v. Group Hospitalization & Med. Servs., 3 F.3d 80, 87 (4th Cir.1993).
In Judge Becker’s scholarly opinion in Pinto v. Reliance Standard Life Insurance Co., 214 F.3d 377, 392 (3d Cir.2000), we cast our lot with the sliding scale approach. Among the eleven courts of appeals that have reported decisions in this area, six have adopted some version of the sliding scale.4 Id.; Vega v. Nat’l Life Ins. Servs., Inc., 188 F.3d 287, 296 (5th Cir.1999) (en banc); Woo v. Deluxe Corp., 144 F.3d 1157, 1161-62 (8th Cir.1998); Chojnacki v. Georgia-Pacific Corp., 108 F.3d 810, 815 (7th Cir.1997); Doe, 3 F.3d at 87; Miller v. Metro. Life Ins. Co., 925 F.2d 979, 984 (6th Cir.1991). In addition, the Ninth Circuit Court of Appeals follows a “substantially similar” approach, though it rejects the sliding-scale metaphor. Abatie v. Alta Health & Life Ins. Co., 458 F.3d 955, 967 (9th Cir.2006) (en banc) (choosing simply to note that “[a] district court, when faced with all the facts and circumstances, must decide in each case how much or how little to credit the plan administrator’s reason for denying insurance coverage”). In Pinto, we held that the sliding scale approach was most faithful to Firestone’s command that the level of conflict be considered as a factor in shaping arbitrary and capricious review. 214 F.3d at 392.
B. Contours of the Sliding Scale
The premise of the sliding scale approach is that courts should examine benefit denials on their facts to determine whether the administrator abused its discretion. Id. at 391. To apply the approach, courts first consider the evidence [162]*162that the administrator acted from an improper motive and heighten their level of scrutiny appropriately. Id. at 392. Second, they review the merits of the decision and the evidence of impropriety together to determine whether the administrator properly exercised the discretion accorded it. Id. at 394. If so, its decision stands; if not, the court steps into the shoes of the administrator and rules on the merits itself.
At its best, the sliding scale reduces to making a common-sense decision based on the evidence whether the administrator appropriately exercised its discretion. This theme, rather than getting bogged down in trying to find the perfect point on the sliding scale, should be district courts’ touchstone.
C. Sorting Individual Cases
Determining how to apply heightened arbitrary and capricious review requires considering both structural and procedural factors. Pinto, 214 F.3d at 392-93. The structural inquiry focuses on the financial incentives created by the way the plan is organized, whereas the procedural inquiry focuses on how the administrator treated the particular claimant. While there is no magic to the order in which these inquiries are conducted, our previous cases have considered structure first. We do the same.
1. Structural factors
Our concern with structure derives from the common law of trusts. As the Supreme Court noted in Firestone, the law of trusts requires that courts take a trustee’s self-interest into account. 489 U.S. at 115, 109 S.Ct. 948 (quoting Restatement (Second) of Trusts § 187 cmt. d (1959)). The Court based this pronouncement primarily on the Second Restatement. Since then, the ALI has published the Third Restatement, which further clarifies that while it is permissible for a trustee to act under a structural conflict of interest, its discretionary decisions “will be subject to especially careful scrutiny.” Restatement (Third) of Trusts § 37 cmt. f(l) (2003). Under ERISA, plan administrators are, for most purposes, treated like common-law trustees. Firestone, 489 U.S. at 110, 109 S.Ct. 948. Like common-law trustees, plan administrators are accorded discretion and judicial deference (if the plan so provides); in return, they assume fiduciary duties of care and loyalty to their beneficiaries. 29 U.S.C. § 1104(a). So long as we have no reason to doubt the administrator’s faithfulness to those duties, this model works well. We, however, are wary of according a fiduciary deference when the structure of the plan gives it financial incentives to act against the participants’ interest. See Restatement (Third) of Trusts § 50 illus. 1.
As an initial note, federal courts of appeals are split on the issue of what is a structural conflict. We have long held that a structural conflict arises when the administrator has a non-trivial financial incentive to act against the interests of the beneficiaries. Pinto, 214 F.3d at 389. Such a conflict is, by itself, sufficient to heighten our review.5 Id. at 390. Our Court’s holdings are in line with black-letter trust law. The Second Restatement, on which the Supreme Court relied in Firestone, defines a “conflict” as merely [163]*163“an interest in the trustee conflicting with that of the beneficiaries.” Restatement (Second) of Trusts § 187 cmt. d (1959). This statement is worded broadly — almost to the point of being tautological — but it applies by its own terms to a situation in which the administrator has an interest {e.g., in profit or a better bottom line) that is adverse to the interests of beneficiaries seeking payment.
In sharp disagreement, the Court of Appeals for the Seventh Circuit holds that it is improper to label those situations “conflicts of interest.” See Rud v. Liberty Life Assur. Co. of Boston, 438 F.3d 772, 776 (7th Cir.2006) (Posner, J.). The problem, it argues, is that we generally assume that parties to a contract are self-interested, and it is inimical to the law of contracts to confuse self-interest with a conflict of interest. Id. This is no doubt logical, yet the Supreme Court has held that ERISA places us in the realm of trust law, not contract law. Firestone, 489 U.S. at 110—11, 109 S.Ct. 948. Moreover, were we to apply contract law, we would review plans de novo from the start, for there is no analog to fiduciary discretion in the common law of contracts. But we are not, and our position, in strict accordance with Supreme Court precedent, follows the common law of trusts.
Pinto listed four non-exclusive structural factors for courts to consider: (1) the sophistication of the parties, (2) the information accessible to the beneficiary, (3) the financial arrangement between the employer and administrator, and (4) the financial status of the administrator. 214 F.3d at 392. In subsequent cases, we have also considered the administrator’s claim evaluation process, according more deference to administrators that use an independent body to evaluate claims (thus lessening the effect of any conflict). Stratton v. E.I. DuPont De Nemours & Co., 363 F.3d 250, 255 (3d Cir.2004). All of these factors relate to whether the plan is set up so that the administrator has strong financial incentives routinely to deny claims in close cases- — in short, whether the administrator’s incentives make treating it as an unbiased fiduciary counterintuitive. Pinto, 214 F.3d at 388. We emphasize that courts should focus on this question and not get bogged down in factors, for this is anything but a mechanistic test. Rather, it is a broad-based inquiry into whether the structure of the plan raises concerns about the administrator’s financial incentive to deny coverage improperly. This makes sense, as ERISA plans come in many forms.
We have held that two aspects of some plans’ financial structure raise particular concern: (1) when a plan is funded on a case-by-case basis, Skretvedt v. E.I. DuPont de Nemours & Co., 268 F.3d 167, 174 (3d Cir.2001), and (2) when it is funded and administered by an outside insurer, Pinto, 214 F.3d at 390. Case-by-case funding simply means that the administrator pays claims out of its operating budget, rather than from segregated monies- that the employer sets aside according to an actuarial formula. This raises concerns because it means that each dollar paid out is a dollar out of the administrator’s pocket. Strat-ton, 363 F.3d at 254. Thus, the administrator has a financial incentive to deny claims.
This concern is compounded when it is an outside insurer, rather than the employer, that funds and administers the plan, for we presume that employers have at least some self-interest in seeing that benefits are paid fairly. After all, employees’ morale will suffer if they perceive that their benefits are illusory. When the plan is funded by an outside insurer, however, the employer is a step removed from the process, making it less likely to feel the [164]*164full effects of employee dissatisfaction with claims handling. Pinto, 214 F.3d at 389.6
We have also noted that when the claimant is a former employee, any dissatisfaction with the claims handling process is less likely to translate into a significant financial disincentive for the employer. Id. at 388. In addition, when the employer is in financial difficulty, the dissatisfaction of employees is less likely to be an incentive favoring them because paying off creditors will probably take priority over keeping up employee morale. Id. at 392.
Importantly, under Pinto, the structural analysis does not ask about the administrator’s behavior. Indeed, as Pinto held, the structure alone can require heightened review. 214 F.3d at 390. Pinto itself concerned a structure in which the plan administrator was an outside insurance company that received an actuarial premium from the employer. Id. Thus, what the insurer/administrator paid out came directly off its bottom line. Pinto noted that this structure creates a high level of financial conflict of interest, as the insurer/administrator has a strong incentive to construe claims in a light most favorable to it. Id. at 389. Thus, Pinto held that this structure alone gives rise to heightened scrutiny. Id. at 390.
When there is a structural conflict of interest mitigated by independent claim evaluation and no evidence of procedural bias, we have heightened our review only slightly. Stratton, 363 F.3d at 254-56. The animating logic of that case is that while there was a conflict of interest, there was also good reason to believe that it was of little moment, and so we held that we would defer to the administrator unless its decision was clearly unreasonable or not a product of an exercise of discretion at all.
When structural bias is not mitigated by independent claim evaluation, we have heightened our review a bit more. See Smathers v. Multi-Tool, Inc./Multi-Plastics, Inc. Employee Health & Welfare Plan, 298 F.3d 191, 199 (3d Cir.2002). There, we emphasized that we were not free to substitute our judgment for that of the fiduciary. Nevertheless, because the record revealed that the administrator had not adequately supported its decision, we concluded that it had not properly exercised its discretion. Id. at 200.
It is worth noting that we have not reported a case in which structural factors alone warranted anything more than moderately heightening our review. This is not fortuitous. Structural conflicts of interest warrant more searching review, but in the absence of evidence that bias infected the particular decision at issue, we defer to an administrator’s reasonable and carefully considered conclusions. See Orvosh v. Program of Group Ins. for Salaried Employees of Volkswagen of Am., Inc., 222 F.3d 123, 129 (3d Cir.2000).
2. Procedural Factors
As Pinto held, courts must also examine the process by which the administrator came to its decision to determine whether there is evidence of bias. 214 F.3d at 393. This sort of evidence can come in many forms, and a review of the caselaw reveals that we have identified numerous procedural irregularities that can raise suspicion. The following is an illustrative, not exhaustive, list of the irregularities identified: (1) reversal of posi[165]*165tion without additional medical evidence, id.; (2) self-serving selectivity in the use and interpretation of physicians’ reports, id.; (3) disregarding staff recommendations that benefits be awarded, id. at 394; and (4) requesting a medical examination when all of the evidence indicates disability, Kosiba v. Merck & Co., 384 F.3d 58, 67 (3d Cir.2004).
In considering procedural factors, the focus is whether, in this claimant’s case, the administrator has given the court reason to doubt its fiduciary neutrality. If it has, then the court must decide how much to heighten its scrutiny. If the irregularities are minor, few in number, and not sustained, then they may not counsel for raising the level much at all, for minor glitches reasonably can be chalked up to low-level carelessness. If, however, they are more serious, numerous, or regular, then they should raise more suspicion. Ko-siba, 384 F.3d at 66; Pinto, 214 F.3d at 393. Given the administrator’s familiarity with the claims process and the duties of a fiduciary, marked deviations from procedural norms cannot but raise questions about its neutrality.
In the face of significant evidence of procedural bias, we have reviewed its decision closely. Pinto, 214 F.3d at 394. When an ERISA administrator is not acting in accord with its fiduciary status, we are naturally wary of according it much of the deference that it would otherwise receive as a result of that status. Id. Evidence that an administrator’s decision was incorrect, coupled with evidence it was biased, can add up to a conclusion that its decision was not the product of reasoned discretion, but of anti-claimant bias,, in which case the decision should be reversed. Id. at 395.
In the face of non-trivial evidence of procedural bias, the standard of review should be raised; the more difficult question is how much. In Kosiba, we discerned non-trivial evidence of procedural bias but, as it was neither egregious nor coupled with evidence of structural bias, we heightened our scrutiny only a moderate amount. 384 F.3d at 68. In Pinto, on the other hand, we found that the evidence of procedural bias was coupled with evidence of structural bias, and so we heightened our review substantially. 214 F.3d at 394.
III. Applying the Sliding Scale to This Case
A. Structural Factors
Addressing the structural factors, the District Court seemed to confuse the structural analyses in Pinto and Stratton. Pinto held that a non-trivial structural conflict gives rise to heightened scrutiny— that is, it pushes the standard of review above the low end of the sliding scale. 214 F.3d at 393. Stratton added that when the structural conflict is trivial, the low end of the scale is appropriate. 363 F.3d at 254-55. What made the conflict trivial in Stratton was that the employer/administrator, while conflicted, was a step removed from the claim evaluation process. Id. Here, on the other hand, the administrator is an outside insurer that makes claims decisions itself. This is the very sort of conflict that Pinto declared to be substantial and worthy of raising the standard of review. 214 F.3d at 393. In addition, Post is a former employee, so it is doubtful that her dissatisfaction with the claims-handling process will filter back to Overlook and translate into pressure on Hartford to deal more precisely with claims.
The District Court correctly noted that the other factors mentioned in Pinto— sophistication of the parties, accessibility of information, and the financial status of [166]*166the administrator — seem not to counsel in favor of heightened scrutiny. Following Pinto, however, the structural factors that do present a conflict of interest are sufficient to require at least moderately heightened review. Id. We now proceed to whether procedural factors counsel us to increase even more our degree of review.
B. Procedural Factors
On the issue of procedural irregularities, the District Court wrote that “procedural anomalies appear to form a pattern of Hartford being overly aggressive in its attempts to reduce or eliminate Post’s [disability] benefits and then attempting to rectify the situation when it realized its error.” The Court named four aspects of the process that appeared irregular, yet it ultimately concluded that they were too minor to heighten further its scrutiny. We address each in turn and two additional matters brought up by Post.
First, Hartford attempted to use Post’s Social Security benefits to offset her disability benefits, despite the Plan not allowing such an offset. After Post’s attorney protested, Plartford relented. This, of course, may have been a good-faith mistake on Hartford’s part, but it is a plan administrator’s responsibility to know the contents of the plan. Our dissenting colleague believes that the Plan itself was confusing enough that Hartford’s mistake was understandable. But Hartford is a large, sophisticated insurance company, and the Plan is its own design. Thus, we are less willing to draw such benign inferences (particularly at the summary judgment stage, where we draw all reasonable inferences in Post’s favor) from Hartford’s supposed confusion about the contents of its own contract.
Second, Hartford terminated Post’s benefits in part because she allegedly refused to undergo a functional capacity evaluation. The record suggests, however, that Post had not refused an evaluation and that Hartford was quick to conclude that she had despite never making a written request. During the appeals process, however, Hartford relented and agreed to reconsider Post’s appeal if she would agree to undergo an evaluation. Of concern is that Hartford did not allow Post to see Dr. Lynch’s report before making its final decision to terminate. Thus she had no opportunity to allow her treating physicians to comment on it.
Third, Hartford’s decision to terminate benefits relied heavily on Dr. Malievskaia’s report, which was not based on a physical examination. While the District Court correctly noted that ERISA does not require that plan administrators give the opinions of treating physicians special weight, Black & Decker Disability Plan v. Nord, 538 U.S. 822, 823-24, 123 S.Ct. 1965, 155 L.Ed.2d 1034 (2003), courts must still consider the circumstances that surround an administrator ordering a paper review. On one hand, nothing in the record specifically suggests that Hartford ordered this review in bad faith, as, we assume, periodic reviews are typical in the industry. On the other hand, we note that at the time of the review the overwhelming weight of evidence in Post’s record argued in her favor.7
Fourth, Hartford surveilled Post. As the District Court noted, while surveillance is [167]*167an aggressive tactic, nothing prohibits its use. Post argues that the bothersome point is that Hartford continued to investigate her claim despite its surveillance revealing that she did not leave her home. We agree. The fact that Post did not leave her home while she was under surveillance is perfectly consistent with, and corroborative of, her claim for disability. Yet Hartford was undeterred in continuing to pursue evidence that Post was not disabled. Indeed, the very fact that its employees characterized the results of the surveillance as “unsuccessful” suggests that its motive was to find evidence to deny Post’s claim.
In addition to these incidents, Post cites Hartford’s request for her tax returns as evidence of bad faith. As the District Court pointed out, the Plan did allow Hartford to reduce Post’s benefits by the amount of any income she was receiving from working; thus Hartford’s request for proof that she was receiving none was not beyond the pale. Nonetheless, Hartford’s pursuit of Post’s tax returns in the face of ambiguous Plan language is accurately characterized as an aggressive tactic.8
Post also cites Hartford’s denial of benefits despite a favorable Social Security decision as evidence of bad faith. Our Court has not passed on the relevance of Social Security decisions in determining the appropriate standard of review, but other courts of appeals and some district courts have held that a disagreement with the Social Security Administration is a relevant — though not dispositive — factor. See Glenn v. MetLife, 461 F.3d 660, 669 (6th Cir.2006) (“[A]n ERISA plan administrator’s failure to address the Social Security Administration’s finding that the claimant was ‘totally disabled’ is yet another factor that can render the denial of further longterm disability benefits arbitrary and capricious.”); Lopes v. Metro. Life Ins. Co., 332 F.3d 1, 6 n. 9 (1st Cir.2003); Whatley v. CNA Ins. Co., 189 F.3d 1310, 1314 n. 8 (11th Cir.1999) (per curiam); Edgerton v. CNA Ins. Co., 215 F.Supp.2d 541, 549 (E.D.Pa.2002); Dorsey v. Provident Life & Accident Ins. Co., 167 F.Supp.2d 846, 856 n. 11 (E.D.Pa.2001). We agree that a disagreement is relevant though not dispositive, particularly (as here) when the administrator rejects the very diagnoses on which the Social Security benefits determination is based.9
In sum, we agree with the District Court that, on this record, each irregularity here [168]*168may appear minor. But given their number and regularity, the standard of review should be further heightened. As in Kosi-ba, we recognize that Hartford may offer plausible explanations for those irregularities, but in setting the standard of review the issue is merely whether the process raises questions. See 384 F.3d at 68. In this case, the sheer number of irregularities coupled with Hartford’s aggressive posture raise concerns, and so the standard of review must be heightened. This procedural posture suggests that we move toward the high end of the sliding scale, much as we did in Pinto. 214 F.3d at 394.
C. Conclusion
Both structural and procedural factors favor a more searching standard of review than was used here. In light of what we believe the standard of review should be, the District Court erred by applying only slightly heightened review. Moving toward the high end of the sliding scale, the District Court must searchingly review both the merits and the process to determine if Hartford’s decision was not the product of reasoned, disinterested discretion. No doubt the evidence on the merits appears close. But a factfinder reviewing the merits could yet determine that the weight of the medical evidence supports Post and that it, coupled with the evidence of bias, yields the conclusion that Hartford did not properly exercise its discretion.
IV. Other Issues
A. Closure of the Record
Generally, only evidence in the administrative record is admissible for the purpose of determining whether the plan administrator’s decision was arbitrary and capricious. Kosiba, 384 F.3d at 67 n. 5; Mitchell v. Eastman Kodak Co., 113 F.3d 433, 440 (3d Cir.1997); Abnathya v. Hoffmann-La Roche, Inc., 2 F.3d 40, 48 n. 8 (3d Cir.1993).
In the wake of Pinto, however, we have modified that holding to allow the consideration of extrinsic evidence when deciding how much to heighten our review. Kosiba, 384 F.3d at 67 n. 5. That evidence must show “potential biases and conflicts.” Id. In particular, we have noted that considering evidence of a plan’s funding mechanism would be appropriate. Id. Here, however, Post’s supplemental exhibits are all medical reports. The first five are reports from doctors that Post consulted between 1993 (just after the accident) and 1996. See Appellant’s Br. 6-9. The last two are summaries of Post’s condition prepared by' her current doctors at the request of counsel in May 2005 (nearly two years after Hartford issued its final denial of benefits). Id. at 18-19, 28. Post has provided no explanation why the reports produced between 1993 and 1996 were not sent to Hartford for its consideration. Similarly, if she wanted Hartford to consider her treating physicians’ responses to Dr. Lynch’s report or their summaries of her medical condition, she should have submitted them (and thus made them part of the administrative record) soon after she received Hartford’s denial of benefits, but she did not. Because all of these documents are medical reports, they are not relevant to the issue of bias; rather, [169]*169they are only relevant to whether Hartford reached the right decision. Under Mitchell, they cannot be considered for that purpose because they were not submitted to Hartford and made part of the record. 113 F.3d at 440. Thus,, the District Court acted properly in not considering them.
B. The Section 1132(a)(2) Claim
The doctrine of res judicata “protects] litigants from the burden of relit-igating an identical issue with the same party or his privy and ... promotes] judicial economy by preventing needless litigation.” Parklane Hosiery Co. v. Shore, 439 U.S. 322, 327, 99 S.Ct. 645, 58 L.Ed.2d 552 (1979). To apply, the following three prongs must be met: “(1) a final judgment on the merits in a prior suit involving (2) the same parties or their privies and (3) a subsequent suit based on the same cause of action.” Lubrizol Corp. v. Exxon Corp., 929 F.2d 960, 963 (3d Cir.1991). Here, the parties agree that prongs two and three are met; their dispute is over whether the Court rendered a final judgment on the merits in their previous suit.
In that suit, the District Court dismissed a cause of action alleging violation of 29 U.S.C. § 1132(a)(2) for failure to state a claim. See Post v. Hartford Life & Accident Ins. Co., No. CIV.A. 02-1917, 2002 WL 31741470, at *2 (E.D.Pa. Dec.6, 2002). Dismissal for failure to state a claim is a final judgment on the merits for res judicata purposes. Federated Dep’t Stores v. Moitie, 452 U.S. 394, 399 n. 3, 101 S.Ct. 2424, 69 L.Ed.2d 103 (1981). Moreover, res judicata bars not only claims that were brought in the previous action, but also claims that could have been brought. CoreStates Bank, N.A. v. Huls America, Inc., 176 F.3d 187, 194 (3d Cir.1999). Thus, for Post to maintain a § 1132(a)(2) claim, she would have to explain to the Court why it could not have been brought in 2002. She has made no attempt to do so.
As Hartford notes, Pdst’s claim has the additional problem that it, too, fails properly to allege a violation of § 1132(a)(2). Post seeks to recover individually for Hartford’s alleged breach of fiduciary duty. Under § 1132(a)(2) this is impossible, for that section allows beneficiaries to recover assets on behalf of the plan only. Mass. Mut. Life Ins. Co. v. Russell, 473 U.S. 134, 140, 105 S.Ct. 3085, 87 L.Ed.2d 96 (1985). In other words, § 1132(a)(2) does not authorize suits for the recovery of individual benefits. Hozier v. Midwest Fasteners, Inc., 908 F.2d 1155, 1162 n. 7 (3d Cir.1990) (“Because plaintiffs here seek to recover benefits allegedly owed to them in their individual capacities, their action is plainly not authorized by either § 409 or § 502(a)(2).”).10
Y. Conclusion
We conclude that the District Court should have applied a more searching review to this case because of the non-trivial evidence of structural and procedural bias. Because that was not the standard applied here, we vacate the District Court’s grant of summary judgment in Hartford’s favor on the § 1132(a)(1)(B) claim and remand for further proceedings.
We affirm, however, its grant of summary judgment on the § 1132(a)(2) claim because principles of res judicata bar that claim.