Opinion for the Court filed by Circuit. Judge SCALIA.
SCALIA, Circuit Judge:The Association of Data Processing Service Organizations, Inc. (“ADAPSO”), a national trade association representing the data processing industry, and two of its members petition this court for review of two orders of the Board of Governors of the Federal Reserve System, pursuant to 12 U.S.C. § 1848 (1982). In No. 82-1910, they seek review of the Board’s July 9, 1982 order approving Citicorp’s application to establish a subsidiary, Citishare, to engage in certain data processing and transmission services. Order Approving Engaging in Data Processing and Data Transmission Activities, 68 Fed.Res.Bull. 505 (1982) (“Citicorp Order”). In No. 82-2108, they seek review of the Board’s August 23, 1982 order, entered after notice and comment rulemaking, amending those portions of Regulation Y which dealt with the performance of data processing activities by bank holding companies. Data *305Processing and Electronic Funds Transfer Activities, 47 Fed.Reg. 37,368 (1982) (as set forth at 12 C.F.R. §§ 225.4(a)(8), 225.-123(e) (1983) (“Regulation Y Order”).1 We consolidated the two appeals'.
The Bank Holding Company Act of 1956, ch. 240, 70 Stat. 133 (codified as amended at 12 U.S.C. §§ 1841-50 (1982)) (the “Act”), requires all bank holding companies to seek prior regulatory approval before engaging in nonbanking activities. The restrictions do not apply to:
activities ... which the Board after due notice and opportunity for hearing has determined (by order or regulation) to be so closely related to banking or managing or controlling banks as to be a proper incident thereto____ In determining whether a particular activity is a proper incident to banking or managing or controlling banks the Board shall consider whether its performance by an affiliate of a holding company can reasonably be expected to produce benefits to the public, such as greater convenience, increased competition, or gains in efficiency, that outweigh possible adverse effects, such as undue concentration of resources, decreased or unfair competition, conflicts of interests, or unsound banking practices.
12 U.S.C. § 1843(c)(8). Section 1848, the source of our review authority, provides that “[t]he findings of the Board as to the facts, if supported by substantial evidence, shall be conclusive.” Id. at § 1848.
On February 23, 1979, Citicorp applied for authority to engage, through its subsidiary Citishare, in the processing and transmission of banking, financial, and economic related data through timesharing, electronic funds transfer, home banking and other techniques. It also sought permission to sell its excess computing capacity and some computer hardware. The Board published notice of Citicorp’s application, which was protested by ADAPSO, and set it for formal hearing. 45 Fed.Reg. 41,533 (July 19, 1980). Before the hearing was held, Citicorp amended its application to add certain activities and to request amendment of Regulation Y to permit the activities it had specified. The Board published an Amended Order for Hearing and invited public comments and participation. 45 Fed.Reg. 76,515 (Nov. 19, 1980). A formal hearing was held before an Administrative Law Judge in which the merits of both the application and the proposed rule were considered. In addition, more than sixty companies and individuals submitted written comments on the proposed rule. On March 29, 1982, the AU decided that the activities proposed by Citicorp were closely related to banking and would produce benefits to the public which would outweigh their costs. In re: Application of Citicorp to Engage in Data Processing and Transmission Activities, AU Recommended Decision, J.A. B-68 to B-123 (“Recommended Decision”). The AU also recommended amendments to Regulation Y that would permit those activities contained in the Citicorp application. On July 9, 1982, the Board adopted the AU’s recommendation to approve the Citicorp application, with certain restrictions. On August 23, 1982, the Board adopted the AU’s recommended amendments to Regulation Y, again with certain restrictions. ADAPSO, and two of its members, participants in the actions below, filed these petitions for review.
I. Standard of Review
We are faced at the outset with a dispute regarding the proper standard of review. These consolidated appeals call for us to review both an on-the-record adjudication and an informal notice and comment rule-making. Petitioners contend that the substantial evidence standard, which presumably authorizes more rigorous judicial review, should govern our review of both orders. Petitioners’ Reply Brief at 26-28. The Board agrees, noting that § 1848 applies a substantial evidence standard to fac*306tual determinations. Respondent’s Brief at 22. Intervenor Citicorp contends that while the substantial evidence standard should govern review of the Citicorp order, Regulation Y should be upset only if arbitrary or capricious. Citicorp Brief at 16-17. Intervenor California Bankers Clearing House Association, addressing only Regulation Y, also advocates review under the arbitrary or capricious review standard. CBCHA Brief at 8-14. The parties’ submissions on this point reflect considerable confusion, which is understandable when one examines decisions defining the standard.of review under this statute.
Both of the Supreme Court's opinions reviewing action of the Board in amending Regulation Y noted that the Board’s determination “is entitled to the greatest deference,” Board of Governors of the Federal Reserve System v. Investment Company Institute, 450 U.S. 46, 56, 101 S.Ct. 973, 981-82, 67 L.Ed.2d 36 (1981) (“ICI ”); Securities Industry Association v. Board of Governors of the Federal Reserve System, — U.S. -, 104 S.Ct. 3003, 3009, 82 L.Ed.2d 158 (1984) (“SIA ”), but neither of them discussed the applicable standard of review, or even referred to § 1848.2 The courts of appeals, however, have applied the substantial evidence standard of § 1848 to Board adjudications such as the authorization in the first order here under review, Securities Industry Association v. Board of Governors of the Federal Reserve System, 716 F.2d 92, 101-02 (2d Cir.1983), aff'd, — U.S. -, 104 S.Ct. 3003, 82 L.Ed.2d 158 (1984), while applying the arbitrary or capricious standard, despite § 1848, to Board rules, including specifically amendments of Regulation Y, National Courier Association v. Board of Governors of the Federal Reserve System, 516 F.2d 1229 (D.C.Cir.1975); Association of Bank Travel Bureaus v. Board of Governors of the Federal Reserve System, 568 F.2d 549 (7th Cir.1978); see Investment Company Institute v. Board of Governors of the Federal Reserve System, 551 F.2d 1270, 1281 (D.C.Cir.1977) (dicta). In fact one appellate opinion has, like this one, addressed precisely the situation in which both an adjudicatory authorization and an amendment of Regulation Y were at issue in the same case—and applied the § 1848 substantial evidence standard to the former but the arbitrary or capricious standard to the latter. Compare Alabama Association of Insurance Agents v. Board of Governors of the Federal Reserve System, 533 F.2d 224, 246 (5th Cir.1976), with id. at 240. This would make a lot of sense if, as the Board has argued in some cases, § 1848 in its totality applies only to adjudication rather than rulemaking, since it is limited to “orders” of the Board, a word which the Administrative Procedure Act (“APA”) defines to mean the product of an adjudication. See 5 U.S.C. § 551(4), (6) (1982). Such a technical interpretation of the provision, however, has been uniformly and quite correctly rejected. See Investment Company Institute, supra, 551 F.2d at 1276-78; Alabama Association of Insurance Agents, supra, 533 F.2d at 234-35. That leaves the courts with the difficult task of explaining why the last sentence of § 1848, unlike all the rest of it, should be deemed to apply only to adjudication and not to rulemaking. Difficult, because there is nothing in either the text3 or the *307legislative history of the section to suggest such a result. The courts applying the arbitrary or capricious standard to Board rulemaking (which, as stated above, include all the courts that have confronted the issue) dispose of this problem either by totally ignoring it, see Alabama Association of Insurance Agents, supra, 533 F.2d at 240, 246; cf Investment Company Institute, supra, 551 F.2d at 1281, or by noting that the parties “do not appear to contest” the point, National Courier, supra, 4 516 F.2d at 1235 n. 8, or by the ipse dixit that “[w]e interpret [the last sentence of § 1848] to apply to findings of fact ‘on the record’ in an adjudicatory hearing as contrasted with a rulemaking proceeding,” Association of Bank Travel Bureaus, supra, 568 F.2d at 552 n. 5.
We think that there is no basis for giving the last sentence of § 1848 anything less than the general application given to the rest of the section. The Supreme Court’s pronouncement that the “greatest deference” is to be given to the determinations of the Board, and the court of appeals decisions applying the arbitrary or capricious test to Board rulemaking, seem to us explicable on quite different grounds— namely, that in their application to the requirement of factual support the substantial evidence test and the arbitrary or capricious test are one and the same. The former is only a specific application of the latter, separately recited in the APA not to establish a more rigorous standard of factual support but to emphasize that in the case of formal proceedings the factual support must be found in the closed record as opposed to elsewhere. We shall elaborate upon this point because it is not uncommon for parties to expend great effort in appeals before us to establish which of the two standards is applicable where in fact their operation is precisely the same.
The “scope of review” provisions of the APA, 5 U.S.C. § 706(2),5 are cumulative. Thus, an agency action which is supported by the required substantial evidence may in another regard be “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law”—for example, because it is an abrupt and unexplained departure from agency precedent. Paragraph (A) of subsection 706(2)—the “arbitrary or capricious” provision—is a catchall, picking up administrative misconduct not covered by the other more specific paragraphs. Thus, in those situations where paragraph (E) has no application (informal rulemaking, for example, which is not governed by §§ 556 and 557 to which paragraph (E) refers), paragraph (A) takes up the slack, so to speak, enabling the courts to strike down, as arbitrary, agency action that is devoid of needed factual support. When the arbitrary or capricious standard is performing that function of assuring factual support, there is no substantive difference between what it requires and what would be required by the *308substantial evidence test, since it is impossible to conceive of a “nonarbitrary” factual judgment supported only by evidence that is not substantial in the APA sense— i.e., not “ ‘enough to justify, if the trial were to a jury, a refusal to direct a verdict when the conclusion sought to be drawn ... is one of fact for the jury,’ ” Illinois Central R.R. v. Norfolk & Western Ry., 385 U.S. 57, 66, 87 S.Ct. 255, 260, 17 L.Ed.2d 162 (1966) (quoting NLRB v. Columbian Enameling & Stamping Co., 306 U.S. 292, 300, 59 S.Ct. 501, 505, 83 L.Ed. 660 (1939)).
We have noted on several occasions that the distinction between the substantial evidence test and the arbitrary or capricious test is “largely semantic,” Aircraft Owners and Pilots Association v. FAA, 600 F.2d 965, 971 n. 28 (D.C.Cir.1979); Pacific Legal Foundation v. Department of Transportation, 593 F.2d 1338, 1343 n. 35 (D.C.Cir.1979); American Public Gas Association v. FPC, 567 F.2d 1016, 1028-29 (D.C.Cir.1977), and have indeed described that view as “the emerging consensus of the Courts of Appeals,” Pacific Legal Foundation, supra, 593 F.2d at 1343 n. 35. See Associated Industries v. Department of Labor, 487 F.2d 342, 349-50 (2d Cir.1973) (Friendly, J.); National Nutritional Foods Association v. Weinberger, 512 F.2d 688, 705. (2d Cir.1975) (Lumbard, J., concurring); Paccar, Inc. v. NHTSA, 573 F.2d 632, 636 (9th Cir.1978) (purporting to avoid the issue, but seemingly not doing so). Leading commentators agree:
Does the extent of required factual support for rules depend in part on whether the standard for review is “substantial evidence” or “arbitrary and capricious”? Although from 1946 until some time during the 1970s the dominant answer probably was yes, a change to a no answer has probably occurred during the 1970s ....
1 K. Davis, Administrative Law Treatise § 6:13 at 512 (2d ed. 1978).
In review of rules of general applicability made after “notice and comment” rule-making, [substantial evidence and arbitrary or capricious] criteria converge into a test of reasonableness.
Review without an agency record thus comes down to review of reasonableness. [T]he question of reasonableness is also the one which the court must now ask itself in reviewing findings of fact under the post-APA substantial evidence rule.
B. Schwartz, Administrative Law 604, 606 (1976).
As noted earlier, this does not consign paragraph (E) of the APA’s judicial review section to pointlessness. The distinctive function of paragraph (E)—what it achieves that paragraph (A) does not—is to require substantial evidence to be found within the record of closed-record proceedings to which it exclusively applies. The importance of that requirement should not be underestimated. It is true that, as the Supreme Court said in Camp v. Pitts, 411 U.S. 138, 142, 93 S.Ct. 1241, 1244, 36 L.Ed.2d 106 (1973), even informal agency action (not governed by paragraph (E)) must be reviewed only on the basis of “the administrative record already in existence.” But that is quite a different and less onerous requirement, meaning only that whether the administrator was arbitrary must be determined on the basis of what he had before him when he acted, and not on the basis of “some new record made initially in the reviewing court,” id. That “administrative record” might well include crucial material that was neither shown to nor known by the private parties in the proceeding—as indeed appears to have been the situation in Camp v. Pitts itself. It is true that, in informal rulemaking, at least the most critical factual material that is used to support the agency’s position on review must have been made public in the proceeding and exposed to refutation. That requirement, however, does not extend to all data, see Administrative Conference of the United States, Recommendation No. 74-4, Preenforcement Judicial Review of Rules of General Applicability, 1 C.F.R. § 305.74-4 (1975); Verkuil, Judicial Review of Informal Rulemaking, 60 Va.L. *309Rev. 185 (1974); and it only applies in rule-making and not in other informal agency action, since it derives not from the arbitrary or capricious test but from the command of 5 U.S.C. § 553(c) that “the agency ... give interested persons an opportunity to participate in the rule making.” See Portland Cement Association v. Ruckelshaus, 486 F.2d 375, 393 n. 67 (D.C.Cir.1973).
Consolidated cases such as those before us here—involving simultaneous review of a rule (whose factual basis is governed only by paragraph (A)’s catch-all control against “arbitrary or capricious” action) and of a formal adjudication dealing with the same subject (whose factual basis is governed by paragraph (E)’s requirement of substantial evidence)—demonstrate why the foregoing interpretation of the two standards is the only interpretation that makes sense. If the standards were substantively different (and leaving aside for the moment consideration of any special effect of § 1848), the Citicorp order, authorizing one bank holding company’s data processing services, would be subject to more rigorous judicial review of factual support than the Regulation Y order which, due to its general applicability, would affect the operations of every bank holding company in the nation. Or, to put the point another way: If the Board had never issued any Regulation Y, and simply determined in the context of a particular application that the provision of timesharing services is “closely related” to banking, that determination, which could be reconsidered and revised in the context of the next adjudication, would require more factual support than the same determination in a rule-making, which would have immediate nationwide application and, until amended by further rulemaking, would have to be applied to all subsequent applications.
This seemingly upside-down application of varying standards is not an issue in the present case since, as we have observed, § 1848 makes it clear that only one standard—the substantial evidence test—applies to review of all Board actions. The relevance of the foregoing discussion here is to determine what that standard means. What we have said suggests that the normal (APA) meaning of the “substantial evidence” terminology connotes a substantive standard no different from the arbitrary or capricious test. One cannot dismiss out of hand, however, the possibility that, in this particular statute, a different meaning was intended—in which case that different standard would govern review of both rulemaking and adjudication. A number of “substantial evidence” review provisions have been attached to rulemaking authority, particularly in recent years. See, e.g., 29 U.S.C. § 655(f) (1982) (Occupational Safety and Health Act); 30 U.S.C. § 816(a) (1982) (Federal Coal Mine Health and Safety Act); 15 U.S.C. § 1193(e)(3) (1982) (Flammable Fabrics Act); 15 U.S.C. § 57a(e)(3)(A) (1982) (FTC Improvement Act of 1975). It is conceivable that some of these were intended, as the Fifth Circuit found with regard to such a provision in the Consumer Product Safety Act, 15 U.S.C. § 2060 (1982), to require the courts “to scrutinize [agency] actions more closely than an ‘arbitrary and capricious’ standard would allow.” Aqua Slide ‘N’ Dive Corp. v. CPSC, 569 F.2d 831, 837 (5th Cir.1978). Congress’s unpropitious use of the “substantial evidence” APA language for such a purpose is plausible, since the standard has acquired a reputation for being more stringent.6 One should not be too quick, however, to impute such a congressional intent. There is surely little appeal to an ineffable review standard that lies somewhere in-between the quantum of factual support required to go to a jury (the traditional “substantial evidence” test) and the “preponderance of the evidence” standard that would apply in de novo review. Moreover, 5 *310U.S.C. § 559 provides that a subsequent statute shall not be held to supersede or modify the APA provisions “except to the extent that it does so expressly.” While the provision for “substantial evidence” review where the APA would otherwise require only “arbitrary or capricious” review is unquestionably an “express” alteration, surely the import of the § 559 instruction is that Congress’s intent to make a substantive change be clear. This can lead to some fairly convoluted inquiries. Suppose, for example, that Congress clearly intended to switch to a stricter test, but was also clearly operating on the mistaken belief that the existing test (“arbitrary or capricious”) was more lenient than the “substantial evidence” standard. Should one give effect to the congressional intent to adopt a stricter standard, or rather to the congressional intent to adopt the “substantial evidence” standard (which is in fact, as we have discussed, no stricter)? Several decisions of this court stand for the proposition that a “substantial evidence” provision in the substantive statute under consideration did not have the effect of requiring increased factual support beyond that demanded by the normal “arbitrary or capricious” rulemaking standard of review. Public Systems v. FERC, 606 F.2d 973, 980 n. 34 (D.C.Cir.1979) (interpreting the predecessors of § 19(b) of the Natural Gas Act, 15 U.S.C. § 717r(b) (1982), and the corresponding provision of the Federal Power Act, 16 U.S.C. § 825i (b) (1982)); American Public Gas Association v. FPC, supra, 567 F.2d at 1028-29 (interpreting the predecessor of § 19(b) of the Natural Gas Act). See also Aircraft Owners and Pilots Association v. FA A, supra, 600 F.2d at 971 n. 28 (not deciding the issue with regard to the “substantial evidence” provision of the Federal Aviation Act, 49 U.S.C. § 1486(e) (1976)).
Fortunately, it is not necessary to engage in these speculations with regard to the “substantial evidence” provision of § 1848. The Supreme Court has evidently rejected the notion that it alters normal APA review requirements, since the Court’s opinions reviewing Board action deem the provision unworthy of mention, and specifically accord the Board “the greatest deference.” See, e.g., SIA, supra, 104 S.Ct. at 3009; ICI, supra, 450 U.S. at 56-57, 101 S.Ct. at 981-982. See also National Courier, supra, 516 F.2d at 1237. We hold, therefore, that the § 1848 “substantial evidence” requirement applicable to our review here demands a quantum of factual support no different from that demanded by the substantial evidence provision of the APA, which is in turn no different from that demanded by the arbitrary or capricious standard.
II. Closely Related to Banking
This appeal requires us, then, to decide whether the Board of Governors acted arbitrarily or capriciously—either because no substantial evidence existed to support its factual premises, or in any other respect—when it concluded, based on the record of the proceedings before it, that the new activities Citicorp proposed were closely related to banking. The test which the Board applied was that developed by this court in National Courier, supra, 516 F.2d at 1237:
As to what kinds of connections may qualify [as closely related to banking], at least the following seem to us within the statutory intent:
1. Banks generally have in fact provided the proposed services.
2. Banks generally provide services that are operationally or functionally so similar to the proposed services as to equip them particularly well to provide the proposed service.
3. Banks generally provide services that are so integrally related to the proposed services as to require their provision in a specialized form.
See, e.g., Citicorp Order, 68 Fed.Res.Bull. at 506 n. 3; Regulation Y Order, 47 Fed. Reg. 37,368 n. 1. These criteria are disjunctive rather than cumulative, see National Cornier, supra, 516 F.2d at 1237-38, and not exclusive, see 49 Fed.Reg. 806 (1984); SIA, supra, 104 S.Ct. at 3006 n. 5.
*311 The “Data Test”
Petitioners do not contend that the National Courier criteria are in themselves erroneous (they have subsequently been approved by the Supreme Court, see SIA, supra, 104 S.Ct. at 3006 n. 5), but they do contend, preliminarily, that the Board erroneously applied them to the kinds of data involved in the proposed services, rather than to the services themselves.
The depth of confusion that surrounds this argument is suggested by the fact that the Board’s brief denies that a data test was applied, Respondent’s Brief at 27-33, while the briefs of intervenors supporting the Board acknowledge that, at least as to the Regulation Y order, it was applied but assert that its application was lawful, Citicorp Brief at 27-32; CBCHA Brief at 27-32. We must address the Board’s contention first.
The Board is quite correct that the Citicorp order specifically examined each of the eight categories.of data processing the applicant proposed to provide. But some of those categories were only described with reference to (1) the data processing technology employed and (2) the nature of the data to be provided or processed, and the Board clearly held that the former “is not determinative of whether [the] activity is permissible,” Citicorp Order, 68 Fed.Res. Bull, at 507. The conclusion is therefore inescapable that, even in the Citicorp order, a data test was employed. With regard to timesharing services, for example, the applicant proposed to provide “data processing and transmission services for financial and non-financial institutions wherein the data being processed and transmitted are financial, banking or economic related.” J.A. C-54. The Board approved this, with the one change (which will be discussed further in another portion of this opinion) that “economic related” was altered to “economic.” Citicorp Order, 68 Fed.Res.Bull. at 507 n. 8. It is true that the application further provided some examples of the specific data processing uses to which the requested authority would be put:
Such packaged financial systems permit customers at various locations to obtain the benefit of Citicorp’s financial information systems and financial analysis expertise with respect to such applications associated with banking as financial modeling, loan analysis, accounting and bookkeeping, budget and profitability analysis, portfolio record-keeping and analysis, foreign exchange exposure, general ledger, bond analysis, international trade settlement, and economic forecasting.
J.A. C-54 to C-55. The Board’s opinion discussed several of these examples,7 concluding that they in fact represented services provided by banks or were so similar to such services that banking organizations are particularly well equipped to provide them. But that can only be understood as an attempted demonstration that its data test is as successful in concrete application as it is in theory.
It is even clearer that the Regulation Y order employs a data test. No more is needed to establish this than the text of the new regulation which the order adopted. Permissible activities of bank holding companies now include:
Providing to others data processing and data transmission services, facilities (including data processing and data transmission hardware, software, documentation or operating personnel), data bases, or access to such services, facilities, or data bases, by any technological means, if:
(i) The data to be processed or furnished are financial, banking, or economic, and the services are provided pursuant to a written agreement so describing and limiting the services;
(ii) The facilities are designed, marketed, and operated for the processing and *312transmission of financial, banking, or economic data; and
(iii) The hardware provided in connection therewith is offered only in conjunction with software designed and marketed for the processing and transmission of financial, banking, or economic data, and where the general purpose hardware does not constitute more than 30 percent of the cost of any packaged offering.
12 C.F.R. § 225.25(b)(7). The Regulation Y order states that the amendment it adopts “will make it permissible for bank holding companies to engage in the data processing and transmission services the Board has approved by order in the [Citicorp] case.” Regulation Y Order, 47 Fed.Reg. 37,369. But as the footnote accompanying that, statement shows, this is a reference to the eight broad technological categories of service {e.g., timesharing) rather than to the particular uses within those categories {e.g., budgeting, bookkeeping, accounting) which petitioners complain a data test avoids. And when the Regulation Y order incorporates by reference “[t]he Board’s findings on the permissibility of the services involved ... set forth in detail in the Board’s [Citicorp] order,” id., that incorporation, like the original findings themselves, must be for the purpose of exemplifying that in concrete application the data test will produce services closely related to banking.
As the insignificant nature of the textual changes in the portion of Regulation Y dealing with the current issue indicates, the present amendment does not fundamentally alter the approach of that regulation to data processing ■ services. With regard to the data test point it is significant, we think, that the Board’s staff recommendation proposing what in substance became the 1971 Regulation Y described its approach as follows:
In our view, the real issue is what kind of data should banking organizations be permitted to process. The technology employed is not the subject of the Act.
Accordingly, we recommend that the Board shift the emphasis of its proposal from the method of processing data to the kind of data being processed. Under this recommendation, bank holding companies would be permitted to process banking, financial, or other economic data, regardless of the tool used in the processing.
Legal Division Memo to Board of Governors (June 7, 1971), Applicant’s Exhibit M-2 at 6. While this is the staff’s description rather than the Board’s we think it an accurate representation of what the Board did in 1971 and perpetuated in the present orders.
The Board’s brief suggests that the ex: elusiveness and allegedly impermissible generality of the data test are avoided by the fact that
[bjefore any holding company may engage in data processing services, it must file an application with the Board and receive a Board order determining that the ‘public benefits’ test of section 4(c)(8) [of the Act] has been satisfied. 12 C.F.R. § 225.4(a), (b). Since data processing activities for a particular company are subject to Board approval upon specific application, the Board will have the opportunity to ensure that no bank holding company's activities exceed the statutory standard.
Respondent’s Brief at 32. It is indeed true, as noted in the Regulation Y Order, 47 Fed.Reg. 37,371, that the Board makes the separate “public benefits” determination required under the Act8 by examining individual applications rather than by regulation.9 But it would be unlawful, in the *313course of that case-by-case examination for that separate purpose, to deny an applicant the “closely related” qualification which the regulation generally confers. Such action would require amendment of the regulation—so that the limitation upon the data test which the Board’s brief puts forward amounts to no more than the possibility that the rule can be amended to forbid what it now impermissibly (according to petitioners) permits. Reliance upon such a possibility would validate every invalid rule.
We must confront, therefore, the stark question whether thé data test, as the determinant of whether data processing services are closely related to banking, is arbitrary or capricious. We think not. It would of course be preferable, from the point of view of accuracy alone, to make every “closely related” determination on a more narrow, specific, case-by-case basis— to ask, as petitioners would have the Board do, whether data processing for budget analysis, for bookkeeping, and for accounting each separately qualifies for the exemption. Indeed, it would be evén more accurate to get even more specific, and to ask whether budget analysis for manufacturing entities, budget analysis for retail sales entities, and budget analysis for personal service entities each separately qualifies. But the whole point of rulemaking as opposed to adjudication (or of statutory law as opposed to case-by-case common law development) is to incur a small possibility of inaccuracy in exchange for a large increase in efficiency and predictability. What the present controversy comes down to is simply whether there is reasonable assurance that the activities embraced within the data test—not all of which have been individually examined or even yet foreseen—will be closely related to banking under one or more of the broad National Courier tests. We think that there is.
As a theoretical matter, to begin with, the test is appealing. The record of this proceeding amply demonstrates, if any demonstration is needed, that banks regularly develop and process for their customers large amounts of banking, financial and economic data, and that they do so (and will presumably continue to do so) through the most advanced technological means. Once that is acknowledged, it is difficult to envision how any provision of data processing services dealing with data of that particular type would not meet at least the second of the National Courier tests:
Banks generally provide services that are operationally or functionally so similar to the proposed services as to equip them particularly well to provide the proposed service.
National Courier, supra, 516 F.2d at 1237. Perhaps it may not be in the public interest for them to provide one or another of such services, for anticompetitive or other reasons. But that relates to the “public benefits” determination, which is made case-by-case rather than in the amended Regulation Y, and whose resolution in the Citicorp order is not under challenge on this appeal..
In addition to its theoretical reasonableness, there is the fact that the Board, in the course of this proceeding, considered specific applications of the principle to various specific data processing uses proposed by Citicorp—finding all of them to be within one of the National Courier criteria, and many to be within the most rudimentary criterion that “[bjanks generally have in fact provided the proposed services,” id. We find adequate support in the record for those conclusions.10
*314It is significant that petitioners have not proposed any alternative to the data test, evidently demanding instead that each new data processing technology, and even each new use of data processing (such as budgeting, financial modeling, accounting) be separately examined and approved for its “closely relatedness.” In a field such as data processing, this will not do; the predictability of a rulemaking approach to the issue is vital. As the ALJ in the present proceeding found:
In view of rapid technological evolution, it is virtually impossible for any firm to predict the data processing services which will be available or offered in the coming years.
Recommended Decision at 10, J.A. B-81. And as the Comptroller of the Currency noted in a rulemaking proceeding dealing with the data processing activities of national banks:
In addition to being difficult to draft, acceptable specific lists or examples of permissible activities would be very difficult to keep current. This difficulty is due to the rapidly changing nature of the data processing field. The current rate of change virtually ensures that any acceptable list would quickly be rendered obsolete. Thus, any acceptable list or set of examples would require repeated updating. Perhaps more significantly, such a list may discourage the development by national banks of new data processing services which, although part of the business of banking, are not on a list of permitted activities or similar in nature to those in a set of examples of permitted activities.
Data Processing by National Banks, 47 Fed.Reg. 46,526, 46,529 (1982). We think these considerations amply justify the approach the Board has taken, an approach to which the petitioners have suggested no feasible alternative.
Economic Data
We turn next to an objection of petitioners that relates not to the data test as such, but to the types of data which the data test embraces.
The previous version of Regulation Y permitted bank holding companies to engage in “storing and processing other banking, financial, or related economic data, such as performing payroll, accounts receivable or payable or billing services.” 12 C.F.R. § 225.4(a)(8)(h) (1972) (emphasis added). In the combined adjudication-rule-making here at issue, the ALJ’s Recommended Decision accepted Citicorp’s proposal that this be expanded to include the processing and transmission of “banking, financial and economic related data.” Recommended Decision at 49, J.A. B-120 (emphasis added). The Board rejected that *315recommendation, agreeing with petitioners here that “economic related data” was “too broad a category and includes data that are not closely related to banking.” Regulation Y Order, 47 Fed.Reg. 37,369; Citicorp Order, 68 Fed.Res.Bull. at 507 n. 8. It adopted, however, an expansion of the previous language to “banking, financial and economic [as opposed to economic-related] data,” and permitted Citicorp to engage in such activities. Id. This expansion of prior authority was based upon its conclusion that “the record in this proceeding supports a finding that banks process economic data.” Regulation Y Order, 47 Fed.Reg. 37,369. See also Citicorp Order, 68 Fed.Res.Bull. at 507 n. 8 (“[t]he record supports a finding that banks engage in the processing and transmission of economic data”).
We agree. The testimony amply establishes the proposition that banks have long served their customers by developing and making available information regarding the national and international economy, including economic projections useful for investment decisions. The existing Regulation Y itself permitted the provision of “general economic information and advice, general economic statistical forecasting services and industry studies.” 12 C.F.R. § 225.4(a)(5)(iv) (1982). The current amendment of Regulation Y, which merely enables banks to make such information available in a new and more useful fashion, thus clearly meets the second of the National Courier criteria, and perhaps the first as well. Petitioners complain that “[njowhere in the record has there been any demonstration that banks have historically provided economic services even approaching the sophistication, scale and business mode approved by the Board in its Orders.” Petitioners’ Brief at 42. Perhaps so, but that is progress. It is not the purpose of the Bank Holding Company Act restrictions to confine banks to the same level, or crudeness, or technological simplicity of services previously provided—but merely to services of a closely related nature.
We also reject petitioners’ complaint that the Board has not defined “economic data.” Petitioners’ Brief at 35-36. In the context of these orders the meaning of the term is clear enough. It includes, as petitioners fear, “agricultural matters, retail sales matters, housing matters, corpo.rate profit matters and anything ‘of value in banking and financial decisions.’ ” Id. at 36 (quoting Citicorp Reply Brief before the Board at 13 & n. 17).
Timesharing Services
Until the 1970s, data processing was performed in the “batch” mode, that is, the application of a computer program was applied to data recorded (keypunched) on small cards, which were physically delivered to the computer. Batch processing requires expensive data pick-up and delivery, often involves lengthy overall turnaround time, and permits only one job to be performed at a time. Technological advances now enable data to be transmitted electronically, typically over a telephone line, to and from the central processing unit and the user’s terminal. That capability plus other technological advances permit “timesharing”—the simultaneous use of a computer by many users, each of whom can interact with the computer, i.e., ask yes-no questions and receive immediate responses. The AU found timesharing particularly well suited to economic, financial, and banking operations. Recommended Decision at 9, J.A. B-80.
Timesharing was approved in both the Citicorp and the Regulation Y Orders, 68 Fed.Res.Bull. at 507; 47 Fed.Reg. 37,369. Of course one of petitioners’ objections to this action is that the approval of this new service on the basis of the data test was impermissible; we have addressed that in a more general context above. Petitioners also claim, however, that timesharing, particularly when combined with customer use (which the orders permit) of applications software for such functions as modeling, forecasting and statistical analysis residing on holding company computers (as opposed to requiring customer creation and use of *316their own applications software) enables customers to use bank computer systems for “open-ended and general, non-financial services,” Petitioners’ Brief at 28, such as “race track handicapping or employee evaluation,” id. at 27.
This objection is unfounded.- To begin with, the Act places a limitation upon the services that bank holding companies can offer, not upon the uses to which others choose to put them. If the services proposed here were so well adapted to such uses unrelated to banking as employee evaluation that it could reasonably be thought that the services were offered for that purpose it would be one thing; but these services are no more invalidated by the mere possibility of such use, “unknown even to the holding company itself,”' Petitioners’ Brief at 29, than was the service which we approved in National Courier invalidated by the obvious possibility that a customer might include some nonfinancial material in the courier packages. And such a possibility (at most) is all that petitioners established. They brought forward not a single instance of actual use of such services for such purposes. Even their theoretical horrible was refuted by testimony asserting that it would be prohibitively expensive to use financial programs for such nonfinancial purposes. Nov. 17, 1981 Tr. at 1178-79, 1214-15.
Moreover, as undocumented and as unrealistic as the petitioners’ fears on this score appear to be, the Board nonetheless included provisions in its orders to calm this concern. In the Citicorp order the Board required that “all proposed data processing services provided by Citicorp to others outside the holding company for banking, financial and economic data must be provided pursuant to a written agreement so describing and limiting the services.” Citicorp Order, 68 Fed.Res.Bull. at 507. The approval conferred by the amended Regulation Y applies only when “the data to be processed or furnished are financial, banking, or economic, and the services are provided pursuant to a written agreement so describing and limiting the services.” 12 C.F.R. § 225.25(b)(7)(i). These agreements will be subject to scrutiny in connection with the Board’s case-by-case “public benefits” determination. In addition, the amended Regulation Y specifically requires that the offered facilities (by which the Board means “data processing and transmission hardware, systems software, documentation and operating personnel,” Citicorp Order, 68 Fed.Res.Bull. at 507 n. 12) be “designed, marketed, and operated for the processing and transmission of financial, banking, or economic data,” 12 C.F.R. § 225.25(b)(7)(h), and the Citicorp order makes it clear that this imposes the obligation to “take the technical steps necessary to ensure” this result. 68 Fed.Res.Bull. at 508. It is unthinkable that any more should be required.
Hardware
In both the Citicorp order and the Regulation Y order, the Board approved bank holding company provision of data processing hardware to their customers. Hardware is the equipment used in data processing systems, such as the mainframe computer, terminals, printers, memory devices, and the like. Software is the coded instructions which control the way data is processed, for example, individual programs. See Recommended Decision at 6-7, J.A. B-77 to B-78. For present purposes, data processing hardware provided by bank holding companies can be divided into two types, which were approved subject to different conditions and must be discussed separately.
Specialized hardware is specifically designed “to provide a permissible data processing or transmission service[], and is not likely to be used, to any significant extent, for nonfinancial purposes.” Citicorp Order, 68 Fed.Res.Bull. at 508 n. 14. The prime example is the automated teller machine (ATM), which is designed to execute banking transactions and has special security features appropriate to that purpose. The Board found that such hardware, when “offered only in conjunction with software designed and marketed for *317the processing and transmission of financial, banking, or economic data,” 12 C.F.R. § 225.25(b)(7)(iii), meets the third National Courier test. Citicorp Order, 68 Fed.Res. Bull, at 508 & n. 14, 509; Regulation Y Order, 47 Fed.Reg. 37,370. As noted earlier, that test reads as follows:
Banks generally provide services that are so integrally related to the proposed services as to require their provision in a specialized form.
516 F.2d at 1237. Petitioners make several attacks upon this finding, none of which seems to us well taken.
First, they assert that the third National Courier test could not conceivably apply because by its terms it pertains to “services,” which the provision of computer hardware is not. Petitioners’ Brief at 54. But surely National Courier is not to be interpreted in that fashion. Since the case dealt with services (courier services) it framed all its tests in those terms. But the object of the tests was to determine what are, within the words of the Act, “activities ... closely related to banking,” 12 U.S.C. § 1843(c)(8) (emphasis added), and they are obviously meant to apply, mutatis mutandis, to all activities, including not just services but also sale of material and equipment. It cannot seriously be thought that a bank holding company’s sale of checkbooks stands in a less favored position, under National Courier than its provision of courier services.
Next, petitioners argue that the third National Courier test is not met because it requires that the new services (or, as we have said, other activities) be integrally related to “traditional bank services” such as “check collection, credit extension and deposit gathering.” Petitioners’ Brief at 53. As far as we can determine, this limitation has been created out of whole cloth. It does not appear in National Courier, which only requires that the integrally related services be services that “banks generally provide.” If one were to hang upon this phraseology, the issue would presumably come down to whether the relevant integrally related services are data processing services in general (which banks now do generally provide) or the new types of data processing services authorized by these orders (at least the most futuristic of which banks now do not generally provide). Or it might be argued, as intervenor Citicorp does, Citicorp Brief at 61, that “services which banks generally provide” really means, as shown by language elsewhere in the National Courier opinion,11 “services which banks are authorized to provide.”
But at this point it begins to become foolish to devote one’s analytic energy to a parsing of the National Courier tests as though they were the statute itself, instead of referring to the underlying intent of the “closely related” requirement of which National Courier is, and only purports to be, a partial elaboration. That is to say, surely one of the most significant elements of the National Courier criterion is its prologue. After noting that the 1970 amendments to the Bank Holding Company Act (which slightly altered the text of the “closely related” provision and added the “public benefits” requirement) mean that the “closely related” test “no longer bears the full load, and may now be thought of as setting off as forbidden to banks those activities which are so clearly of a purely commercial nature that the predominantly adverse effects of a bank’s engaging in them may be presumed,” 516 F.2d at 1237, we continued:
Against this background, and reminding ourselves that the matter is one expressly committed by the statute to the Board, we think we owe considerable deference to the Board’s judgment that a particular activity is “closely related to banking.” Rather than define that term with any precision, therefore, we simply require that the Board go about making its “closely related” decision in a rea*318soned fashion consistent with the legislative intent.
The Board must, we think, articulate the ways in which banking activities and the proposed activities are assertedly connected, and must determine, not arbitrarily or capriciously, that the connections are close. As to what kinds of connections may qualify, at least the following seem to us within the statutory intent.'...
Id. (emphasis added).
Whether or not the provision of specialized computer hardware comes within the third National Courier test, we think the Board came to its decision in a reasoned fashion consistent with the legislative intent, and that that decision is not arbitrary or capricious. The Board noted, with adequate record support, that customers do not buy software and hardware but data processing. Citicorp Order, 68 Fed. Res.Bull, at 508-09; Regulation Y Order, 47 Fed.Reg. 37,370. Moreover, the Recommended Decision which the Board adopted discussed the impact of large scale integrated (LSI) circuits upon data processing technology, which now permit simple software functions, and in the future will permit more complex software functions, to be built into the hardware; and enable some of the computing function to be removed from the central computer and located in a customized “micro-processor” or “minicomputer” located at the point of use. Recommended Decision at 7-8, J.A. B-78 to B-79. This is the way some data processing is conducted now, and much more will be conducted in the future. Finally, the Recommended Decision also noted, with adequate record support, that a software producer marketing a “package” of software plus hardware is able to get a manufacturer's discount on the hardware, and is thus able to provide the full service to the purchaser at a more competitive price. Recommended Decision at 8, J.A. B-79. On the basis of these factors the Board concluded that “the activity of providing software without the corresponding authority to provide related hardware is of questionable economic feasibility.” Citicorp Order, 68 Fed.Res.Bull. at 508 (footnote omitted). It is not the economic infeasibility in itself that impresses us, but the underlying cause of that infeasibility, recognized by the Board—that in both market contemplation and technological reality the service in question is a unitary one. The issue boils down, as Citicorp suggests, Citicorp Brief at 58, to almost “a tautology.” In effect, to authorize the provision of banking, financial and economic data processing is to authorize the provision of banking, financial and economic hardware and software. Whether it be considered “integrally related” to the authorized service or, perhaps more realistically, simply part of it, the provision of specialized hardware is reasonably included.
General purpose hardware is, as the name would suggest, hardware that is designed to perform data processing functions in addition to banking, financial and economic. The Board approved bank holding company provision of this hardware, subject to the same condition that it be offered only in conjunction with banking, financial or economic data software, and subject to the additional condition that it “not constitute more than 30 percent of the cost of any packaged offering.” 12 C.F.R. § 225.25(b)(7)(iii).12 The Board acknowledged that the sale of general purpose hardware “is not itself an activity that is closely related to banking,” but found that with the limitations the Board imposed it would be permissible as “incidental” to the provision of permissible data processing services, Citicorp Order, 68 Fed.Res.Bull. *319at 508; see also Regulation Y Order, 47 Fed.Reg. 37,370.
The notion of permissibility of “any incidental activities that are necessary to carry on” activities closely related to banking has been embodied in Regulation Y since 1971. 12 C.F.R. § 225.21(a)(2). We approved it in National Courier, noting that “[i]n enumerating the activities that could be carried on, [Congress] certainly could not have meant to forbid engagement in such other ‘incidental’ activities as were reasonably necessary to carrying out those that were enumerated.” 516 F.2d at 1240. We disapproved its application to bank courier handling of nonfinancial materials for the following reason:
The justification for the ‘incidental’ courier services ... is not, as far as we can tell, that the carriage of non-financially related material is in any way necessary to the successful operation of the courier service affiliates. Rather, it is that the provision of such service would serve ‘the convenience of the public.’
Id. (footnote omitted). Here, by contrast, the Board’s justification was precisely that the provision of banking, financial and economic data processing services could not be successful unless they were offered in conjunction with the necessary hardware. Even where the hardware itself was not required to be specialized or to contain any software—so that the hardware could not be regarded as in itself the provision of banking, financial or economic data processing—the nature of the data processing market was such, the Board found, that hardware and software were a single package. The provision of data processing software in isolation, even where no specialized or software-inclusive hardware was required, was “of questionable economic feasibility.” Citicorp Order, 68 Fed.Res.Bull. at 508, referred to in Regulation Y Order, 47 Fed.Reg. 37,370. There was record evidence to support this conclusion, and, as the excerpt from National Courier above suggests, the conclusion is sufficient to support the Board’s action. Indeed, the Fifth Circuit Court of Appeals has permitted an even lesser connection to sustain bank holding company provision of liability insurance as “incidental” to the closely-related-to-banking activity of providing property damage insurance for the collateral in bank loans. Alabama Association of Insurance Agents, supra. While that court found evidence to sustain the proposition that, “from the consumer’s point of view, packaged property damage and liability policies are more desirable than the same policies separately sold,” 533 F.2d at 245, it did not make the further finding that the sale of the one had been shown to be “of questionable economic feasibility” without the sale of the other. In that respect, Alabama Association goes further than we were prepared to go in National Courier, though there was, it must be acknowledged, the added factor that liability insurance in itself to some degree increased the security of the banks’ loans. In any case, we think that the reasoning of both National Courier and Alabama Association supports the proposition that the economic necessity of offering a service that is not closely related to banking in order to sell another service that is, justifies the provision of the one as “incidental” to the other.
There is an obvious limitation upon this principle: At some point the tail begins to wag the dog. If it should be found, for example, that data processing services cannot be sold in an economically feasible manner without manufacturing data processing hardware; and if the banks’ profits from the latter should exceed their profits from the former; surely the provision of data processing services would be incidental to hardware manufacture rather than viceversa. But the Board has adequately taken that limitation into account, by specifying, as described above, that the cost of the hardware (including both specialized and general purpose hardware) cannot exceed 30 percent of the cost of the package. The Board derived this figure by noting from the record that hardware costs for the data processing industry as a whole represent about 25 percent of total costs, Citicorp Order, 68 Fed.Res.Bull. at 509,—so that a “package” in which hardware accounted *320for about that percentage of the cost could reasonably be considered primarily a sale of data processing rather than a sale of processing hardware. We think that a reasonable way to proceed, and we cannot say that an element of a permissible service which constitutes less than one-third the cost of that service is not “incidental.”
III. Procedural Objection
Petitioners raise a procedural objection to the Citicorp order. They claim that the Board approved Citicorp’s request without stating its holding or reasoning regarding four components of that request, in violation of the APA’s requirement that agency orders in formal proceedings address “all the material issues of fact, law, or discretion presented on the record.” 5 U.S.C. § 557(c). They complain that the Board did not address the sale of applications software products, the provision of facilities management services, the sale of turnkey systems products, and the provision of software-related services.
We do not agree. What is a “material issue” that must be separately addressed obviously depends upon one’s view of the logical or operational differences between various aspects of the problem in question. If an appellant before this court chooses to analyze a freedom-of-speech issue by discussing verbs, nouns, adjectives and pronouns separately, our decision will not necessarily disregard a “material issue” if it ignores this distinction. So also here, petitioners cannot insist upon imposing their structure of analysis upon the Board. The characteristic feature of petitioners’ approach throughout this proceeding has been the insistence that each separate element of the provision of data processing services be analyzed separately; this is simply contrary to the Board’s view, which we have found reasonable, that “the particular technology by which a data processing activity is provided is not determinative” and the “permissible data processing activities may be provided by any technologically feasible method.” Regulation Y Order, 47 Fed.Reg. 37,369. The Board’s decision in the Citicorp order clearly disposed of the various elements petitioners complain of here, whether or not it addressed each of them separately; and the reasons for its dispositions are "fully set forth. That is all that reason or even the APA requires.
Applications software, which petitioners define as software that provides instructions for the performance of specific tasks, is part of the software included in on-site data processing, electronic funds transfer, and authentication programs, all of which were approved and fully discussed. Facilities management by employees of the bank holding company and “turnkey” systems (that is, systems including both hardware and software in which control is turned over to the purchaser) were described by the AU as the two modes of providing on-site data processing, Recommended Decision at 7-8, 12-13, J.A. B-78 to B-79, B-83 to B-84; that activity, in all of its modes, was approved and fully discussed. Citicorp Order, 68 Fed.Res.Bull. at 508-09. Finally, as to software related services, such as consulting and systems analysis: Citicorp’s application did not include these services as a separate activity, and neither we nor, judging by its brief, the Board, see Board’s Brief at 63, regards the operation of a consulting service to give advice regarding data processing to be the same thing as data processing, which is all that the orders approve. Of course some consultation and advice will be necessary as an incident to the sale of data processing, but authorization for that was neither separately sought nor is separately required, since it is obviously included within Regulation Y’s existing authorization for “incidental activities,” discussed above. We therefore reject the procedural objection to the Board’s action.
* * * # # *
The challenge which the Board faced in the current proceeding was the perennial regulator’s problem of adjusting the outlines of regulation to accommodate technological change. Virtually unprecedented difficulties of adjustment are presented by *321the torrent of recent scientific innovation in the fields of electronics and telecommunications, which possess a peculiar capability to destroy the categories of enterprise upon which regulation is based. Communications merges into data processing and data processing into banking. Intervenor California Bankers Clearing House Association perceptively observes, CBCHA Brief at 29 n. 20, that this proceeding parallels the Federal Communications Commission’s Second Computer Inquiry, the order resulting from which we upheld in Computer and Communications Industry Association v. FCC, 693 F.2d 198 (D.C.Cir.1982).
The record of the present proceeding displays a careful and conscientious effort by the Board to cope with these difficulties. We are not inclined to complicate its task further by attempting to exercise close and necessarily inexpert supervision of its judgments. That would be particularly inappropriate under a governing statute such as this one, which commits it to the Board to apply a standard of such inherent imprecision (“closely related to banking”) that a discretion of almost legislative scope was necessarily contemplated. If there is a problem in such broad delegation, it would assuredly not be solved by effectively taking the delegation from the Board and placing it in our own hands. Having assured ourselves that the Board has acted reasonably, consistently and with procedural regularity in giving content to the statutory standard, our task is at an end.
In addition to the major points discussed above, we have considered the other objections urged by petitioners and find them insubstantial. For these reasons, the petitions are
Denied.
. 12 C.F.R. § 225.25(b)(7) (1984) is the slightly revised successor to 12 C.F.R. § 225.4(a)(8) (1983). Since the revisions are minor and do not affect this opinion, all references are to the 1984 Code.
. In SIA, the Court noted that the Board's "factual findings ... are substantially supported by the record," 104 S.Ct. at 3009, but as will appear below that may be a description of the requirements of the arbitrary or capricious standard.
. 12 U.S.C. § 1848 reads as follows:
Any party aggrieved by an order of the Board under this chapter may obtain a review of such order in the United States Court of Appeals within any circuit wherein such party has its principal place of business or in the Court of Appeals in the District of Columbia, by filing in the court, within thirty days after the entry of the Board’s order, a petition praying that the order of the Board be set aside. A copy of such petition shall be forthwith transmitted to the Board by the clerk of the court, and thereupon the Board shall file in the court the record made before the Board, as provided in section 2112 of title 28. Upon the filing of such petition the court shall have the jurisdiction to affirm, set aside, or modify the order of the Board and to require the Board to take such action with regard to the matter under review as the court deems prop*307er. The findings of the Board as to the facts, •if supported by substantial evidence, shall be conclusive.
. National Courier's failure even to refer to § 1848 was noted in a later decision of this court which applied the "substantial evidence" provision of another statute to informal (as opposed to formal) adjudication. Aircraft Owners and Pilots Association v. FAA, 600 F.2d 965, 971 n. 26 (D.C.Cir.1979). Our decision today' explains why National Courier, in ignoring the provision, and Aircraft Owners, in applying it, were not reaching different results. Aircraft Owners acknowledged this possibility in a footnote, 600 F.2d at 971 n. 28.
. Subsection 706(2) reads as follows:
[The reviewing court shall—]
(2) hold unlawful and set aside agency action, findings, and conclusions found to be—
(A) arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law;
(B) contrary to constitutional right, power, privilege, or immunity;
(C) in excess of statutory jurisdiction, authority, or limitations, or short of statutory right;
(D) without observance of procedure required by law;
(E) unsupported by substantial evidence in a case subject to sections 556 and 557 of this title or otherwise reviewed on the record of an agency hearing provided by the statute; or
(F) unwarranted by the facts to the extent that the facts are subject to trial de novo by the reviewing court.
. See, e.g., the quotation from K. Davis, supra, page 684. The reason for this reputation, one may surmise, is that under the APA the substantial evidence test applies almost exclusively to formal adjudication (formal rulemaking is rare), which is, by contrast to rulemaking, characteristically long on facts and short on policy— so that the inadequacy of factual support is typically the central issue in the judicial appeal and is the most common reason for reversal.
. There is no doubt that the Board understood them to be, as they were phrased, merely examples of the data processing uses that would be made available. Where its opinion meant to limit the applicant to the specific uses that were recited as currently contemplated, it said so. See the portion of the Citicorp order dealing with home banking services, 68 Fed.Res.Bull. at 509-10, the conclusion of which is reflected in the Regulation Y Order, 47 Fed.Reg. 37,370.
. Although the language of the statute is subject to the interpretation that the "public benefits" finding is merely part of the finding that the activity is "so closely related to banking ... as to be a proper incident thereto,” 12 U.S.C. § 1843(c)(8), it is now well established that they are “two separate determinations," SIA, supra, 104 S.Ct. at 3006.
. This is required by the Board's regulations, 12 C.F.R. § 225.21(a)(2). The Supreme Court has intimated that it is also required by law, saying that the Board "must [make the determination] *313on a case-by-case basis,” and citing for that proposition a portion of the legislative history, H.RXonf.Rep. No. 1747, 91st Cong., 2d Sess. 16-18, reprinted in 1970 U.S.Codf. Cong. & Ad. News 5561, 5567-69. SIA, supra, 104 S.Ct. at 3006.
. The petitioners’ challenge to the evidentiary support for the Board’s approval of budget and accounting services, which they present as a separate challenge to the orders, Petitioners’ Brief at 46-47, is in reality merely a challenge to this pragmatic validation of the data test. We digress briefly to address that challenge. The Board concluded that ”[t]he record of this proceeding demonstrates that banks in fact provide data processing services that assist banks in performing their own credit functions and that *314perform financial bookkeeping and accounting operations for other businesses.” Citicorp Order, 68 Fed.Res.Bull. at 507. While this specific finding was made only in the Citicorp order, the Regulation Y order incorporated it by reference when it noted that ”[t]he Board’s findings on the permissibility of the services involved are set forth in detail in the Board’s Citishare order.” Regulation Y Order, 47 Fed.Reg. 37,369.
Neither the Board nor the intervenors cited us any portion of the record that supports this finding. The reason, apparently, is that with the exception of the provision of economic data (discussed later in this opinion) the evidence in the present proceeding focused upon the technological means of delivering data processing services, rather than upon the subjects those services covered. The latter had been extensively considered in the 1971 rulemaking that produced the prior version of Regulation Y—in which proceeding, we may note, ADAPSO participated. Thus, for evidence to support the Board’s statement on this score its brief refers to an American Bankers Association survey introduced in the 1971 proceeding. Respondent's Brief at 43 n. 46, 48 (citing Supplemental Statement of American Bankers Association, Apr. 30, 1971, Appendix I at 4). We have been unable to find the survey in the record. We think it unnecessary to examine or rely upon the survey itself, however, since the Applicant’s Exhibits include a June 7, 1971 Memorandum to the Board from its Legal Division which provides information based upon the survey. See page 686, supra. It shows that since the 1950s banks have performed such functions as payroll service, account reconciliation, sales analysis and inventory analysis, which seem to us unquestionably bookkeeping' and accounting functions. While that information is not extensive (as the full survey itself presumably is) we think it adequate, in the absence of any contradiction regarding such functions from petitioners, to provide the necessary record support.
. There is no question that courier service for data processing material has been found “closely related to banking" not because of its connection to banking per se but because of its relation to financial data processing, an activity banks are permitted to engage in.
516 F.2d at 1239.
. This means cost to the bank holding company. See Regulation Y Order, 47 Fed.Reg. 37,-370. The Citicorp order, and the text of the Regulation Y order (as opposed to the language of the amended regulation) say "approximately 30 per cent.” Citicorp Order, 68 Fed.Res.Bull. at 509; Regulation Y Order, 47 Fed.Reg. 37,370 (emphasis added). The Citicorp order also concedes that "in individual cases, the cost of hardware may exceed the percentage limitation,” so long as hardware sales do not “consistently exceed” it. Citicorp Order, 68 Fed.Res.Bull. at 509.