United Stationers, Inc. (USI) sued the United States after the IRS failed to take action on its claim for a qualified research tax credit. The district court ruled for the government, largely adopting the recommendations of a magistrate judge. USI appeals that judgment. The issue is whether USI is entitled to a qualified research credit under Internal Revenue Code § 41 with respect to its development of certain computer software programs. USI has not shown its entitlement to the qualified research credit and we therefore affirm.
I. Background
The facts of this case are not disputed. The parties submitted detailed stipulations with attached exhibits, including summaries of the software development projects at issue. These stipulations and summaries form the basis of the recitation that follows.
USI is the nation’s largest wholesaler of office supplies. In response to both rapid corporate growth and industry changes in the 1980s, it attempted to correct operational inefficiencies by automating and computerizing its business. To this end, it purchased a software package, DCS, from a consulting firm for use in developing eight separate, but related, software programs. USI chose DCS for some of its file structure and data access segments; it could not have used DCS as purchased, nor could it have purchased any other software to meet its needs without modification. Instead, USI used parts of DCS as building blocks for designing application codes specific to its needs.
USI developed all eight of the software projects at issue during the fiscal year 1988. Each performs a different operational function although several work together to correct a particular inefficiency. The first program, for example, the Document Retention and Retrieval System (DRRS), maintains files, customer histories and purchase records; it is, in short, a centralized invoice data base. Customers can access certain information in DRRS and request reprints of invoices or sales histories. DRRS is integrated with the second program, the Central Invoice Project (CIP), which furnishes a paperless invoicing and record-keeping system. The goal of CIP is to print, distribute and retain all invoices at USI’s corporate head*442quarters in order to correct inefficiencies associated with USI’s far-flung warehousing and distribution network. The third project is Concept 90, a computerized marketing program developed to address increased competition from volume discounters. USI hoped that by quickly responding to competitor pricing but carefully controlling its own price concessions it could increase sales volume on certain items. Unilink, the fourth project, is a computer-to-computer order entry system through which customers can enter orders on their own computers and transmit the orders directly to USI. On Unilink, customers can review previous orders and are immediately notified if an item is out of stock. USI developed a related project, called Distributed Order Entry, to allow inter-divisional billing in the event that the division receiving the order was out of stock. USI also developed the Facility Database Project, FDP, the fifth program, to replace its outdated inventory program and as an aid in forecasting and replenishing inventories. FDP creates a hierarchy of inventory data, with information about each region’s inventory first, specifics on each item in that inventory second and pricing possibilities for each item last. The sixth project was the Automated Inbound Shipment Processing Program. As its name implies, USI developed this program to replace its manual warehousing procedures with automated incoming and outgoing shipping procedures. The seventh project, the Receiving System, is integrated with this Automated Inbound Shipment Processing Program. The eighth project is called the Forecasting/Replenishment Application Project. It creates a more timely and accurate method of inventory control. This project replaced a system based largely on an individual buyer’s experience in anticipating product demand.
The project summaries also describe the risks and benefits of each program. Most of the summaries include a general statement about the uncertainty at the outset of achieving the projects’ anticipated operational efficiencies or economic benefits. For example, the CIP summary states that “there was no guarantee at the outset of [CIP] that [USI], in centralizing the invoicing functions ... would be able to both provide customers with accurate invoices on a timely basis and have easy access to that invoice data after they had been sent to customers.” Similarly, Concept 90, the marketing program, involved “operational risk.” And another project summary states that “[t]here was no guarantee that the FDP approach would either perform as expected or provide an adequate return on the Company’s investment.” Most, but not all, of the project summaries contain similar general risk warnings, presumably as estimated at the commencement of the project. Some summaries also include assessments at completion of the success of each project. For example, DRRS “generated cost savings to [USI] as well as improving its ability to service its customers.” Other projects were less successful, like Concept 90 (“the economic benefit ... was far less than originally anticipated”) and the Forecasting/Replenishment Application (“this system has worked fairly well”). Again, most but not all of the summaries include such a general assessment of the project’s success. None of these eight programs have been resold to third parties by USI.
In its original tax returns for the fiscal year ended August 31, 1988, USI claimed a deduction under Code § 174 of $156,457 of expenses related to the development of these eight programs. The IRS allowed this deduction. In an amended return for the same year, USI claimed a credit under § 41 of the same amount for qualified research activities based upon the IRS’s allowance under § 174. The IRS took no action on this claim. USI filed this suit six months later to recover a refund based on the claimed credit.
The district court referred the case to a magistrate judge, who issued a report determining that USI’s development of the programs did not constitute qualified research and that USI was therefore not entitled to the claimed refund. See United Stationers, Inc. v. United States, No. 92 C 6065, 1997 WL 159526 (N.D.Ill. March 28,1997) (United Stationers I). USI filed objections to this report. The district court agreed (with a minor deviation) with the magistrate judge’s report. See United Stationers, Inc. v. United States, 982 F.Supp. 1279 (N.D.Ill.1997) (United Stationers II). The district court
*443UNITED STATIONERS, INC. v. U.S. 443
Cite as 163 F.3d 440 (7th Cir. 1998)
ruled that the programs did not qualify for § 41’s qualified research credit because USI neither “discover[ed] information which is technological in nature,” id. at 1284, nor did so through a “process of experimentation.” Id. at 1286. The district court further held that, even if the programs could meet the § 41 general requirements, they fitted within an enumerated exclusion from the allowable credit because they were for “internal use.” Id. at 1287. Finally, the district court found that the programs did not qualify under an exception to this internal use exclusion. Id. at 1287-88. USI appeals these holdings.
II. Discussion
[1-3] In this case, as with all claimed tax credits, the taxpayer bears the burden of showing entitlement to the credit. See, e.g., Interstate Transit Lines v. Commissioner of Internal Revenue, 319 U.S. 590, 593, 63 S.Ct. 1279, 87 L.Ed. 1607 (1943); United States v. Wisconsin Power & Light Co., 38 F.3d 329, 337 (7th Cir.1994). To the extent entitlement turns on the legal question of interpretation of § 41, we review the district court’s legal conclusions de novo. See, e.g, Resolution Trust Corp. v. Gallagher, 10 F.3d 416, 418 (7th Cir.1993). As to questions of fact, including issues of evidentiary sufficiency, we review for clear error. See, e.g., Wisconsin Poiver & Light, 38 F.3d at 337.
A. The Qualified Research Tax Credit
Section 41, the Credit for Increasing Research Activities, allows taxpayers a credit for certain expenses incurred conducting qualified research. Subsection 41(d) defines qualified research in pertinent part as follows:
(d) Qualified research defined.&emdash;For purposes of this section&emdash;
(1) In general.&emdash;The term “qualified research” means research&emdash;
(A) with respect to which expenditures may be treated as expenses under section 174,
(B) which is undertaken for the purpose of discovering information&emdash;
(i) which is technological in nature, and&emdash;
(ii) the application of which is intended to be useful in the development of a new or improved business component of the taxpayer, and
(C)substantially all of the activities of which constitute elements of a process of experimentation for a purpose described in paragraph (3).
Such term does not include any activity
described in paragraph (4).
26 U.S.C. § 41(d)(1). This provision sets out a four-pronged test for qualified research. First, the research must have qualified as a business deduction under § 174. See id. § 41(d)(1)(A). Second, the research must be undertaken to “discover information which is technological in nature.” Id. § 41(d)(1)(B)®. Third, the taxpayer must intend to use the information to develop a new or improved business component. Id. § 41(d)(l)(B)(ii). Finally, the taxpayer must pursue a “process of experimentation” during substantially all of the research. Id. § 41(d)(1)(C).
Only two of these requirements are at issue here. A § 174 deduction has been allowed, satisfying the first requirement, and the government concedes the third requirement, involving usefulness for a new or improved business component. We will address the two other requirements in turn.
1. Discover Information Which is Technological in Nattire
Section 41(d)(1)(B)® requires research to be undertaken “for the purpose of discovering information which is technological in nature.” The district court held that, although the programs were “technological in nature,” USI did not develop the programs to discover information. United Stationers II, 982 F.Supp. at 1284. Citing dictionary definitions and the legislative history of § 41, the district court instead found that “Stationers [had] merely applied, modified, and at most, built upon, pre-existing, technological information already supplied to it.” Id. USI now claims that the district court’s interpretation of this prong comports with neither the Code’s plain language nor its legislative history. While we do not necessarily adopt the district court’s view that the issue turns on
*444the breadth of the definition of the word “discover,” we believe that the terms of the requirement in question have not been met.
USI cites a dictionary which defines “discover” as “to make known (something secret, hidden, unknown, or previously unnoticed)” or “to obtain for the first time sight or knowledge of.” Appellant’s Br. at 15 (citing Webster’s Third New International Dictionary, Unabridged 647 (1993)). USI invites us to end our definitional inquiry here, content with an interpretation which equates discovery with mere newness. We decline this invitation. Congress attempted to give further meaning to § 41(d)(l)(B)(i) and we can look to legislative history for guidance. See, e.g., Sundstrand Corp. v. Commissioner of Internal Revenue, 17 F.3d 965, 967-68 (7th Cir.1994).
The Conference Report accompanying the 1986 amendments states that research that discovers information which is technological in nature must rely on “principles of the physical or biological sciences, engineering, or computer science.” H.R. Conf. Rep. No. 99-841, at 11-71 (1986). Specifically, “[r]esearch does not rely on the principles of computer science merely because a computer is employed. Research may be treated as undertaken to discover information that is technological in nature, however, if the research is intended to expand or refine existing principles of computer science.” Id. n. 3. Thus Congress set a demanding standard. Not just any computer programming would qualify for a § 41 credit; qualifying research must go beyond the current state of knowledge in that field — expand or refine its principles. See Norwest Corp. v. Commissioner, Tax Ct. Rep. (CCH), ¶ 52,-758, at 4653-54, 1998 WL 341634 (U.S. Tax Ct. June 29,1998). Therefore, in the context of § 41(d)(l)(B)(i) discovery demands something more than mere superficial newness; it connotes innovation in underlying principle. See id.
This understanding of “discovery” also brings the requirements for the credit in line with the purposes of both the original statute in 1981 and the amendments in 1986. Congress first created a research and experimentation tax credit in 1981 as part of the Economic Recovery Tax Act of 1981 (ERTA). ERTA adopted § 174’s definitions of “research” and “experimentation” but failed to specifically indicate how the credit applied to computer software development.1 Congress hoped that ERTA would energize the overall economy and specifically would stimulate business investment in technological research. See S.Rep. No. 97-144, at 76-77 (1981), U.S.Code Cong. & Admin.News 1981, pp. 105,181-182; H. Rep. No. 97-201, at 111 (1981). In 1986, in response to taxpayer abuse, Congress amended the research tax credit provision to include an express statutory definition of qualified research. The 1986 amendments restricted the definition of qualified research to that which “discover[s] information which is technological in nature” in an attempt to limit the deduction to only high technology research.2 Congress clearly intended, then, that qualifying research pass a high threshold of innovation and be of broad effect. See Norwest, ¶ 52,758, at 4653-54, 1998 WL 341634; see also TSR, Inc. v. Commissioner, 96 T.C. 903, 916-20, 1991 WL *445110430 (1991).3
We believe that our interpretation&emdash; as well as that of the district court&emdash;comfort-ably fits the Code’s language and its legislative history and compels the conclusion that USI’s programs did not discover information which is technological in nature. USI’s projects did not refine or expand the principles of computer science. Instead, they modified an existing software program in an attempt specifically to improve USI’s inventory control system. Further, although these applications were new to USI and helped it eliminate various operational inefficiencies, they were not of broad economic effect. USI accuses the district court of imposing a “revolutionary” standard where Congress intended no more than “evolutionary” research to qualify. USI’s amicus similarly cautions against raising the bar of the research tax credit too high. We are respectful of these warnings but are no less convinced of our conclusion. Wherever the line is drawn, these projects fall short. The more or less routine modification of a commercially-available software package to apply to USI’s inventory control computers simply does not measure up to the demanding requirements of § 41(d)(1)(B)(i). See H. Rep. No. 99^26, at 180 (1985); compare Nonvest, ¶ 52,758, at 4663-65, 1998 WL 341634 (examples of computer programs that meet the high standard) with id. at 4667, 4668, 1998 WL 341634 (examples of computer programs that do not meet the high standard).
2. Process of Experimentation
Section 41(d)(1)(C) provides that qualified research involve “substantially all of the activities which constitute elements of a process of experimentation.” 26 U.S.C. § 41(d)(1)(C). The district court, again looking to dictionary definitions and legislative history for guidance, held that this test requires qualifying research to have significant design uncertainty from its outset. See United Stationers II, 982 F.Supp. at 1285. The court then found USI’s claim of uncertainty “not persuasive:” “while the aspired benefits of the projects were in doubt, the development of the means that would allow [USI] to potentially achieve those benefits was not.” Id. (citation omitted). USI now challenges both the district court’s legal interpretation of this process of experimentation test and its factual findings regarding design uncertainty in the eight projects. We reject both challenges.
USI again begins with its plain reading of the subsection. According to USI’s dictionary, an experiment is “... a test, trial, or tentative procedure, an act or operation for the purpose of discovering something unknown or of testing a principle, supposition, etc.” Appellant’s Br. at 23 (citing Random House College DICTIONARY 465 (rev. ed.1984)). The projects fit squarely within this definition, USI claims, because software development involves debugging, a process of testing and correcting computer programs. We find this argument unpersuasive. Debugging programs amounts simply to fine-tuning computer operating instructions. Although we are reluctant to establish bright-line rules&emdash;§ 41 eases will always be highly fact-intensive&emdash;we think that a process of experimentation involves something more than simply debugging a computer program. See Norwest, ¶ 52,758, at 4669, 1998 WL 341634 (computer programming project did not qualify because it “merely required conducting good coding and eliminating bugs through testing”).
The legislative history of § 41(d)(1)(C) bolsters this conclusion. The Conference Report explains the meaning of the process of experimentation test:
The term process of experimentation means a process involving the evaluation of more than one alternative designed to achieve a result where the means of achieving that result is uncertain at the outset. This may involve developing one *446or more hypotheses, testing and analyzing those hypotheses (through, for example, modeling or simulation), and refining and discarding the hypotheses as part of a sequential design process to develop the overall component.
Thus, for example, costs of developing a new or improved business component are not eligible for the eredit if the method of reaching the desired objective (the new or improved product characteristics) is readily discernible and applicable as of the beginning of the research activities, so that true experimentation in the scientific or laboratory sense would not have to be undertaken to develop, test, and choose among viable alternatives .... [Ejngineers who design a new computer system, or who design improved or new integrated circuits for use in computer or other electronic products, are engaged in qualified research because the design of those items is uncertain at the outset and can only be determined through a process of experimentation relating to specific design hypotheses and decisions as described above.
H.R. Conf. Rep. No. 99-841, at 11-72 (1986). USI would have us hold that this Report dictates a per se credit for software development. This suggestion, however, ignores the Report’s mention of only systems design (not software design) as well as its focus on overcoming design uncertainty through a systematic — almost scientific — methodology. The Report suggests that qualifying research must from its outset involve some technical uncertainty about the possibility of developing the product. See Norwest, ¶ 52,758, at 4655, 1998 WL 341634 (discussing the “required uncertainty”); cf. TSR, 96 T.C. at 920-21,1991 WL 110430.
USI and amicus again caution that this articulation of the relevant standard is too narrow. We are not persuaded, however, and conclude that USI’s projects did not involve a process of experimentation as envisaged by Congress. The eight projects involved the modification of a commercially-available software package to fit USI’s par-tieular operational needs. Most of the project summaries include general statements about the uncertainty of actually realizing the anticipated operational, efficiencies and resultant economic benefits. None of the summaries, however, describes any technical uncertainty about actually developing the programs. Compare Norwest, ¶ 52,758, at 4670, 1998 WL 341634 (programming project did not constitute qualified research because, in part, it did not involve technical risk). Nor were there any doubts about the ability of computers to perform the invoicing, billing and marketing tasks. That is, there was simply no technical uncertainty from the outset. We therefore conclude that USI’s development of the eight programs did not involve a process of experimentation and that the district court did not clearly err in concluding that the projects did not involve the level of uncertainty necessary to clear this hurdle.
B. The Internal Use Exclusion
Section 41(d)(1) excludes from the qualified research credit any computer software developed “primarily for internal use by the taxpayer.” 26 U.S.C. § 41(d)(4)(E). For lack of other authority, the district court consulted the legislative history and proposed Treasury regulations to determine what “internal use” means. See United Stationers II, 982 F.Supp. at 1286. The district judge employed a “totality of the circumstances” test and concluded that USI’s programs “were the type of internal use software that Congress specifically sought to exclude from the § 41 research credit.” Id. at 1287. USI, of course, contests this conclusion and constructs two elaborate arguments in an attempt to rescue the projects from this internal use exclusion. Neither argument, despite its admirable ingenuity, is particularly compelling.
Again, nothing in the Code or in case law4 explicitly defines internal use. We therefore plunge again into the muddy waters of legislative history. USI points us to 1981 Congressional records to support its reading of the exclusion. Only software in*447tended exclusively for internal use, USI argues, should be excluded because the original intent of the research credit was to stimulate the economy; therefore, any program that provides a stimulus to or has an impact on the economy, however indirect or insubstantial, should escape the internal use exclusion. This proposed analysis sweeps too broadly. Congress intended the original credit to reward taxpayers for a significant contribution to the store of public technological knowledge. See S.Rep. No. 97-144 (1981); H. Rep. No. 97-201 (1981). To avoid this exclusion, then, a taxpayer claiming a credit for developing software must to an appropriate extent make the program available to the public. Something more than a speculative or attenuated impact on the economy is required.
USI also pins its hope on the 1986 Conference Report which explicitly states that software development projects would constitute internal use software where they are “used internally, for example, in general and administrative functions (such as payroll, bookkeeping, or personnel management) or in providing noncomputer services (such as accounting, consulting, or banking services).” H.R. Conf. Rep. No. 99-841, at 11-73 (1986). USI relies on this language to urge us to draw elusive distinctions between the provision of general and administrative services and a taxpayer’s core revenue-generating activities. Core business activities have an external impact and should not be subject to the internal use exclusion, USI asserts. It maintains that the projects at issue here— inventory controls and the like — are integral to its core revenue activities as a wholesale distributor. USI therefore concludes that its eight projects cannot be internal use software.
We cannot accept this analysis. USI seems to be asking us to view cost accounting methodology (specifically, the difference between general overheads and directly assigned costs) as a determinative principle in identifying internal use. In effect, USI appears to be arguing that, if the programs involve activities that directly impact parties outside the taxpayer, they are not for internal use. As we stated before, however, this formulation does not fully capture the idea of contributing to the store of public technological knowledge. In any event, we agree with the district court that all the relevant facts— the totality of the circumstances — must be taken into account. See United Stationers II, 982 F.Supp. at 1286-87; see also Prop. Treas. Reg. § 1.41-4(e)(4), 62 Fed.Reg. 81, 83 (Jan. 2, 1997) (Credit for Increasing Research Activities). And, on the record before us, the district court’s conclusion that the software development projects were for internal use is not clearly erroneous. USI developed all of the software to help it track its huge inventory. Even the two programs to which customers have limited access, DRRS and Unilink, were developed primarily for use by USI in streamlining its operations. The services these software programs expedite — marketing, ordering, invoicing, shipping, receiving, pricing, etc. — even though they may have a direct impact on customers, suppliers and other third parties do not rescue the programs from the internal use exclusion.
C. Innovative and Risky Exception to the Exclusion
The legislative history of § 41 introduces an exception to the internal use exclusion which would permit a qualified research credit for the development of internal use computer software that meets the general requirements for the credit that we have already discussed, is innovative, involves risk and results in software that is not commercially available. See H.R. Conf. Rep. No. 99-841, at 11-73 (1986); see also Prop. Treas. Reg. § 1.41-4(e)(5)(i)-(iii), 62 Fed.Reg. 81, 83 (Jan. 2, 1997) (Credit for Increasing Research Activities) (adopting this same standard).5 The parties stipulated that all eight of the programs were not commercially available. The district court held that the programs “all fall under the plain meaning of the *448definition [of innovative] included in the legislative history” and therefore meet the requirements of the first prong of the special exception. United Stationers II, 982 F.Supp. at 1287-88.6 However, the district judge found against USI on the second prong, concluding that the projects did not involve substantial uncertainty because of technical risk. See id. at 1288. USI now challenges the court’s legal and factual disposition of the risk prong, while the government challenges the district court’s factual findings as to innovativeness. We consider each of these contentions, taking the government’s argument first.
According to the Conference Report, computer software is innovative if it “results in a reduction in cost, or improvement in speed, that is substantial and economically significant.” H.R. Conf. Rep. No. 99-841, at 11-73 (1986). Applying this principle, the government argues that the district court did not properly focus on the economic significance of the efficiencies generated by USI’s programs. It points out that some 65% of the original investment in the projects did not produce the anticipated economic returns. But we do not think that this quantitative approach is necessarily dispositive; the facts of every case will be different. Cf. Nonvest, ¶ 52,758, at 4656, 1998 WL 341634 (rejecting an attempt to quantify the required improvement). Instead, we believe that Congress’s use of the adjectives “substantial” and “significant” indicates that internal use software to escape the exclusion must be more clearly innovative than software merely meeting the § 41 requirements in general. See id.; see also S.Rep. No. 99-313, at 695 (1986) (internal use software must meet a “high threshold of innovation”); H. Rep. No. 99^126, at 178 (1986) (same). After a thorough review of the project summaries, we believe that all of the programs resulted in improvements in speed or reductions in cost. The district court concluded that these efficiencies met the more onerous requirements of the relevant test. We cannot say on this record that the district court clearly erred in finding that the programs were innovative.
But the Conference Report also defines the risk necessary to meet the requirements of this exception to the internal use exclusion. The Report states that a computer software research project must involve “significant economic risk (as where the taxpayer commits substantial resources to the development and also there is substantial uncertainty, because of technical risk, that such resources would be recovered within a reasonable period).” H.R. Conf. Rep. No. 99-841, at 11-73 (1986). USI argues that the district court improperly focused on the uncertainty of developing the programs rather than the uncertainty that the programs would recover their costs. Whatever the focus, however, USI’s projects do not meet the exception’s definition because they did not involve the requisite technical risk. In the context of the general requirements for § 41, we have already found that there was little if any technical risk that USI would not succeed in developing the programs. Here, in the context of an exception to an express exclusion, Congress again employed the weighty adjective “substantial” to indicate a higher threshold of risk than for software merely meeting the § 41 requirements in general. We do not purport to define the exact contours of this test — again, the inquiry will be fact-driven, and proposed regulations might further guide this determination — but we can be confident that because USI’s projects did not meet the less demanding general requirements of § 41’s qualified research definition, they necessarily do not pass the higher threshold of risk required by the exception to the internal use exclusion. We therefore also agree with the district court on this requirement.
The judgment therefore is Affirmed.
. The House Report on the ERTA research credit indicated that the credit would only apply to expenses for “developing new or significantly improved programs or routines that cause computers to perform desired tasks (as distinguished from other software costs where the operational feasibility of the program or routine is not seriously in doubt).” H. Rep. No. 97-201, at 114 (1981). The Conference Report did not, however, include these software guidelines. See H. Conf. Rep. No. 97-215, at 223 (1981).
. The Senate Finance Committee report on these amendments well illustrates Congress’s concerns:
After reviewing available information and testimony on the actual use of the credit to date, the committee believes that the statutory credit provision should set forth an express definition of qualified research expenses for purposes of the credit. The committee believes that the definition has been applied too broadly in practice, and some taxpayers have claimed credit for virtually any expenses relating to product development. According to early data on the credit, the Treasury has reported, many of these taxpayers do not engage in high technology-
S. Rep. No. 99-313, at 694-95 (1986) (emphasis added); accord H. Rep. No. 99-426, at 178 (1985).
. USI also asserts that the term "discovery” in § 41(d)(l)(B)(i) must have the same meaning as in § 174 and § 41(d)(1)(B)(ii). Legislative history suggests, however, that "Congress intended to treat the discovery test under section 41 more narrowly than the discovery test created under section 174.” Norwest, ¶ 52,758, at 4653, 1998 WL 341634; see also id. (arguing that the timing of the enactment of the sections as well as their express language are reliable indicia that discovery means different things in §§ 41 and 174).
. In Norwest the parties agreed that all the programs constituted internal use software. See Norwest, ¶-52,758, at 4650, 1998 WL 341634.
. This is the one point on which the district judge disagreed with the magistrate. Compare United Stationers I, 1997 WL 159526, at *5.