In a complaint filed in the United States District Court for the District of Massachusetts, SMS Systems Maintenance Services, Inc. (SMS) accused Digital Equipment Corporation (DEC) of violating Section 2 of the Sherman Act, 15 U.S.C. § 2, by integrating three-year warranties with sales of computer systems. SMS, an equipment servicer, asserted that deploying warranties in this manner unfairly constrained consumers' ability to choose their preferred service providers and thereby paved the way for a monopoly in the services aftermarket for DEC computers. The district judge granted summary judgment in DEC's favor. See SMS v. DEC, 11 F. Supp. 2d 166 (D. Mass. 1998). Although our reasoning does not mirror that of the lower court, we nonetheless affirm.

* We sketch the facts, viewing them as favorably to SMS as reason and the record will permit. See Conward v. Cambridge Sch. Comm., 171 F.3d 12, 17 (1st Cir. 1999) (articulating summary judgment standard). We furnish additional details as they become relevant to the ensuing analysis.

DEC manufactures an array of hardware, ranging from personal computers (PCs) to mainframes. In the market for mid-range computers, DEC battles other heavyweights (e.g., IBM, Sun Microsystems, and Hewlett-Packard) for the attention and affection of consumers. In April 1994, DEC introduced its "Alpha" line, consisting largely of mid-range servers. These models were more powerful and more versatile than their predecessors and embodied certain distinctive technological advances. DEC included a three-year warranty as part of the mid-range Alpha package. Although multi-year warranties for PCs had become standard fare in the early 1990s, a three-year warranty in the mid-range server market was uncommon in 1994. One-year warranties were the norm -- indeed, DEC itself provided a one-year warranty in respect to its pre-Alpha products and continued to offer one-year warranties in connection with sales of its established "VAX" line of mid-range computers even after it introduced the Alpha models.1

DEC's conception of a warranty as an instrument of competition is scarcely original. See, e.g., 3A Phillip E. Areeda & Herbert Hovenkamp, Antitrust Law, ¶ 761, at 55 (1996) (referring to warranty protection as a tool of "non-price competition" that has particular importance for firms competing in robust product-differentiated markets for durable equipment). A warranty functions essentially as an insurance policy. See generally Thomas J. Holdych and Bruce D. Mann, The Basis of the Bargain Requirement: A Market and Economic Based Analysis of Express Warranties, 45 DePaul L. Rev. 781, 794-99 (1996). The customer pays the purchase price and receives not only the purchased product itself but also the manufacturer's promise to repair defects and supply replacement parts without extra charge (usually under certain conditions and during a certain interval). Because a warranty is a mechanism through which a consumer can protect himself against the uncertainties inherent in owning a product that likely will require parts and service over time, the product's allure increases as the warranty terms become more generous. This attraction is magnified in some cases because a strong warranty signals a manufacturer's faith in the quality of its product. In theory, then, warranties boost sales (and, ultimately, profits).

Given this general experience, SMS's claim that DEC's warranty is anticompetitive appears odd at first blush. There is, however, a certain offbeat logic to SMS's position. The aftermarket for servicing computers is both dynamic and lucrative. While manufacturers usually seek to service the hardware that they produce, other firms compete with them for this business. Many of these independent service organizations (ISOs) operate nationally or regionally and some specialize in servicing particular brands of equipment. A manufacturer's customers may in fact prefer to use these ISOs for a variety of reasons, including loyalty, convenience, response time, pricing, and perceived quality of service.

Against this backdrop, SMS -- an ISO that operates nationally and specializes in servicing DEC equipment -- puts a sinister cast on DEC's introduction of "mandatory" warranties (that is, warranties that accompany the product at no extra charge). SMS contends that current users of DEC equipment, known in the industry argot as its "installed base," are effectively "locked-in" to buying DEC computers in the future because of the magnitude of their sunk costs (e.g., outlays related to training employees to work with DEC systems and to the acquisition of expensive software that is compatible with those systems). This lock-in phenomenon, SMS warns, creates an environment in which a warranty can function as a vehicle for aftermarket monopolization by creating a disincentive for computer purchasers to consult service firms other than the manufacturer itself. In SMS's view, a purchaser who has a warranty will not readily take his service business to an ISO because no consumer wants to pay twice for the same service -- and the longer the warranty, the less the opportunity for ISOs to compete. SMS predicts that, if such practices are left unchecked, lost business opportunities will ruin ISOs, eliminate competition, and eventually enable manufacturers to raise aftermarket prices.

Before moving to an analysis of SMS's claim, we must put a Trojan horse out to pasture. DEC suggests that this case involves only a "single product," because the challenged three-year warranty is nothing but an "attribute" of the hardware. If this were true, the primary equipment market would constitute the relevant market and DEC's share of it (according to its estimates, at least) would be far below what SMS would need to show a credible threat of monopolization, and, thus, the case would vanish quietly into the night.

The single product defense is customarily invoked to rebut allegations of an illegal tie brought under Section 1 of the Sherman Act (15 U.S.C. § 1). See, e.g., Jefferson Parish Hosp. Dist. No. 2 v. Hyde, 466 U.S. 2, 18-25 (1984). In the context of SMS's Section 2 monopoly claim, DEC appears to wield the defense in a different manner, i.e., as a means of eliminating the possibility that its warranty has any relevant connection with the services aftermarket. Although we are tempted to dismiss DEC's assertion simply as a matter of common sense, the best indication of its infirmity resides in the record.

The success of a single product defense ordinarily hinges on observations of actual market practices. If evidence shows that there is significant demand for separate components of what is alleged to be a single product and the product is in fact sold in those forms, there is no single product. See Eastman Kodak Co. v. Image Tech. Servs., Inc., 504 U.S. 451, 462-63 (1992); 10 Areeda & Hovenkamp, supra, ¶ 1744, at 203. Conversely, if all or substantially all competing firms always or almost always (discounting occasional accommodations for customers with special needs) sell two seemingly separate items together, then the two attributes may be treated as components of a single product. See generally 10 Areeda & Hovenkamp, supra, ¶ 1744, at 197-98.

In mounting its single product defense, DEC relies heavily on the affidavit of Professor Jerry Hausman, a distinguished economist, who explains uncontroversially that warranties are tools used by manufacturers to promote the attractiveness of their products and thus to gain a competitive edge. Using this statement as a springboard, DEC vaults to the conclusion that hardware and warranty are a single product. But DEC leaps without looking. The record is pellucid that, while some other computer manufacturers bundle warranties with their machines, the practice, especially with regard to mid-range computers, is anything but universal.

Vendors of mid-range computers offer a variety of ancillary packages with their products, including service contracts, warranties of differing lengths, or nothing at all. Indeed, a DEC official, Jane Heaney, affirmed in the course of pretrial discovery that differing service packages are items that customers can (and do) use to differentiate between competing computer brands. In competitive markets, suppliers' offerings generally reflect consumer demand. See, e.g., Digital Equipment Corp. v. Uniq Digital Tech., Inc., 73 F.3d 756, 762 (7th Cir. 1996). We therefore conclude -- at least for the purpose of summary judgment -- that the mid-range computer market treats warranty and equipment as separate products.

Having laid to rest DEC's attempt to shortstop the requisite analysis, we turn to the parameters of SMS's burden. To prove a violation of Section 2 of the Sherman Act, the plaintiff must show both that the defendant has monopoly power in a relevant market and that it has maintained or increased that power through anticompetitive conduct. See United States v. Grinnell Corp., 384 U.S. 563, 570-71 (1966); Town of Concord v. Boston Edison Co., 915 F.2d 17, 21 (1st Cir. 1990); Olympia Equip. Leasing Co. v. Western Union Tel. Co., 797 F.2d 370, 373 (7th Cir. 1986). SMS's monopolization hypothesis begins with the proposition that, before all else, we must define the relevant market. Relying on traditional product market definition cases, see Brown Shoe Co. v. United States, 370 U.S. 294, 325 (1962); United States v. E.I. du Pont De Nemours & Co., 351 U.S. 377, 404 (1956), SMS contends that, because DEC's aftermarket products are "unique" and servicing DEC computers requires specialized expertise, the relevant market must be the aftermarket for servicing DEC mid-range servers.

The argument, if successful, would place SMS at a distinct advantage. Market share often serves as a proxy for market power. See, e.g., Town of Concord, 915 F.2d at 30; Grappone, Inc. v. Subaru of New Engl., Inc., 858 F.2d 792, 797 (1st Cir. 1988). If we were to agree with SMS that the relevant market is DEC's aftermarket for services, then, given that DEC predominates in that market (as evidenced by its large market share), SMS would have traveled far on the road to establishing DEC's monopoly power. But we believe that SMS's single-minded focus on traditional concepts of product market definition oversimplifies the route it must take. Cases involving aftermarkets are sui generis.

The purpose of defining a relevant market is to assist in determining whether a firm has market power. See Israel Travel Advisory Serv., Inc. v. Israel Identity Tours, Inc., 61 F.3d 1250, 1252-53 (7th Cir. 1995); see also U.S. Healthcare Inc. v. Healthsource, Inc., 986 F.2d 589, 598 (1st Cir. 1993) (noting the importance of remembering what antitrust question the court is trying to answer when it engages in market definition). In its most basic iteration, the monopolist's market power consists of having sufficient economic muscle to permit it to raise prices well in excess of competitive levels without inducing customers to turn elsewhere. See Town of Concord, 915 F.2d at 31. It follows inexorably that a traditional product or geographic market definition exercise would be the starting point in an aftermarket case if, but only if, a large market share in such a derivative market invariably translated into (or at least reliably indicated) monopoly power in that market.

This condition does not obtain as a matter of course. Most firms that service their manufactured products can be expected to have a very high percentage of the services aftermarket for those products. See generally Daniel Wall, Aftermarket Monopoly Five Years After Kodak, 11 Antitrust 32, 32-33 (1997). Virtually by definition, manufacturers possess relative expertise in repairing their own products (an expertise that increases in importance as the good increases in technological complexity) and consumers typically are reluctant to incur the transaction costs associated with locating and hiring alternative service providers.

But the naked fact that a manufacturer has a high percentage of the market for servicing its own products does not mean that it can raise the price of services or parts with impunity in that market. See 10 Areeda & Hovenkamp, supra, ¶ 1740, at 170-71. Reputation is important to a firm that constantly competes for new customers, and a manufacturer's behavior in the aftermarket probably will be scrutinized by customers shopping for the firm's products in the primary market. If the firm has a bad reputation, that will prompt potential customers to go elsewhere. Moreover, such a firm eventually will suffer defections from its installed base as well, for firms concerned with the long term cannot afford to bite the hands that feed them. See id. ¶ 1740, at 153; see also Joseph Kattan, Market Power in the Presence of an Installed Base, 62 Antitrust L. J. 1, 15-16 (1993); Wall, supra, at 33. Under such circumstances, it ordinarily captures the reality of the marketplace to envision a firm's behavior in the aftermarket as having a direct effect on the "cross-elasticity of demand," du Pont, 351 U.S. at 400, with respect to its products in the primary market.

Of course, markets do not always function in the manner we have just described. Kodak is a case in point. There, a group of ISOs complained that a copy-machine manufacturer which competed with ISOs in servicing its brand of copiers had engaged in exclusionary behavior in its parts and services aftermarket (Kodak had stopped supplying the ISOs with the parts required to service Kodak copy machines, while at the same time entering into pacts with independent parts manufacturers that prohibited them from supplying the ISOs with copier parts). See Kodak, 504 U.S. at 455-56. Then, the manufacturer told its customers that they could not purchase parts unless they agreed to use the manufacturer's aftermarket services. See id. at 456-58. Coupling evidence of this scheme with evidence that Kodak was charging supracompetitive prices in the aftermarket, the ISOs accused Kodak of improperly expanding its monopoly in the services aftermarket. See id., 112 S.Ct. 2072

Kodak countered that, as a matter of law, a manufacturer cannot wield monopoly power in a derivative market because competitive primary markets, by definition, always check aftermarket behavior. The Court granted certiorari to consider the bona fides of this argument. See id. at 454-55. Ultimately, it rejected Kodak's contention as a matter of law and affirmed the remand for further proceedings because data in the record suggested that Kodak might in fact have been exacting monopoly prices in its derivative market and there was no sign that competition in the primary market deterred it from doing so. See id. at 477-78.

Kodak, then, stands for the proposition that the foremarket does not always exert sufficient competitive pressure to insulate the aftermarket from monopolistic practices. It does not hold, as SMS entreats, that the foremarket never exerts sufficient competitive pressure to keep the aftermarket pristine. And the fact that the primary market at times may fail to discipline a derivative market does not mean that the latter necessarily constitutes the relevant market for antitrust analysis. Rather, a litigant who envisions the aftermarket as the relevant market must advance hard evidence dissociating the competitive situation in the aftermarket from activities occurring in the primary market. Cf. id. at 477 (noting that the plaintiff adduced evidence that the defendants were able to -- and did -- charge supracompetitive prices in the aftermarket). Put another way, a court may conclude that the aftermarket is the relevant market for antitrust analysis only if the evidence supports an inference of monopoly power in the aftermarket that competition in the primary market appears unable to check. See id. at 481-82; see also 10 Areeda & Hovenkamp, supra, ¶ 1740, at 170-71.

We summarize succinctly. In any market with some degree of product differentiation, goods of a single brand will enjoy a certain degree of uniqueness. See 3A Areeda & Hovenkamp, supra, ¶ 761, at 55-56. While that uniqueness may account for a manufacturer having a large market share in the services aftermarket for its own brand, that fact, without more, does not suffice to establish that the manufacturer enjoys monopoly power in that market. Unless the evidence shows that the manufacturer can exert raw power in the aftermarket without regard for commercial consequences in the foremarket, the aftermarket is not the relevant market. This means that, for antitrust cases involving derivative markets, courts cannot reflexively resort to traditional product market definition methods, but must look to other modes of scrutinizing the existence and measure of market power. As Kodak teaches, in aftermarket situations, market power vel non must be assessed by weighing the complete package of primary equipment, parts, and services.2

We turn next to the question of whether SMS has managed to create a genuine issue of material fact as to DEC's alleged monopoly power. We think it has not.

* Laboring to dispel the customary expectation that competition in the primary market may serve as a substantial deterrent to anticompetitive behavior in an aftermarket, SMS first asserts that this is a case in which information costs have distorted market behavior. The argument has its genesis in Kodak, where the Court suggested that, under some circumstances, the foremarket's inability to obtain needed information about the aftermarket could foil the former's ability to act as a check on the latter. See Kodak, 504 U.S. at 473-76. Because this information could be difficult for ordinary copy-machine buyers to obtain, and Kodak apparently treated informed purchasers more favorably than uninformed purchasers (through price and term discrimination), the Court suggested that these facts might explain Kodak's ability to charge supracompetitive prices in the aftermarket without suffering adverse consequences in the foremarket. See id. at 473-74.

DEC counters that the purely prospective nature of its warranty policy removes this case from Kodak's sphere of influence. DEC explains that the only customers affected by the warranty policy are those who purchase new DEC mid-range systems; contracts between ISOs and current owners of DEC platforms are unimpaired. Furthermore, purchasers are keenly aware of the warranty terms. If a customer does not like the warranty, DEC continues, he will shop elsewhere, either for a different brand of computer or for a DEC computer without a warranty. See, e.g., Digital, 73 F.3d at 762 (explaining that customers will shop for other brands if a manufacturer does not satisfy their needs).

We agree that both the transparency of DEC's warranty policy to buyers in the primary market and the policy's prospective outlook militate in DEC's favor. Let us begin with transparency. The Kodak Court's discussion of information costs stemmed exclusively from the concern that the type of information necessary for allowing the primary market to check anticompetitive behavior in the aftermarket was unavailable to the average copy machine purchaser. See 504 U.S. at 473-76. Given the specific means that Kodak had adopted to stifle competition, the information relevant to the primary market would be lifecycle costing, which, the Justices concluded, was difficult to access. See id. at 473-74.

The information deficits that concerned the Kodak Court are largely absent here. SMS cites the three-year warranty as the instrument of market dominance. But, unlike the precise cost of owning a machine for its useful life (the Kodak scenario), the warranty's existence is obvious to any purchaser in the primary market and, therefore, SMS's ruminations about each purchaser's need to engage in lifecycle costing for installed hardware are beside the relevant point.3 Hence, the only datum that bears on alleged anticompetitive activity in the aftermarket is readily available to the foremarket. Given that the primary market's ability to check a firm's behavior in its own aftermarket hinges in substantial part on whether information about the latter is sufficiently reflected in the former, the transparency of DEC's allegedly monopolistic policy represents a salient departure from the Kodak scenario. Accord PSI Repair Serv., Inc. v. Honeywell, Inc., 104 F.3d 811, 819-20, 822 (6th Cir.) (concluding that the fact that the defendant's policies were "generally known" was a central datum in barring a Kodak-type claim), cert. denied, 520 U.S. 1265 (1997).

The prospective nature of the warranty further distinguishes this case from Kodak. Kodak had entered into agreements that precluded the distribution of Kodak parts to ISOs. See Kodak, 504 U.S. at 458. These agreements were tantamount to a retroactive change in the rules because many customers had purchased Kodak machines against a background understanding that they would be able to procure parts from ISOs. Accordingly, this tergiversation detrimentally affected the expectations of those who had already sunk considerable sums of money into the acquisition of Kodak equipment. Such "bait and switch" tactics can be problematic from an antitrust standpoint when they enable exploitation. See Digital, 73 F.3d at 763 (concluding that "[t]he material dispute that called for a trial [in Kodak] was whether the change in policy enabled Kodak to extract supra-competitive prices from customers who had already purchased its machines"); Lee v. Life Ins. Co. of N. Am., 23 F.3d 14, 20 (1st Cir. 1994) (noting that Kodak's policy change jeopardized its customers' ability to protect themselves). Like the easy availability of information, the purely prospective nature of the warranty helps to take this case out of Kodak's precedential orbit. See PSI Repair, 104 F.3d at 820, 822.

Although this constellation of factors undermines SMS's rote reliance on Kodak, the scenario would be fully dispositive of SMS's Section 2 claim only if all purchasers in the primary market were free agents. But SMS says that is not the case. It endeavors to draw an analogy between the Kodak plaintiffs' need to buy copier parts and the perceived need of those who already own DEC computers to purchase any new hardware from DEC (rather than its competitors). The linchpin of this analogy is SMS's insistence that repeat purchasers of computers are more or less compelled to patronize the manufacturer with whom they originally dealt because of the concomitant investment in, among other things, specialized training and software. As to these "locked-in" firms, SMS asseverates, integrating a three-year warranty into new computer systems is akin to the "policy change" that the Kodak Court decried because, at the time of the initial purchases, these consumers had no idea that future generations of DEC computers would bear such warranty terms.

This argument has some conceptual footing. In theory, aftermarkets may be insulated from the competitive atmosphere of primary markets if, and to the extent that, current owners of a manufacturer's products find it prohibitively dear to "switch" to another product because of the large investment they have made. See Kodak, 504 U.S. at 476. Because of switching costs, these owners might be harmed even if information about anticompetitive behavior is reflected in the primary market (for example, the manufacturer may offer better purchase terms to first-time buyers). Withal, we do not believe that the evidence suffices to create a genuine issue as to whether this theoretical possibility has been realized here.

Even though SMS's lock-in argument relies heavily on Kodak, there is an important factual distinction between the two cases in regard to the nature of the alleged switching costs. When the Kodak Court spoke of switching costs, it referred specifically to the cost of purchasing new copying equipment. See Kodak, 504 U.S. at 476-77. What prompted the Court to suggest that switching would be necessary was Kodak's policy of offering parts only to those customers who also purchased service from it. This meant that if a customer did not turn to Kodak for service, it could not get parts, and its copy machine eventually would be rendered useless.

Here, by contrast, DEC neither withholds parts nor otherwise precludes any hardware purchaser from using another service provider. If a customer prefers to retain an ISO, it does not need to switch to another computer system. Indeed, the record shows not only that customers do continue to hire ISOs to service DEC machines that are under warranty, but that SMS itself bids for (and has procured) such contracts. Thus, the only switching cost at issue is the cost of writing off the portion of the equipment's purchase price that represents the warranty. If the behavior of consumers and ISOs is any indication, the switching cost is not particularly significant. (Although SMS presents some conclusory testimony that this cost is the rough equivalent of the outlays for regular service contracts, it has absolutely no solid evidence to back up the claim -- and, indeed, this testimony is at variance with its announced position that the warranty is in actuality a "loss leader" for DEC. See infra note 5.) At any rate, SMS has not proffered significantly probative evidence sufficient to create a fact question as to whether this alleged switching cost is material to a large enough segment of DEC's installed base to harm competition.

To sum up, while the warranty at issue here may act, at least in some cases, as a subtle disincentive that inhibits customers from consulting ISOs (because the customer would, by retaining an ISO, in effect write off whatever extra cost the manufacturer had built into the computer's purchase price to permit it to offer the warranty), there is nothing inherently anticompetitive about the warranty. Unlike Kodak, this is not a case in which a manufacturer effectively forces consumers to use one, and only one, service outfit. Moreover, nothing in the record before us (except for conclusory, self-serving testimony by SMS officers, which we need not credit on summary judgment) suggests that the existence of the warranty dissuades a sufficient number of customers who otherwise would have used ISOs' from doing so. Testimony of unbearable switching costs by a mere handful of disaffected customers does not satisfy SMS's burden of showing that consumers are generally worse off as a result of DEC's warranty policy. See, e.g., 10 Areeda & Hovenkamp, supra, ¶ 1740, at 152-53.

That is essentially what the Eighth Circuit concluded in a post-Kodak case when it indicated that, as long as a warranty does not limit a customer's choice of service provider, there is usually no antitrust problem. See Marts v. Xerox, Inc., 77 F.3d 1109, 1112 (8th Cir. 1996); see also Digital, 73 F.3d at 761-62 (reaching a similar conclusion with respect to installed operating systems). We agree with this conclusion. Many manufacturers offer the market bundled products that have separable components. Every day, consumers purchase these products because "buyers often find package sales attractive; a seller's decision to offer such packages can merely be an attempt to compete effectively -- conduct that is entirely consistent with the Sherman Act." Jefferson Parish, 466 U.S. at 12.

Of course, some consumers may not like all the components, but the overall package fits their needs, they are not constrained in looking elsewhere for alternative components, and they will not have to scrap their prior investments just because the new package is not a perfect fit. Given this commonplace, it cannot be that every time some consumer or group of consumers does not like one component of a complex product, the manufacturer will be exposed to an antitrust claim on the ground that it has circumscribed consumer choice. The antitrust laws do not require manufacturers to customize their offerings to the precise specifications of each and every customer.

At the expense of carting coal to Newcastle, we note a second serious shortcoming that permeates SMS's lock-in argument. Throughout its analysis in this case, SMS assumes that all current DEC customers fall into the locked-in category. This taxonomy does not hold together: being part of an installed base and being locked-in are not synonymous. A lock-in phenomenon must be shown, not assumed. See Kodak, 504 U.S. at 476; see also 10 Areeda & Hovenkamp, supra, ¶ 1740, at 152-53. Moreover, in undertaking this inquiry, the nature of the product must be taken fully into account, for that will inform an inquiring court about the degree of effective control the manufacturer may be able to exert over a consumer's purchasing decisions. That is singularly important here, for the evidence required to show that an installed base is locked-in to future purchases of computer systems is quite different from that needed to show that a copy machine owner is locked-in to buying parts for that machine.

It is an article of faith, for antitrust purposes, that unless a substantial number of preexisting customers are locked-in, defections from the manufacturer's installed base, coupled with losses in the foremarket, in all probability will sabotage any effort to exploit the aftermarket. Our review of the instant record affords no reason to believe that an appreciable number of DEC users are, in fact, significantly locked-in to making repeat purchases of DEC mid-range computers. SMS's most direct evidence of lock-in comes from the testimony of individuals employed by two of DEC's customers. These individuals -- none of whom bore the ultimate responsibility for making decisions regarding new computer purchases and none of whom exhibited any specific or detailed knowledge of how management might go about calculating switching costs -- testified generally that once a company has committed to a certain brand of hardware and has acquired ancillary software (including an operating system), it would be costly to migrate to another platform. Significantly, however, the testimony fails properly to account for the nature of a company's decision to purchase new computer hardware and, therefore, we cannot put much stock in it.

The evidence to which we have just referred assumes that a company's decision to replace its computer system will be dictated solely by the hardware that it currently uses. Under certain circumstances, the assumption might hold. The best example probably derives from the analogy that Kodak offers. Although the owner of a copy machine might hope that he will not require parts during the machine's useful life, the hope defies reality. When the moment of need occurs, the corporate decisionmaker must be able to purchase a Kodak part -- or else, scrap the machine. The fragility of the equipment forces the owner to buy parts, and this need, linked with owners' general unwillingness to sacrifice the remaining useful life of machinery, enables the manufacturer to control and exploit the situation.4

The record evidence shows that decisionmaking surrounding the purchase of a new mid-range computer, however, is qualitatively different from the decision to buy copier parts. Typically, a much wider variety of factors will enter into the computer user's decision. Although we do not discount the impact of a server's speed and power (which will lead to hardware upgrades), the record indicates that the availability of software applications designed to meet current or anticipated business needs usually is the determinative factor in this calculus.

This reality is largely borne out by a marketing survey and report that SMS considers to be one of the most significant pieces of documentary evidence in this case. The report's observations are particularly noteworthy because many of the survey respondents were chosen from DEC's installed base and the purpose of the study was to gauge their receptivity to DEC's Alpha and VAX computers. The survey concludes that end users do not focus principally on hardware when making decisions about buying new systems. Rather, in the words of the report, "applications drive the choice of platforms," because software applications represent the business solutions to the problems confronting the computer user. There is no record evidence that contradicts, or even comes close to undercutting, this telling point.

A fortiori, if software is the primary determinant in commercial decisionmaking on computer system purchases, the power of affecting a customer's purchasing decisions resides more with those who design and sell software than with hardware manufacturers. That is exactly what the survey indicates. To complicate matters further, software companies tend to develop their applications so that they can be used with the more popular operating systems. In turn, given the efficiencies of networks, many consumers tend to opt for software that is usable on such systems. The evidence in this case shows that operating systems other than DEC's VMS -- such as Windows NT and UNIX -- are becoming increasingly prevalent (which perhaps explains why DEC engineered the new Alpha line to operate with all three) and many other hardware manufacturers make systems that can accommodate these other operating systems. If DEC is to remain the company of choice for its installed base, it cannot rest on its laurels.

The complex nature of the decision to purchase a new computer system means that an analysis of switching costs in this context cannot parrot the linear inquiry that proved possible in Kodak. Here, the cost of shifting to another system must take into account the efficiency gains of buying new software -- gains that often may dwarf hardware price in dollar terms. This is true even if one considers the switching costs that are associated with retraining employees and discarding software designed to run exclusively on a particular platform. The record demonstrates convincingly that, in this industry, both software vendors and hardware manufacturers offer migration support for new customers in the form of significant discounts on training, installation, and software conversion, thus internalizing much of the switching costs.

The impressions that we have gathered from the marketing survey are largely confirmed by the very witnesses on whose testimony SMS relies. Quite aside from the statements these witnesses make about the magnitude of the cost of randomly switching from one currently functioning computer system to another -- testimony which, as is clear from our discussion, addresses an irrelevant scenario -- all of them acknowledge that, all things considered, if a new computer system would bring more benefits, there would be no objection to the switch. The actual behavior of these witnesses' employers illustrates the point: their firms were in fact in the process of switching some of their systems to other platforms, citing the sorts of reasons we have catalogued.

In fine, the record does not support the conclusion that a substantial number of installed base customers are locked-in -- and SMS has failed to make the case. The power of software vendors, the rapid progress of software solutions, the unpredictability of when consumers will seek to purchase a brand new system, the ready willingness of competitors to absorb migration costs, and the uncertain calculus of cost versus efficiency gains that obtains when a firm moves to new applications all distinguish the computer purchase context from the copier context. This plethora of factors strongly implies that, in most instances, DEC has no effective control over whether a customer will remain loyal when opting to purchase a new computer system.

To address these issues, which are central to assessing the existence of a lock-in, SMS ought to have focused on proof of the actual behavior of an installed base in terms of such consumers' tendency to return to the same manufacturer. It did not marshal any such proofs. Consequently, the record lacks the empirical evidence required to give us a proper glimpse into whether members of an installed base actually behave in conformity with SMS's switching cost hypothesis. Although we do not doubt that some customers may experience the lock-in that SMS envisions, the record affords no reason to believe that such customers are numerous or that they are representative of the installed base. And -- in marked contrast to Kodak -- there is no evidence, direct or circumstantial, that DEC has attempted to locate or exploit vulnerable customers within its installed base (or that it has the capacity to identify and isolate such customers).

These last points are particularly important. In a product-differentiated market (such as the mid-range computer hardware market), there always will be a subset of customers whose subjective preferences, given their specific business needs, will align them more closely with one manufacturer. As we remarked in a closely related context, however, this kind of preference does not translate into the kind of economic power that antitrust law aims to mitigate. See Grappone, 858 F.2d at 797 ("Of course, virtually every seller of a branded product has some customers who especially prefer its product. But to permit that fact alone to show market power is to condemn ties that are bound to be harmless, including some that may serve some useful social purpose."). Sophisticated consumers with such preferences will know beforehand that they will lock themselves in by their choice of manufacturer and do so willingly. What would be of concern is if a firm were able to extend its control by improper means over a sufficiently sizable number of customers who did not have such a preference. The facts at hand do not fit this model. The evidence, taken most favorably to SMS, does not demonstrate a lock-in, certainly not one that raises antitrust concerns. See 10 Areeda & Hovenkamp, supra, ¶ 1740, at 152-53.

Because this case arises on an appeal from the entry of summary judgment, it behooves us to stress further SMS's failures at the empirical level.

Here, unlike in Kodak, the record is devoid of any evidence of supracompetitive prices or other oppressive terms of business in the aftermarket.5 Here, unlike in Kodak, the record furnishes no reason to believe that new customers are irrelevant to DEC; to the contrary the record shows unequivocally that DEC intended to sell Alpha computers vigorously to new as well as old customers in an effort to grow its market share.6 Here, unlike in Kodak, there is also no proof indicating that DEC has attempted to discriminate between customers who are knowledgeable and those who are not, or that it discriminates in price (or in offering different warranty terms) between new and repeat purchasers.7 Such discrimination, the Court explained, would be one way that manufacturers would attempt to prevent sophisticated purchasers from disciplining the aftermarket through their decision to forego buying from the manufacturer. See Kodak, 504 U.S. at 475. Moreover, data such as these -- that new customers remain important to DEC and that the warranties offered to new and installed base customers have exactly the same terms -- are further signs of the absence of aftermarket exploitation. See 10 Areeda & Hovenkamp, supra, ¶ 1740, at 159.

All of these factors were crucial to the Kodak Court's conclusions, forming the bases for the Court's attempts to explain, through hypotheses involving information deficits and switching costs, the possible causes of the market's apparently irregular behavior. See id. at 473-78. DEC's alleged behavior, however, is worlds apart from Kodak's. In the absence of comparably suspicious practices, there is no reason to assume that DEC's interest in maintaining its position in a fiercely competitive primary market is not checking any exploitative tendencies in the derivative market.

Represented by able counsel, SMS strives resourcefully to fill these evidentiary gaps. When queried at oral argument, it referred us to the affidavit of Joseph Scordino (SMS's director of marketing). That affidavit does not remedy the empirical shortcomings that we have identified. The closest it comes is the following: "It has been my experience that when [DEC] does not face competition, its prices tend to be higher. For example, [DEC] does not discount its maintenance services on the Alpha 8000s. Thus, the elimination of competition from SMS and other Independent Service Organizations, should result in price increases to customers." We fail to see how we can allow this self-serving speculation to pass for competent evidence. The third sentence is conclusory conjecture of a kind that begs the very question at issue, namely, whether the foremarket sufficiently disciplines the aftermarket. The same is true for the first sentence, because, if competition in the foremarket checks activity in the aftermarket, the absence of competitors in the aftermarket will not lead to higher prices. The second sentence, which deals with Alpha 8000s, is irrelevant; those computers are high-end servers and belong to a different product market (and the record is bereft of evidence that the market structure for high-end servers has any bearing, direct or by reasoned analogy, on the market structure for mid-range servers).

SMS also cites the report of its retained expert, Dr. William Bleuel, to the effect that DEC has a much greater share of the services aftermarket than it should enjoy, given low customer satisfaction. If DEC did not have monopoly power, this argument runs, it would not have been able to keep its large share of the aftermarket in spite of rampant dissatisfaction. Expert opinions, however, are no better than the data and methodology that undergird them -- and, on this score, Dr. Bleuel's conclusions are highly suspect. Although we find Dr. Bleuel's report deficient on several levels, it suffices to discuss only the shortcomings of his data collection.

Dr. Bleuel did not conduct a customer satisfaction survey, but based his opinion on his interpretation of certain internal DEC documents. The relevant part of his report begins with the intuitively obvious proposition that when customer satisfaction declines, chances are that a company will lose some business. Then, citing to sources which it neither attaches nor discusses, the report concludes that DEC's customer satisfaction ratings have been declining. Dr. Bleuel's conclusion may or may not be correct, but an expert must vouchsafe the reliability of the data on which he relies and explain how the cumulation of that data was consistent with standards of the expert's profession. See Wessmann v. Gittens, 160 F.3d 790, 804-05 (1st Cir. 1998); see also Kumlo Tire Co. v. Carmichael, 119 S. Ct. 1167, 1176 (1999) (directing courts "to make certain that an expert, whether basing testimony upon professional studies or personal experience, employs in the courtroom the same level of intellectual rigor that characterizes the practice of an expert in the relevant field").

Not only did Dr. Bleuel fail to discuss in his report the nature of the data and its meaning, but he failed to explain whether the information-gathering technique used in the DEC documents was valid, whether the data was sufficiently representative to permit him to draw any relevant conclusions, and whether the sampling methodology used to compile these documents corresponded to methods that might be considered legitimate in his discipline. Expert testimony that offers only a bare conclusion is insufficient to prove the expert's point. See Mid-State Fertilizer Co. v. Exchange Nat'l Bank, 877 F.2d 1333, 1339 (7th Cir. 1989) ("An expert who supplies nothing but a bottom line supplies nothing of value to the judicial process.").

We add a coda. When a party brings a Section 2 claim, it is not enough simply to show that there is monopoly power. Monopoly power may be obtained through legitimate means. An antitrust problem arises only when an improper use of that power, to the detriment of the forces of competition, occurs. Thus, to make out a Section 2 claim, the plaintiff must show that the alleged monopolist has engaged in improper exclusionary conduct. See Kodak, 504 U.S. at 482-83; Aspen Skiing Co. v. Aspen Highlands Skiing Corp., 472 U.S. 585, 600-05 (1985); Barry Wright Corp. v. ITT Grinnell Corp., 724 F.2d 227, 230 (1st Cir. 1983).

The improper conduct identified by SMS is the inclusion of a three-year warranty on new equipment in the purchase package.8 But a warranty has obvious virtue as a tool of competition, and the commercial justification for its use is compelling. Indeed, people ordinarily associate warranties with consumer welfare and highly competitive markets.

To avoid the presumption that DEC's warranty is no more than a legitimate sales tool, SMS points to two internal documents which it claims show that DEC intended to use the warranty as a device to capture the totality of the services aftermarket. These documents appear to be a part of an ongoing series of correspondence between DEC functionaries relative to the launching of the Alpha product line. One document makes mention of the warranties having "100% penetration with all product sales" and the other makes a similar point. Read in context, neither document so much as hints at an intent to suffocate the aftermarket. The mere fact that the terms "warranty," "penetration," and "100%" appear in a single writing does not signal an improper attempt to distort market forces.

In the last analysis, SMS's monopoly power hypothesis defies common sense. If the three-year warranty is the only means toward monopolization, it will, under the assumptions that SMS adopts, prove to be a remarkably ineffective tool. SMS has taken the position that the warranty actually represents a loss to DEC because, all things being equal, DEC can make more money providing services than it can offering warranties. See supra note 5. If this is true, SMS's argument takes on the flavor of the predatory pricing cases and, therefore, must afford some basis for a belief that DEC can recoup these losses at some point in the future in order to make the enterprise worthwhile.

But DEC has placed no restriction on the aftermarket for service of its products. Parts are available to all market participants, and there is no indication in the record either that there are restrictions on their availability or that such a policy is even being contemplated. Meanwhile, ISOs are free to service all DEC computers, including those that are under warranty, both during the warranty period and after its expiration. On these facts, it seems highly unlikely that DEC will recoup the sums that SMS says that it foregoes when it offers the warranties in the first place. All of this indicates that the more likely explanation for the warranty relates to its efficiency as a means of growing market share in the primary market. In any event, sacrifice now/recoup later claims do not necessarily bespeak monopolistic behavior, see, e.g., Barry Wright, 724 F.2d at 231-36, and we believe that this dimension of SMS's logic is yet another element that distances its case from Kodak while simultaneously bringing it within the orbit of Matsuthita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 587 (1986) (noting that evidence must be even more persuasive to stave off summary judgment when the theory set forth by the party opposing the motion is implausible).

That SMS may have lost business as a result of DEC's policy is not, in and of itself, a concern of the antitrust laws. Antitrust law is designed to protect competition, not competitors. See Spectrum Sports, Inc. v. McQuillan, 506 U.S. 447, 458-59 (1993). If there is no objective indication of harm to competition, we cannot stifle a firm's ability to compete in the primary market just because some of its aftermarket competitors complain that they have lost business as a result. After all, such harm frequently occurs due to the specialized and dependent nature of the investment made by aftermarket firms rather than as the anticompetitive exercise of market power by manufacturers. See 3A Areeda & Hovenkamp, supra, ¶ 762, at 65-66 (explaining the phenomenon in the context of self-distribution decisions made by manufacturers).

We need go no further. The Kodak Court resorted to economic theory in order to highlight the possibility that, under the notably sinister circumstances of the case (evidence of supracompetitive prices in the aftermarket, price discrimination favoring certain consumers, and no resulting change in sales in the primary market), information deficits and switching costs better explained the available market data than did the defendant's glib assertion that the aftermarket is never independent of the foremarket. Theory is powerful when it explains reality. But when evidence for the trumpeted reality is lacking, the theory is of no practical value. Here, the record contains no significantly probative evidence that DEC is engaged in sinister practices or otherwise suffocating competition, and, hence, theoretical possibilities alone are inadequate to block the swing of the summary judgment ax.

Affirmed.