OPINION

RICHARD A. POSNER, Circuit Judge, Sitting by Designation.

This is a suit for patent infringement, brought by a firm named Pinpoint, which owns U.S. Patent No. 5,758,257 (filed in 1994, issued in 1998), entitled a "System Method for Scheduling Broadcast of and Access to Video Programs and Other Data Using Customer Profiles," against Amazon.com. Pinpoint also asserts a second patent, U.S. Patent No. 6,088,722, but it is identical in all but one respect (which I discuss later in this opinion) that bears on this suit.

I dismissed an identical earlier suit on a jurisdictional ground: Pinpoint had not owned the patent when it brought suit. Pinpoint, Inc. v. Amazon.com, Inc., 347 F. Supp. 2d 579 (N.D. Ill. 2004). But it had acquired the patent later and so was able to refile the suit, and it has done so. Amazon asks me to revisit the issue of claim construction. The original case had been reassigned to me after the district judge who had had the case previously had construed the claims on which Pinpoint was (and is) suing; she construed the claims as not being limited to a method of scheduling in the temporal sense of the word or to a method that uses mathematical formulas to create customer and content profiles. Amazon challenges only these two aspects of Judge Conlon's claims construction, having abandoned its other objections to her claims construction at the hearing on its motion to reopen it.

Pinpoint is content with her construction but argues that she erred in thinking that the preamble to each of the claims in the patent is a limitation on the claims and thus on the scope of the patent. The rule is that the preamble is part of a claim if without it the claim would not describe a complete and therefore useful invention. Boehringer Ingelheim Vetmedica Inc. v. Schering-Plough Corp., 320 F.3d 1339, 1345 (Fed. Cir. 2003); Catalina Marketing Int'l, Inc. v. Coolsavings.com, Inc., 289 F.3d 801, 808-10 (Fed. Cir. 2002); Griffin v. Bertina, 285 F.3d 1029, 1033 (Fed. Cir. 2002). Thus Pinpoint's objection is not to the claims construction as such, that is, to the meaning that Judge Conlon gave to the claims in suit, but to a part of her reasoning process.

The objection is unfounded. The preamble to claim 17, for example, is "A method of scheduling customer access to data from a plurality of data sources, comprising the steps of:". Without the reference to scheduling access (explained more fully later in this opinion), the method described in the claim creates and updates customer profiles (mathematically expressed customer program preferences) and content profiles (mathematized program characteristics) and relates them by means of an "agreement matrix" that indicates which programs a customer is likely to prefer in particulate date and time slots. But without the preamble the method described in the claim does not convey the data generated by the matrix to the consumer and so fails to indicate what utility of the invention has, and utility is of course one of the requirements of a patentable invention.

Also, the references in the claim to " said data" are references back to the preamble. So without the preamble, "said data" would have no antecedent — it would be unidentified. This is a further indication that the preamble is part of the claim. NTP Inc. v. Research in Motion, Ltd., 392 F.3d 1336, 1358-59 (Fed. Cir. 2004); Eaton Corp. v. Rockwell Int'l Corp., 323 F.3d 1332, 1339-40 (Fed. Cir. 2003); Rapoport v. Dement, 254 F.3d 1053, 1059 (Fed. Cir. 2001).

But before I can reconsider Judge Conlon's claims construction I must decide whether the doctrine of the law of the case permits me to do so. This depends in the first instance on whether the doctrine even applies to a refiled case. The claim construction that Amazon wishes to reopen was made in the original, not the refiled, case. Orthodox statements describe the doctrine as a limitation on the reexamination of rulings in subsequent stages of the same case. E.g., White v. Godinez, 301 F.3d 796, 803 (7th Cir. 2002); Schiavo ex rel. Schindler v. Schiavo, 403 F.3d 1289, 1291 (11th Cir. 2005) (per curiam); Prairie Band Potawatomi Nation v. Wagnon, 402 F.3d 1015, 1018 (10th Cir. 2005). This implies, and courts sometimes say, that the doctrine does not apply between separate suits. E.g., Rekhi v. Wildwood Industries, Inc., 61 F.3d 1313, 1317-18 (7th Cir. 1995); Harbor Ins. Co. v. Essman, 918 F.2d 734, 738 (8th Cir. 1990); 18B Charles Alan Wright et al., Federal Practice and Procedure § 4478, pp. 638-39 (2d ed. 2002). That is certainly true in general, but not in the unusual setting of this case.

The purpose of the doctrine is to provide some measure of stability in a litigation, insulating the litigation to a degree from the vagaries of the different judges (as well as from casual changes of mind when the same judge handles the case throughout) who may handle the case in succession. Waid v. Merrill Area Public Schools, 130 F.3d 1268, 1272 (7th Cir. 1997); United States v. Feldman, 825 F.2d 124, 130 (7th Cir. 1987); Suel v. Secretary of Health Human Services, 192 F.3d 981, 984-85 (Fed. Cir. 1999). That purpose is equally engaged when, as happened here, a case is dismissed on a technical ground, without prejudice, and is then refiled without material change.

But the doctrine is not a straitjacket. Not only does it not govern when the law has changed between judges (Amazon to the contrary notwithstanding, there has been no significant intervening change in law here); there is no duty to adhere to the prior ruling if the second judge (or the same judge, on further reflection, when there has been no changes of judge in midstream) is strongly convinced that the earlier ruling was incorrect. McMasters v. United States, 260 F.3d 814, 818 (7th Cir. 2001); Walsh v. Mellas, 837 F.2d 789, 796-97 (7th Cir. 1988); Chevron USA, Inc. v. Bronster, 363 F.3d 846, 849-50 (9th Cir. 2004); Southco, Inc. v. Kanebridge Corp., 324 F.3d 190, 194 (3d Cir. 2003). That's the position I have arrived at with regard to the two issues pressed by Amazon, after the rebriefing and rearguing of the claim-construction issue that I ordered. So I shall proceed to the merits of Amazon's motion.

Under "Background of the Invention," Pinpoint's patent explains that the patented "invention relates to a system and method for controlling broadcast of and/or customer access to data such as video programs in accordance with objective profile data indicative of the customer's preferences for that data." In a context of multiple video channels that makes it difficult to pick out a preferred program just by channel surfing, viewers would be grateful if the programs they are likely to want to see in particular day-and-time slots actually are shown them in those slots. So, from "customer questionnaires, customer demographics, relevance feedback techniques, default profiles, and the like," Pinpoint (if it were practicing the patent) would create an initial "customer profile that would rank and weight numerically the customer preferences for different types of program at different times of the day and week. The weights would be based on the preferences of each customer for particular program characteristics relative to the preferences of other customers for those characteristics; and so, for example, a customer who had an extreme preference for movies with a romantic content would be given a high number to denote his or her preference for romantic movies.

A content profile would be created for each program ("the initial content profiles are determined from questionnaires completed by `experts' or some sort of customer's panel, are generated from the text of the video programs themselves, and/or are determined by adopting the average of the profiles of those customers who actually watch the video program") characterizing it similarly in numerical terms, and so a movie that was ultraromantic would get a high rating for romance. Customer and content profiles would then be compared mathematically in an agreement matrix with a view to selecting the programs that matched the customer's preferences most closely. So the ultraromantic movie would be recommended to the ultraromantic moviegoer for a particular time slot and either the cable television company or other program distributor would transmit the recommended program to the customer in place of the program otherwise scheduled for that time slot or a set-top box would assign the program to an empty channel. The customer profile would be updated on the basis of the customer's response to the recommendations; the more frequent the response, the closer the profile would converge to his true preferences.

What I have described thus far is, however, merely the "preferred embodiment" of the invention. The patent law requires the patent applicant to specify the best way ("best mode," in patent-speak) of actually making the patented device or, as here, practicing the patented method. 35 U.S.C. § 112; Cardiac Pacemakers, Inc. v. St. Jude Medical, Inc., 381 F.3d 1371, 1378-79 (Fed. Cir. 2004). This is to make sure that the patent conveys information of practical utility, which is the quid that the patent law demands as the quo for giving inventors exclusive rights to practice their inventions. Teleflex, Inc. v. Ficosa North America Corp., 299 F.3d 1313, 1330 (Fed. Cir. 2002); Eli Lilly Co. v. Barr Laboratories, Inc., 251 F.3d 955, 963 (Fed. Cir. 2001). The embodiment of the patent that reflects the best mode is called the preferred embodiment. Other embodiments specified in the patent can also be covered by the patentee's claims, however, and it is the claims that determine the patent's scope.

A section of the patent entitled "Alternative Embodiments of Systems Which Use Agreement Matrix" describes embodiments that are remote from the preferred embodiment. Subsection B of this section describes "techniques of the invention" for matching customer profiles with content profiles in "kiosks" — for example, computers in book stores that would recommend books to customers. Subsection C similarly refers to using the techniques of the invention "for the optimum selection of any chunks of information such as stock market data, print information (e.g., for personalized newspapers), or multimedia information which can be downloaded over networks such as the Internet." Thus "the techniques of the invention may be used to match a potential purchaser to real estate on the market by creating profiles of the characteristic features of a house. . . . The potential purchaser can request his or her `dream home' by giving example houses, by specifying desired characteristics such as range of prices, or by a combination of the two. The agreement matrix would match the customer's profiles to the profiles of the available homes. . . ."

The question is whether these alternative embodiments, which do not use the term "scheduling," or the concept of scheduling in its temporal sense, nevertheless are covered by the patent claims.

The patent lists many claims, but only three (numbers 17, 41, and 43) are argued to be infringed by Amazon. I quote them:

17. A method of scheduling customer access to data from a plurality of data sources, comprising the steps of:

creating at least one customer profile for each eligible recipient of said data, said customer profile indicating the customer's preferences for data having predetermined characteristics;

creating content profiles for each data source of said data, said content profiles indicating the degree of content of said predetermined characteristics in data from each data source;

monitoring which data sources are actually accessed by each recipient; and

updating, without input from each customer, each customer profile in accordance with the content profiles of the data sources actually accessed by that customer to automatically update each customer's actual preferences for said predetermined characteristics.

41. A method of scheduling customer access to data from a plurality of data sources, comprising the steps of:

creating a customer profile for each customer of said plurality of data sources, said customer profile indicating the customer's preferences for predetermined characteristics of the data sources;

monitoring which data sources are actually accessed by each customer; and

updating each customer profile to reflect the frequency of selection of the data sources by customers with customer profiles substantially similar to said each customer profile.

43. A method of scheduling customer access to data from a plurality of data sources, comprising the steps of:

creating at least one customer profile for each eligible recipient of said data, said customer profile including a profile of data previously accessed by said customer;

creating content profiles for each data source of said data, said content profiles reflecting the customer profiles of those customers who have previously accessed said data from each data source;

relating said at least one customer profile with the content profiles for the data available from each data source to the customer;

determining a subset of data having content profiles which are determined in said relating step to most closely match said at least one customer profile; and

presenting said subset of data to said customer for selection.

Notice that all the claims refer in their preamble to "scheduling." The word is defined expressly at the beginning of a section entitled "Scheduling Video Delivery in Accordance with Customer and Content Profiles," as follows: "The introduction of time dimensions makes possible the scheduling of video programs, i.e., the assignment of programs to days and to time slots in accordance with each customer profile." Clearly this is "scheduling" in the temporal sense. Every time the word is used it denotes temporal scheduling — scheduling a program to be seen at a particular time of day. This implicit definition coincides with and reinforces the explicit one. See Irdeto Access, Inc. v. Echostar Satellite Corp., 383 F.3d 1295, 1300-03 (Fed. Cir. 2004); Bell Atlantic Network Services, Inc. v. Covad Communications Group, Inc., 262 F.3d 1258, 1268 (Fed. Cir. 2001); Vitronics Corp. v. Conceptronic, Inc., 90 F.3d 1576, 1582 (Fed. Cir. 1996).

The "kiosk" embodiments do not involve temporal scheduling — and therefore don't use the word "scheduling" but instead use "selection" or "matching." The change of wording is revealing. For while, as Pinpoint emphasizes, alternative embodiments of an invention can be encompassed by the patent claims, this is not automatic. ACCO Brands, Inc. v. Micro Security Devices, Inc., 346 F.3d 1075, 1079 (Fed. Cir. 2003); Unique Concepts, Inc. v. Brown, 939 F.2d 1558, 1561-63 (Fed. Cir. 1991). The patent applicant might, to demonstrate the utility of his invention, list all sorts of potential applications that he was not prepared to claim patent protection for. Or, as may have happened here, he will write narrow claims in order to get them allowed and then slip additional subject matter into the specifications in the hope of being able to claim ownership over that subject matter in an infringement suit later on. Maxwell v. J. Baker, Inc., 86 F.3d 1098, 1106-08 (Fed. Cir. 1996); Genentech, Inc. v. Wellcome Foundation Ltd., 29 F.3d 1555, 1563-64 (Fed. Cir. 1994). In neither case is patent protection enlarged. The scope of the patent depends on the claims allowed by the Patent and Trademark Office, not on the description of various embodiments in the specifications part of the patent. The claims in issue use a term, "scheduling," that means only one thing in the patent.

Pinpoint's choice of terms to use in the patent may well have been intended to skirt prior art that might have invalidated a broader patent, such as the Strubbe patent (U.S. Patent No. 5,223,924). That patent matches characteristics of television programs with customer preferences to predict which programs a particular customer will like. But it doesn't assign the programs to particular day and time slots; and Pinpoint's patent actually states that this is a reason not to invalidate its patent on the basis of the Strubbe patent.

That the word "scheduling" as used in the patent has a temporal dimension is further shown by the fact that Pinpoint's initial patent application included claims that expressly embraced the kiosk embodiments in subsection B. The patent examiner rejected those claims as being obvious in light of prior art consisting of a patent on an "Apparatus and Method of Selecting Video Programs Based on Viewers' Preferences" (the Graves patent, U.S. Patent No. 5,410,344). That patent involves selecting video content by comparing "viewer preference files" to a program's "content coding." The examiner ruled that the patent covered any audiovisual content, not just video, and therefore rendered Pinpoint's kiosk embodiments obvious. The examiner further ruled that using the invention in an in-store kiosk rather than on a television set wasn't novel. Pinpoint at first attempted to distinguish the kiosk claims by contrasting their sophisticated agreement matrix with the Graves patent's more primitive matching system, but ultimately abandoned the attempt and dropped the kiosk claims from the patent application.

Pinpoint may not circumvent the rejection of its kiosk claims by claiming that the kiosk embodiments are embraced by the claims that the patent examiner did allow — claims that involve scheduling, as the kiosk embodiments, like the Graves invention, do not (if they did, the examiner would have accepted them). An interpretation of a term in the specifications that encompasses the prior art that inspired the examiner's objection is forbidden, Fantasy Sports Properties, Inc. v. Sportsline.com, Inc., 287 F.3d 1108, 1115 (Fed. Cir. 2002); J M Corp. v. Harley-Davidson, Inc., 269 F.3d 1360, 1367-68 (Fed. Cir. 2001). This "ensures that claims are not construed one way in order to obtain their allowance and in a different way against accused infringers." Chimie v. PPG Industries, Inc., 402 F.3d 1371, 1384-85 (Fed. Cir. 2005).

The alternative embodiments in subsection C were not the subject of claims in the original patent application, so the examiner had no occasion to reject them. But there is nothing to distinguish them from the kiosk embodiments so far as scheduling is concerned. So they cannot be used to embrace systems of matching that do not involve date and time scheduling.

"Customer profile" and "content profile" are likewise not terms defined in the patent. But — coming now to Amazon's second challenge to Judge Conlon's claims construction — whenever they are used other than in the alternative embodiments they mean mathematical constructs of customer preferences and program contents. No nonquantitative characterization, definition, or expression of such preferences and contents is described. The "Background of the Invention" says that Pinpoint's "invention relates to a system and method for controlling broadcast of and/or customer access to data such as video programs in accordance with objective profile data indicative of the customer's preference for that data" (emphasis added); and the only type of "objective data" that the patent refers to are the numerical values for program characteristics preferred by particular customers and found in particular programs. Claim 17 refers to the content profiles as "indicating the degree of content of said predetermined characteristics in data from each data source" and the word I have italicized is used in comparisons, and in this patent evokes the ranking in the preferred embodiment of, for example, movies by the degree of their romantic content. The lead inventors of the patent point out in a related Pinpoint patent (U.S. Patent No. 5,754,939) that any preferred characteristic (director, producer, title, MPAA rating, etc.) "can be replaced by a (usually larger) set of numeric attributes, and hence that any profile can be represented as a vector of numbers denoting the values of these numeric attributes." That would include a preference for one director over another, a preference that, just like a preference for romantic over violent movies, can be expressed mathematically in order to determine intensity ("degree").

It is true that claim 43 does not use the word "degree," and that claim 1 in the other patent in suit, U.S. Patent No. 6,088,722, describes a "method of scheduling customer access to data of a plurality of data objects, comprising the steps of: creating at least one customer profile for a customer of said data, said customer profile indicating the customer's preferences for data having predetermined characteristics; creating content profiles for each of said data objects, said content profiles indicating at least one of the presence and the degree of content of said predetermined characteristics in data of each of said data objects" (emphasis added). Pinpoint argues that the distinction between presence and degree shows that a content profile need not be mathematical and that omission of "degree" from claim 43 shows that the content profile to which that claim refers is indeed nonmathematical. Pinpoint's underlying idea is that you use math only to indicate variation, not mere presence or absence. Yet the content and customer profiles in all the specifications include characteristics with quantifiable values. And it is apparent from the abstract of the 722 patent that that patent, like the 257 patent, envisages profiles and an agreement matrix in which degree as well as presence figures, with degree measured quantitatively rather than qualitatively (as in 100°F versus "hot"). For it states that "from these profiles, an `agreement matrix' is calculated by comparing the recipient's profiles to the actual profiles of the characteristics of the available video programs, movies, or other data. The agreement matrix thus characterizes the attractiveness of each video program, movie, or other data to each prospective customer" (emphasis added).

Had Pinpoint wanted to claim a simple matching system, in which, for example, a customer is asked what type of movie he prefers as between romantic and violent, and an expert is asked to classify movies as either romantic or violent, and the customers who prefer romantic get recommended the movies that the expert has classified as romantic and the ones who prefer violent get the ones the expert has classified as violent, it could have said so. But so simple a matching system, as we know, would be obvious on the basis of the prior art, such as the Strubbe patent. That is why the preamble, abstract, background, and claims sections of the patent describe an invention that schedules video or other programs in particular time slots and decides what to schedule for particular viewers by ingeniously attaching numerical weights to program characteristics and customer preference for such characteristics (in particular time slots) so that matches can be determined with greater precision than if a purely verbal comparison were used.

I conclude that the Pinpoint patent in suit claims only a system of temporal scheduling based on mathematically comparing mathematically expressed customer preferences with mathematically expressed program contents.

When I dismissed the first suit, I voided all previously entered orders, because they had been entered in a suit over which the district court had no jurisdiction. I hereby reinstate Judge Conlon's claims construction as modified by this opinion.