Before us are cross-appeals in a suit arising out of a contract dispute. The appeals are rich with issues, and to discuss them intelligibly we shall have to simplify matters brutally.
In 1988, Olympia Hotels Corporation filed suit against Racine Hotel Partners Limited Partnership charging breach of contract and, by a subsequent amendment to the complaint, conspiracy as well, all in violation of Wisconsin law. The basis of federal jurisdiction was diversity of citizenship. Racine questioned the existence of diversity, claiming that Olympia’s principal place of business was in Wisconsin, the state of which Racine’s partners were citizens, Carden v. Arkoma Associates, — U.S.-, 110 S.Ct. 1015, 108 L.Ed.2d 157 (1990), and not in Texas as Olympia claimed. The jurisdictional issue became moot when Racine filed compulsory counterclaims that charged Olympia not only with breach of contract and with fraud in the inducement of the contract — both claims under Wisconsin law — but also with violations of the federal RICO statute. The RICO claims were not frivolous, and therefore they gave the district court federal-question jurisdiction over the counterclaims, and (because these were compulsory counterclaims) ancillary jurisdiction, equivalent to pendent jurisdiction, over the complaint. Baker v. Gold Seal Liquors, Inc., 417 U.S. 467, 469 n. 1, 94 S.Ct. 2504, 2506 n. 1, 41 L.Ed.2d 243 (1974); Robbins v. Lynch, 836 F.2d 330, 334 (7th Cir.1988); 6 Wright, Miller & Kane, Federal Practice and Procedure § 1414, at p. 99 (2d ed. 1990). Any defects in the invocation of diversity jurisdiction became academic. Well, not completely academic, because ancillary, like pendent, jurisdiction is discretionary, Giardono v. Jones, 867 F.2d 409, 414 (7th Cir.1989), whereas diversity jurisdiction is mandatory. If the district judge had realized that the basis for her exercising jurisdiction over the case might be discretionary rather than mandatory, she might have decided to resolve the issue of Olympia’s citizenship, and if she found that Olympia was a citizen of Wisconsin after all — so that there was no diversity jurisdiction; the only basis for federal jurisdiction was the doctrine of ancillary jurisdiction— have declined to exercise ancillary jurisdiction over the state law claims. But it is too late for all that. The district judge had *1366jurisdiction and exercised it, and the exercise was valid even if she misconceived the basis of that jurisdiction.
The principal counterclaims were not filed until March 23, 1989; and with trial scheduled for May 15, the judge decided to sever the counterclaims (except for Racine’s claim for breach of contract) from the plaintiffs claims and try them later. (Again, if the only basis of jurisdiction over those claims was ancillary jurisdiction and she knew this, the sequence would have been curious — taking the ancillary claims before the main one.) They remain pending in the district court, and there is no trial date. The trial on the plaintiffs claims began as scheduled and lasted two weeks. The voir dire of the jury was conducted by a federal magistrate over Racine’s objection. At the close of all the evidence the judge directed a verdict for Olympia on Racine’s counterclaim for breach of contract and for Racine on Olympia’s claim of civil conspiracy. The jury then returned a verdict for Olympia on its breach of contract claim, awarding $1.2 million in damages. The judge entered judgment for that amount under Rule 54(b) of the Federal Rules of Civil Procedure, and Racine has appealed, with Olympia cross-appealing from the dismissal of its claim of civil conspiracy.
The facts bearing on the legal issues are simple enough. Racine, established to create a first-class hotel in the city of that name, hired Olympia, a hotel-management firm, to build and operate the hotel. The contract had a term of twenty-five years and provided that Olympia would have complete control of the hotel (Racine’s principals had no experience in the hotel business) and would use its best efforts to make the hotel a success. The hotel was built and went into operation, but it was not a success and after several years of operation Racine gave Olympia written notice of default. Racine complained that Olympia had not used its best efforts to make a success of the hotel and that it had reimbursed itself out of the hotel’s revenues for expenses not actually incurred in the hotel’s operation. This suit followed shortly.
The first issue is our appellate jurisdiction. Rule 54(b) authorizes the district court to make immediately appealable a judgment that disposes, with finality, of one or more claims, even though other claims remain pending in the district court so that the suit as a whole has not been finally disposed of by that court. It has seemed to us implicit in the rule that the retained and the appealed claims must be factually distinct, for otherwise the court of appeals may be forced to analyze the same facts in successive appeals, a form of piecemeal appealing not authorized by the rule. Horn v. Transcon Lines, Inc., 898 F.2d 589, 592 (7th Cir.1990); Indiana Harbor Belt R.R. v. American Cyanamid Co., 860 F.2d 1441 (7th Cir.1988); ODC Communications Corp. v. Wenruth Investments, 826 F.2d 509, 512 (7th Cir.1987) (per curiam); FDIC v. Elefant, 790 F.2d 661, 664 (7th Cir.1986); A/S Apothekernes Laboratorium for Specialpraeparater v. I.M.C. Chemical Group, Inc., 725 F.2d 1140 (7th Cir.1984).
This is a borderline case. The claims for breach of contract and for civil conspiracy that were tried and that the parties are trying to bring before us in their appeals concern the parties’ conduct after the contract was signed and the hotel built and in operation, while the counterclaims that await trial in the district court concern promises that Olympia is alleged to have made when the contract was first being negotiated. Yet the claims and counterclaims are of course closely related; if they were not, the counterclaims would not be compulsory; and we shall see that Racine’s principal ground of appeal is that the district judge should have let it prove fraud as a defense to Olympia’s claim for breach of contract. The fact that one claim appears in the complaint and another in a counterclaim, moreover, does not make them different claims for purposes of Rule 54(b). Curtiss-Wright Corp. v. General Electric Co., 446 U.S. 1, 9, 100 S.Ct. 1460, 1465, 64 L.Ed.2d 1 (1980); Automatic Liquid Packaging, Inc. v. Dominik, 852 F.2d 1036, 1037 (7th Cir.1988).
*1367This is not, however, a ease in which a party “merely gave different legal characterizations to the same facts.” A/S Apothekernes Laboratorium for Specialpraeparater v. I.M.C. Chemical Group, Inc., supra, 725 F.2d at 1143. It is a case in which although the claims arise from a single dispute or factual setting, there are many factual differences between them, as in Automatic Liquid Packaging, supra; Jack Walters & Sons Corp. v. Morton Building, Inc., 737 F.2d 698, 701-03 (7th Cir.1984), and Minority Police Officers Ass’n v. City of South Bend, 721 F.2d 197 (7th Cir.1983). Indeed, it appears that virtually the only facts that overlap are those that are not in dispute.
The only rub is the statement in a number of opinions illustrated by Automatic Liquid Packaging, Inc. v. Dominik, supra, 852 F.2d at 1038; Local P-171, Amalgamated Meat Cutters v. Thompson Farms, 642 F.2d 1065, 1071 (7th Cir.1981), and Cinerama, Inc. v. Sweet Music, S.A., 482 F.2d 66, 69 (2d Cir.1973) (Friendly, J.), that a Rule 54(b) judgment may not be entered if two claims are the same for purposes of res judicata; that is, if failing to join them in the same suit would constitute the “splitting” of a single claim. If this were the law, then such a judgment could never be entered in a case in which a compulsory counterclaim remained pending in the district court, and yet we know it can be. Cold Metal Process Co. v. United Engineering & Foundry Co., 351 U.S. 445, 452, 76 S.Ct. 904, 908, 100 L.Ed. 1311 (1956); Automatic Liquid Packaging, Inc. v. Dominik, supra, 852 F.2d at 1037-38. A compulsory counterclaim is by definition a counterclaim that must be filed to preserve the defendant’s rights, Baker v. Gold Seal Liquors, Inc., supra, 417 U.S. at 469 n. 1, 94 S.Ct. at 2506 n. 1; if made the basis of a separate suit it would be barred either by the-force of Rule 13(a) itself, or by the prohibition in the doctrine of res judicata against splitting one’s cause of action. Asset Allocation & Management Co. v. Western Employers Ins. Co., 892 F.2d 566, 572 (7th Cir.1989); cf. Southern Construction Co. v. Pickard, 371 U.S. 57, 60, 83 S.Ct. 108, 110, 9 L.Ed.2d 31 (1962) (per curiam).
In the present case, even though the two sets of claims arise out of the same transaction (the contract for the management of Racine’s hotel), the pertinent facts — the facts bearing on liability, damages, etc. — are different, so that, in the event that the counterclaims eventually are tried and the judgment appealed to us, we shall not have to reacquaint ourselves with the same facts that we had to learn in order to decide the present appeal. The fact that two claims are one for purposes of res judicata may be relevant in appraising the possibility of a Rule 54(b) judgment, as in Judge Friendly’s Cinerama case, in which one claim was for the principal of an unpaid loan and the other for the unpaid interest on the loan, or in our Automatic Packaging case, where the counterclaim was “the mirror image of the complaint,” 852 F.2d at 1038 — where indeed the complaint and counterclaim were “the same claim, expressing the parties’ opposed interpretations of” their contract. Id. But the res judicata status of the two claims is not conclusive under Rule 54(b), and is probably a diversion from the main issue.
So too we think that some of our previous cases place too much weight on the existence and extent of factual overlap between the two claims. What is true is that if the overlap is complete the claims are the same, the only possible difference being the legal theory in which they have been wrapped. If the same set of facts is alleged as a breach of contract and as a breach of a statutory duty, or as a violation of federal and of state securities laws, or as a fraud and as mutual mistake, then as a practical matter there is only one claim, and a Rule 54(b) judgment cannot be entered. If however there is some but not complete factual overlap between nominally separate claims, this circumstance should invite an exercise of discretion by the district court rather than a determination by us that the retained and appealed claims are or are not separate. “It was therefore proper for the District Judge here to consider such factors as ... whether the nature of the claims, already determined was such that no appellate court *1368would have to decide the same issues more than once even if there were subsequent appeals.” Curtiss-Wright Co. v. General Electric Co., supra, 446 U.S. at 8, 100 S.Ct. at 1465 (emphasis added; footnote omitted).
In sum, it is not a decisive consideration, in determining whether claims are one or more than one for purposes of Rule 54(b), either that they are one claim for purposes of res judicata or that there is considerable but not complete factual overlap between them. If the claims are legally distinct and involve at least some separate facts, the district court has the power to enter a Rule 54(b) judgment, and it becomes a matter of the district judge’s discretion, reviewable for but only for abuse thereof, whether to exercise the power and enter such a judgment. The conditions that define the district court’s power are satisfied here, and there was no abuse of discretion. Not only is “one appropriate use of Rule 54(b) ... the entry of judgment on the principal claims of the suit while reserving disposition of counterclaims that may require a longer time to resolve,” FDIC v. Elefant, supra, 790 F.2d at 664, but we shall see that there are compelling pragmatic reasons (the very stuff of discretionary judgments) for deciding this appeal on the merits rather than dismissing it.
On the merits, the first question presented by Racine’s appeal is whether a federal magistrate is empowered to conduct the voir dire in a civil jury trial over the objection of one or more of the parties. The parties describe the question as being whether Gomez v. United States, — U.S. -, 109 S.Ct. 2237, 104 L.Ed.2d 923 (1989), which held that the magistrates statute, 28 U.S.C. § 636, does not authorize magistrates to conduct voir dire in felony trials (at least where the defendant objects, United States v. Wey, 895 F.2d 429, 431 (7th Cir.1990)), should be “extended” to civil cases. We conceive the question differently, as whether the statute authorizes magistrates to conduct voir dire in a civil case. We cannot find where it does, unless it is in section 636(b)(3), authorizing the assignment to a magistrate of “such additional duties as are not inconsistent with the Constitution and laws of the United States.” The location of this provision in the middle of the statute rather than at the end makes us doubt that it was intended to be as comprehensive a catch-all as its words literally suggest. If it were, Gomez would have been decided differently, since section 636(b)(3) is not limited to civil cases. After Gomez, the dictum in In re Establishment Inspection, 589 F.2d 1335, 1340-41 (7th Cir.1979), that section 636(b)(3) should be read as broadly as its words suggests is not' authoritative. (And dictum it was, because all that was at issue was the magistrate’s power to issue an inspection warrant.) Nor would there be much point to the elaborate provisions in section 636(c) for the conduct of civil trials (including jury trials) by a magistrate with the consent of both parties if a district judge could compel the parties, against their wishes, to submit to a magistrate’s conducting vital stages in the trial, such as the voir dire of the jury.
It might even violate the Constitution to allow a magistrate to conduct so vital a stage of a trial without the parties’ consent. Although this is not a pure diversity suit — it may not be any kind of diversity suit — the federal courts have jurisdiction over it only by virtue of the grant of judicial power in Article III of the Constitution. There is no suggestion that the district court was exercising power conferred by any other provision of the Constitution, as it might have been if this were a “core” controversy in a bankruptcy suit or a suit to which the United States was a party. Article III confers the judicial power of the United States on judicial officers who have guarantees of tenure and of undiminished compensation that federal magistrates lack. Parties may be able to consent to have their legal disputes resolved by non-Article III officers even if the dispute is within the judicial power conferred by Article III, since after all they can consent, if they want, to have the dispute resolved by a panel of private arbitrators, by a retired judge, or for that matter by a random number generator. But they cannot be *1369forced to try a dispute that is subject to federal jurisdiction only by virtue of Article III before a judge who is not authorized to exercise the power conferred by that article. Lovelace v. Dall, 820 F.2d 223, 225 (7th Cir.1987) (per curiam); Adams v. Heckler, 794 F.2d 303, 307 (7th Cir.1986); Geras v. Lafayette Display Fixtures, Inc., 742 F.2d 1037, 1041 (7th Cir.1984); Pacemaker Diagnostic Clinic of America, Inc. v. Instromedix, Inc., 725 F.2d 537 (9th Cir.1984) (en banc).
The magistrate here did not conduct the entire trial, however, but only the voir dire; and it can be argued that the voir dire is no more an essential, nondelegable stage of trial than pretrial discovery, which the statute — without thereby engendering any constitutional qualms — authorizes magistrates to conduct without the parties’ consent. 28 U.S.C. § 636(b)(1)(A). We are doubtful whether these are symmetrical exercises of judicial power. Pretrial discovery is conducted largely by the parties on their own, and of course out of court; judicial supervision is minimal. The voir dire, in contrast, is a vital stage of every jury trial. It is the jurors’ first encounter with the court; and the presence of the judge who will preside at trial helps impress on the jurors the gravity of their mission. It is also the judge’s best opportunity to “size up” the jury, because it will probably be the only occasion on which any of the jurors speak in the judge’s presence. Sizing up the jury is important to the judge’s rulings on evi-dentiary questions, on motions for mistrials and new trials, and on other matters requiring an assessment of the particular jury’s ability and attentiveness; on these questions we defer broadly to the trial judge’s judgment, in part because of his superior opportunity to evaluate the jurors. United States v. Bruscino, 687 F.2d 938, 941 (7th Cir.1982) (en banc). The trial is diminished when different judicial officers preside at the voir dire and at the presentation of evidence; the pretrial, much less so.
Whether the trial is so diminished by the use of magistrates to conduct the voir dire that Article III is violated we do not decide. Nor do we invoke the hoary “canon of construction” which teaches that a statute should be so interpreted as to prevent constitutional doubts from arising — a principle that if taken seriously would make the Constitution spill over its already wide banks. The constitutional doubts that we have expressed are rooted in a serious concern about the rights of persons to a trial before a federal judge, a concern likely to be shared by legislators in their reflective moments. For that reason these doubts reinforce our conclusion that section 636 was not intended and should not be read to confer on magistrates, by means of a vague clause buried deep in the statute, the power to conduct jury voir dire without the consent of both parties.
Nor in fact does Olympia contend otherwise, its argument being that the error was harmless in this case. It is indeed harmless in the sense that Racine has made no effort to show how it was harmed. But issues of entitlement to a particular kind of tribunal are in general not subject to the harmless error rule, Walberg v. Israel, 766 F.2d 1071, 1074 (7th Cir.1985), and this regardless of whether the entitlement is jurisdictional in the sense of not being waivable. For example, it is reversible error to deny a party his right to a jury trial; he is not required to show that he probably would have won before a jury. Rose v. Clark, 478 U.S. 570, 578, 106 S.Ct. 3101, 3106, 92 L.Ed.2d 460 (1986). It is reversible error for a judge to refuse to recuse himself in a case in which the law requires recusal, and reversal will not be avoided by an argument that the losing party has failed to demonstrate what difference another judge would have made. Rose v. Clark, supra, 478 U.S. at 577, 106 S.Ct. at 3105, citing Tumey v. Ohio, 273 U.S. 510, 535, 47 S.Ct. 437, 445, 71 L.Ed. 749 (1927). It is reversible error to deny a party to a jury trial the peremptory challenges to which the rules of procedure entitle him, although it will rarely if ever be possible to show that the trial would have come out differently with a different jury. United States v. Ruuska, 883 F.2d 262, 268 (3d Cir.1989), and cases cited there.
Rulings that could never be reviewed if the harmless error rule were applied, be*1370cause their nature is such as to make a demonstration of harm impossible, make an appealing claim for exemption from the rule (a judge-made exemption — Fed.R. Civ.P. 61 and Fed.R.Crim.P. 52(a), the civil and criminal harmless error rules, are unqualified), precisely so that they can be reviewed by appeal, the normal route for correcting trial errors and assuring minimal uniformity of legal obligation. The ruling allowing the magistrate to conduct jury voir dire in this case was such a ruling.
The principle we are exploring should not be applied fanatically (what principle should be?). Hanson v. Parkside Surgery Center, 872 F.2d 745, 749-50 (6th Cir.1989), held that it was a harmless error to impanel an eight-person jury instead of the six-person jury required by the local rule — the result was to give the parties more, not less, due process. And a minor defect in the conduct of the voir dire — harmlessness squared — is not reversible error. McDonough Power Equipment, Inc. v. Greenwood, 464 U.S. 548, 554-56, 104 S.Ct. 845, 849-50, 78 L.Ed.2d 663 (1984). What is troubling, though, is that the principle which lifts the harmless error rule when application would systematically defeat attempts at appellate correction of an unlawful practice is not applied consistently. In particular, a criminal defendant cannot get his conviction reversed by demonstrating that the prosecutor procured the indictment by knowing use of perjured testimony or by other prosecutorial misconduct. Bank of Nova Scotia v. United States, 487 U.S. 250, 108 S.Ct. 2369, 101 L.Ed.2d 228 (1988); United States v. Mechanik, 475 U.S. 66, 106 S.Ct. 938, 89 L.Ed.2d 50 (1986). So violations of the rule against obtaining an indictment through perjury are unreviewable. Id. at 80, 106 S.Ct. at 946 (dissenting opinion). There is inconsistency within inconsistency, since racial discrimination in the selection of the grand jury vitiates the conviction. Vasquez v. Hillery, 474 U.S. 254, 106 S.Ct. 617, 88 L.Ed.2d 598 (1986), weakly distinguished in Bank of Nova Scotia v. United States, supra, 108 S.Ct. at 2375.
Although it may not be possible to extract from the cases a completely coherent principle for deciding when the harmless error rule applies and when it does not, the clincher here is that Gomez refused to apply the rule. 109 S.Ct. at 2248. It was a criminal case, but the criminal and civil harmless error rules are the same. The Court emphasized the fundamental character of a defendant’s right to be tried by a person with jurisdiction to preside, id., and the right is no less fundamental in a civil case. That it was a partial trial — just the jury voir dire — made no difference in the Court’s eyes in Gomez, and we cannot think why it should make a greater difference here. Racine is entitled to a new trial.
By now it should be apparent why we said that there are pragmatic reasons for the use of Rule 54(b) to permit an immediate appeal in this case. If we dismissed the appeal, the next stage in this proceeding would be a trial of Racine’s fraud and RICO counterclaims, resulting in a final judgment from which one or both parties would appeal — and Racine’s appeal would bring up its claim that voir dire by the magistrate was improper, thus requiring that the entire proceeding be tried over. An appellate court must do what it can to expedite the trial process in this era of steep and growing caseloads. In compliance with this precept we shall now, for the guidance of the parties and the district judge on remand, resolve the principal remaining issues raised by the appeal and the cross-appeal.
The first is whether Racine should have been permitted to plead fraud as an affirmative defense to Olympia’s claim for breach of contract. The district judge’s ground for excluding this defense was not, as one might have expected, the belatedness of the attempt to plead the defense, an attempt made only weeks before the trial was to (and did) begin. That ground would have failed. Olympia has never argued that it would have been prejudiced by the late pleading, for it had long been well aware that Racine was contending — albeit the contention was first formalized in the counterclaim — that it had been induced to *1371sign the contract by the misrepresentations that form the basis of its counterclaim. The district judge’s ground was, rather, that by seeking damages for breach of contract Racine had affirmed the contract and therefore could not seek to rescind it, or even defend itself against its being enforced, on the basis of fraud — or on any other basis, for that matter.
The common law doctrine of election of (contract) remedies that the district judge invoked in so ruling has two aspects, a procedural and a substantive. The procedural aspect derives from the overriding goal of common law pleading, which was by successive rounds of pleading to narrow the issues until there was just one for trial. 5 Wright & Miller, Federal Practice and Procedure § 1202, at p. 61 (1969). It was essential to the attainment of this goal that a party be forbidden to plead in the alternative, for that would generate two or more issues for trial. He must therefore elect his remedy. Applied to this case that would mean that Racine had to sue either for breach of the contract or to disaffirm the contract as having been induced by fraud. Common law pleading was superseded long ago, however — in the federal courts by the Federal Rules of Civil Procedure, which expressly abolish election of remedies. Fed.R.Civ.P. 8(e)(2). Those rules of course apply in all federal civil litigation, even if the issue being litigated is one of state law, as noted with specific reference to election of remedies in Koedding v. Slaughter, 481 F.Supp. 1238, 1237 (E.D.Mo.1979), aff’d, 634 F.2d 1095 (8th Cir.1980).
In its substantive aspect, however, the doctrine of election of remedies is not affected by the federal rules of procedure. In that aspect the doctrine is a part of the law of remedies rather than of procedural law. It seeks to prevent double recovery. Wynfield Inns v. Edward LeRoux Group, Inc., 896 F.2d 483, 488 (11th Cir.1990). If Racine was, as it claims, overcharged in the administration of the contract, it would be entitled to damages for breach. And if it was induced to sign the contract by fraud, it would be entitled to damages arising from the fraud, including any overcharges by Olympia. But it would not be entitled to collect the overcharges twice, once as damages for breach of contract and the second time as damages for fraud.
That is all there is to the doctrine of election of remedies, viewed as a doctrine of the law of remedies rather than as a pleading doctrine. Election of remedies, the Supreme Court of Wisconsin has held (quoting with emphatic approval a decision from New Hampshire), “should be confined to cases where the plaintiff may be unjustly enriched or the defendant has actually been misled by the plaintiffs conduct or the result is otherwise inequitable or res judicata can be applied.” Schlotthauer v. Krenzelok, 274 Wis. 1, 6, 79 N.W.2d 76, 79 (1956). See also Bank of Commerce v. Paine, Webber, Jackson & Curtis, 39 Wis.2d 30, 36-37, 39, 158 N.W.2d 350, 352, 354 (1968); Tuchalski v. Moczynski, 152 Wis.2d 517, 520, 449 N.W.2d 292, 293 (Ct.App.1989) (“the real purpose of the doctrine is to prevent double recovery”).
The reference in Schlotthauer (repeated in Bank of Commerce) to equity and to res judicata place in perspective the broad terms in which the doctrine is stated in a number of other Wisconsin decisions, such as Beers v. Atlas Assurance Co., 231 Wis. 361, 368, 285 N.W. 794, 797 (1939): “a party may not affirm a contract and later on disaffirm it and ask for rescission.” See also Jolin v. Oster, 55 Wis.2d 199, 205, 198 N.W.2d 639, 642 (1972); Stadler v. Rohm, 40 Wis.2d 328, 335-36, 161 N.W.2d 906, 909-10 (1968). The point of these dicta is not that an election of remedies is required but that once made it may be final either by operation of res judicata or because a change of ground would create unreasonable surprise or unduly complicate or delay the litigation.
It would be unreasonable to make Racine choose between the two forms of damages before trial and verdict. Suppose its contract damages were $10,000 and its fraud damages $20,000. The jury might find that Racine had a good claim for breach of contract but not for fraud, and in that event Racine would be entitled to $10,000. *1372If Racine had been forced to elect before trial, and had elected to drop the contract count, Racine would get nothing. Alternatively the jury might find that Racine had a good claim for fraud but not for breach of contract, and then Racine would be entitled to $20,000. If it had elected its contract remedy and dropped the fraud remedy, it would again get nothing. Both results are unsound and they are prevented by confining the doctrine of election of remedies to the limited office of preventing duplicative damages awards, and thus of placing a ceiling of $20,000 on Racine’s damages in our hypothetical case.
The district judge did not even apply the doctrine of election of remedies consistently. Applied consistently it would bar Racine from ever proving fraud in the inducement of the contract. But the judge has not dismissed Racine’s counterclaims, which are based on fraud. It may be questioned whether she has applied the doctrine at all, but the important point is that it is inapplicable unless and until the jury returns a verdict that provides duplicative recovery.
We turn to the district judge’s dismissal of Racine’s counterclaim for breach of contract. Her ground was that Racine had failed to produce evidence from which the jury could have determined what its damages were. This ground has two flaws. The first, which Racine’s practical-minded lawyers sensibly do not bother to argue but which we mention for possible future reference, is that a failure of proof of damages does not justify the dismissal of a claim for breach of contract, as it does most tort claims. The victim of a breach of contract is always entitled to nominal damages if he proves a breach but no damages. Vasselos v. Greek Orthodox Community, 24 Wis.2d 376, 129 N.W.2d 243 (1964); Farnsworth, Contracts 838-39 (1982); Dobbs, Remedies 817-18 (1973). The victim of a tort, usually not. Restatement (Second) of Torts, § 907, comment b (1979); Prosser and Keeton on the Law of Torts 845 (5th ed. 1984). This is because, in general, there is no tort without harm. AMPAT/Midwest, Inc. v. Illinois Tool Works Inc., 896 F.2d 1035, 1041 (7th Cir.1990).
The second flaw in the district court’s disposition of the counterclaim for breach of contract, which is fatal, is that a plaintiff is not required to propose a specific damages figure to a jury, even if by his reticence he is angling for a compromise verdict. It is common for the plaintiff’s counsel to leave the determination of the figure to the jury. Counsel may fear that if he proposes a figure that he can defend with reference to the evidence, it will strike the jury as too high, but that if he proposes a lower figure his opponent will argue that the plaintiff lacks the courage of his convictions. We have found no case either expressly approving or expressly disapproving the practice, but while it would no doubt be within our power to lay down a rule requiring the plaintiff always to specify his damages — and indeed we believe this to be the better practice — we do not find the existing practice so troubling as to require a circuit-wide rule.
The essential thing is that the evidence enable the jury to come up with a figure that represents a reasonable estimate of the plaintiff’s damages. Eastman Kodak Co. v. Southern Photo Materials Co., 273 U.S. 359, 379, 47 S.Ct. 400, 405, 71 L.Ed. 684 (1927). That condition was satisfied. Racine presented evidence that Olympia had billed it for certain expenses in fact incurred by Olympia’s management in other properties under its control; these expenses were specified and the jury could easily have added them up and awarded the sum as damages. Racine also presented evidence that Olympia had failed to use its best efforts to make the hotel a success, and it asked the jury to make Olympia refund a portion of the management fees that Racine had paid under the contract. In effect Racine was asking the jury to estimate the discount to which Racine was entitled because Olympia had failed to exert itself to the degree required by the contract. The task was speculative, but would not have been less so if Racine had plucked a number out of the air and asked for that as its discount. If Olympia failed *1373to use its best efforts, a jury question, then it was a reasonable inference that the hotel would make less money for its owner, Racine. How much less was unclear. But as it was not Racine’s fault that this quantity could not be estimated with precision, the jury was allowed to guess, Thorp Sales Corp. v. Gyuro Grading Co., 111 Wis.2d 431, 442-44, 331 N.W.2d 342, 348-49 (1983); Brogan v. Industrial Casualty Ins. Co., 132 Wis.2d 229, 240, 392 N.W.2d 439, 444 (Ct.App.1986), within limits not tested since the issue was never submitted to the jury.
But we disagree with Racine’s argument that to help the jury determine what the contract term “best efforts” meant, Racine should have been allowed to present evidence concerning the parties’ negotiations before the contract was signed. The contract contains an integration clause, and the district judge was correct that the parol evidence rule forbade inquiry into precontractual discussions or agreements concerning the meaning of best efforts. Of course if the contract was procured by fraud, the integration clause would not prevent inquiry into the parties’ discussions before the contract was signed; the integration clause would go down the drain with the contract of which it was a part. Morse Chain Co. v. T.W. Meiklejohn, Inc., 237 Wis. 383, 388, 296 N.W. 106, 109 (1941). The best efforts issue would be academic anyway, for remember that you cannot collect both damages for breach of contract and damages for having been induced to make the contract. But the issue of fraud had yet to be determined (had in fact been severed for a later trial) when Racine’s counterclaim for breach of contract was tried.
The parol evidence rule is maligned in some circles as the vestige of an era when judges were hostile to plaintiffs and mistrusted juries, and thus as an arbitrary barrier to getting at the truth. But it has stubbornly refused to die, In re Spring Valley Meats, 94 Wis.2d 600, 288 N.W.2d 852 (1980), and in fact it serves an important social purpose, which it shares with other doctrines (such as the “four corners” rule) that overlap it. Patton v. Mid-Continent Systems, Inc., 841 F.2d 742, 745 (7th Cir.1988); FDIC v. W.R. Grace Co., 877 F.2d 614, 620-22 (7th Cir.1989). Not all parties to contracts want to entrust their fate to the vagaries of juries unversed in the usages of business, and among the devices that they use to reduce this risk is the integration clause, which precludes inquiry into preliminary discussions unless, as in Central Auto Co. v. Reichert, 87 Wis.2d 9, 17, 273 N.W.2d 360, 365 (Ct.App.1978), such inquiry is necessary in order to disambiguate the contract. It is not necessary here, as it was not in Western Contracting Corp. v. Dow Chemical Co., 664 F.2d 1097, 1100 (8th Cir.1981), or in St. Louis Union Trust Co. v. United States, 617 F.2d 1293, 1300 (8th Cir.1980). It was in Petition of Minnesota Power & Light Co., 435 N.W.2d 550, 562-63 (Minn.App.1989), but only because the scope of the best efforts clause was conceded to be uncertain in that case. The term “best efforts” is a familiar one in contract parlance, and its meaning is especially plain in a case such as this where the promisor has similar contracts with other promisees. In such a case “best efforts” means the efforts the promisor has employed in those parallel contracts where the adequacy of his efforts have not been questioned. If Olympia worked as hard for Racine as it did for its other, but noncomplaining, customers, then it was using its best efforts within the meaning of the contract.
We also disagree with Racine’s contention that Olympia owed it the superi- or fidelity of an agent — a fiduciary. A fiduciary, unlike an ordinary contract promisor, undertakes to treat the affairs of the promisee as if they were the promisor's own affairs. That is the practical content of all that high falutin’ talk of utmost good faith and loyalty, full disclosure, the punctilio of an honor most sensitive, etc. Shevel v. Warter, 256 Wis. 503, 41 N.W.2d 603 (1950); Faultersack v. Clintonville Sales Corp., 253 Wis. 432, 34 N.W.2d 682 (1948); Meinhard v. Salmon, 249 N.Y. 458, 464, 164 N.E. 545, 546 (1928) (Cardozo, C.J.). The promisor is to treat the promisee as well, as loyally, as considerately, as faith*1374fully, as the promisor would treat himself. Racine bases this construal of the contract on a provision that entitled Olympia — “solely in an agency function,” as the contract put it — to make deposits in and disbursements from the bank accounts that Racine established in the name of the hotel. This provision made Olympia the agent of Racine, all right, but only with regard to the management of the bank accounts. The agency was a part of Olympia’s contractual undertaking; it was not the whole of it. It did not make the best efforts clause, for example, a fiduciary endeavor. Olympia is in the business of managing hotels, using a form of contract that Racine has not attempted to show is atypical of the hotel management business. We decline the invitation to recast hotel contract managers as trustees or guardians of the hotel owners, cf. Kohl v. F.J.A. Christiansen Roofing Co., 95 Wis.2d 27, 34-35, 289 N.W.2d 329, 333 (Ct.App.1988), other than in the performance of financial functions in which the manager is exercising a traditional fiduciary role, such as the management of bank accounts.
The last question we discuss is whether the district judge was correct to withdraw the issue of civil conspiracy from the jury, on the ground that Olympia had failed to present sufficient evidence to warrant a reasonable jury in finding an actionable conspiracy. When Racine soured on Olympia it approached another hotel management concern, Aircoa, with which it signed a contract for hotel management services that was expressly contingent on Racine’s terminating its contract with Olympia. Olympia claims that Aircoa and Racine had agreed sotto voce that, if necessary, Racine would cancel the contract on trumped-up charges.
If the claim is factually supported, Air-coa is guilty of the tort of intentional interference with contract. Wis.Stat. § 134.01; Onderdonk v. Lamb, 79 Wis.2d 241, 255 N.W.2d 507 (1977); Frandsen v. Jensen-Sundquist Agency, Inc., 802 F.2d 941, 947 (7th Cir.1986). Since a tort is a species of unlawful behavior, an agreement to commit a tort is a conspiracy, and Wisconsin law makes civil conspiracy a tort. We have not been told why Olympia preferred to prove a conspiracy rather than to name Aircoa (a large firm) as an additional defendant; maybe Olympia could not have obtained personal jurisdiction over Aircoa in Wisconsin. In any event, Olympia took the conspiracy route. This was its privilege, however one might wish to question as an original matter the growing tendency to recast breach of contract suits as tort suits in an effort to get a shot at punitive damages — especially when courts are increasingly prone to award punitive damages for breach of contract directly, upon proof that the breach was willful even though not necessarily tortious. Patton v. Mid-Continent Systems, Inc., supra, 841 F.2d at 750. Wisconsin has resisted that trend, Loehrke v. Wanta Builders, Inc., 151 Wis.2d 695, 701, 445 N.W.2d 717, 720 (Ct.App.1989)— only to succumb to the seductions of conspiracy doctrine.
The tort of interference with contract, unlike civil conspiracy, serves a distinct role in a common law system of remedies. The contract breaker may lack the wherewithal to respond in damages. Frandsen v. Jensen-Sundquist Agency, Inc., supra, 802 F.2d at 947. This is often the case when the breach is an employee’s breach of an employment contract induced by a rival employer, the tortfeasor. The rationale for the tort is therefore similar to the rationale for the doctrine of respondeat superior. The present ease is not one to which the rationale extends, although we do not suggest that the rule is cabined by its rationale; rules often, for the sake of clarity and definiteness, carry farther than the policy behind them. The defendant in Olympia’s tort suit is the contract breaker. The tort claim is an indirect method of obtaining punitive damages without having to prove a willful breach; for the fact that Racine may have promised Aircoa to terminate the contract on trumped-up charges if necessary does not show that it did terminate the contract on such grounds.
At all events, we agree with the district judge that Olympia failed to present admissible evidence of conspiracy. It proved a contract between Racine and Aircoa contin*1375gent on the termination of the contract with Olympia, but not that the parties to the contract further agreed that if necessary Racine would invoke fraudulent grounds for the termination. Maybe a rational jury could infer that Racine was determined to terminate the contract by hook or by crook. But the conspiracy charged is a conspiracy to interfere tortiously with Racine’s contract with Olympia. You cannot interfere tortiously with your own contract, so the conspiracy had to be that Aircoa would interfere tortiously; but there is not the slightest evidence that Air-coa intended or agreed to do anything more than sign with Racine if Racine managed to get out of its contract with Olympia.
We have ignored a few minor issues that may recur on remand but we have decided the major ones and by doing so have charted the future course of this suit in the district court; perhaps this will facilitate a settlement. To recapitulate, Racine is entitled to a new trial at which the jury voir dire will be conducted by the district judge rather than by a magistrate and at which Racine will be entitled to prove fraud as an affirmative defense and to submit its contract damages claim to the jury without naming a specific figure. The judgment in Racine’s favor on Olympia’s civil conspiracy. claim is affirmed, as are the district judge’s rulings on the parol evidence rule and the agency issue. Costs in this court to Racine. In accordance with Circuit Rule 36, the new trial that we are ordering shall be before a different district judge. Lest there be any doubt as to the consequences for the severed counterclaims which remain pending in the district court, we direct that they be reassigned to the same judge in order to facilitate an orderly disposition of the entire case.
Affirmed in Part, Reversed in Part, and Remanded With Directions.