POSNER, Circuit Judge.

This is a diversity suit, governed by Indiana law, in which substantial damages are sought on the basis of promissory estoppel. The suit pits Garwood Packaging, Inc., which created a packaging system designed to increase the shelf life of fresh meat, and its two principals, Garwood and McNamara, against Allen Company (an investment company) and a vice-president of Allen named Martin. We shall refer to the plaintiffs collectively as "GPI" and the defendants collectively as "Allen." The district court granted summary judgment in favor of Allen and dismissed the suit.

There is a threshold question: whether GPI's appeal was timely. The notice of appeal was filed within 30 days of the entry of judgment, but the judgment had not dismissed the suit as to one of the defendants and was therefore, on its face anyway, not final and appealable. GPI then dismissed its suit against the remaining defendant and filed a new notice of appeal, but did so more than 30 days after the dismissal of that defendant had removed the cloud on the finality of the district court's judgment. The district judge granted GPI's motion to file a late notice of appeal, but failed in doing so to make a finding that GPI's tardiness had been due to "excusable neglect." A district court may extend the time for filing a notice of appeal only if the appellant demonstrates to the court's satisfaction "excusable neglect or good cause." Fed.R.App.P. 4(a)(5)(A)(ii). The ruling is regarded as discretionary, e.g., United States v. Brown, 133 F.3d 993, 996 (7th Cir. 1998); Brotherhood of Ry. Carmen v. Chicago North Western Transportation Co., 964 F.2d 684, 686 (7th Cir. 1992); Silivanch v. Celebrity Cruises, Inc., 333 F.3d 355, 362 (2d Cir. 2003), and so when there is no indication that discretion was actually exercised, a remand is necessary unless the issue is so one-sided (which is not the case here) that it could have been resolved only one way. Cf. Prizevoits v. Indiana Bell Tel. Co., 76 F.3d 132, 133-34 (7th Cir. 1996).

No remand is necessary here on a different ground, or rather grounds. One is that the notice of appeal may not have been premature, because the judgment may already have been final. The defendant who was not formally dismissed from the case at the same time as the other defendants had never been served with the complaint, and it was much too late to serve him by the time the judgment was entered against the other defendants. Since he had never become and never could become a party, the judgment that did not mention him was nevertheless final, complete, and appealable. Manley v. City of Chicago, 236 F.3d 392, 395 (7th Cir. 2001); Ordower v. Feldman, 826 F.2d 1569, 1571-73 (7th Cir. 1987); Federal Savings Loan Ins. Corp. v. Tullos-Pierremont, 894 F.2d 1469 (5th Cir. 1990), and cases cited there.

It might be objected that these cases conflate a good reason for entering a final decision with entry of the final decision itself. Suppose a plaintiff filed suit and then appealed the same day, before the district court even looked at the case, and defended his precipitate action by saying that he knew he would lose in the district court under circuit precedent (which he would urge the court of appeals to overrule, but which would bind the district court) and since his suit was doomed in the district court it was as if there were a final judgment and he should therefore be able to appeal immediately. But in the cases that we have cited, as in the present case, the defendant who hadn't been dismissed with the others had never actually become a party because he had never been served. The significance of the fact that he could no longer be served was that the dismissal of the suit could not be regarded as a dismissal without prejudice as to him; it was therefore securely final.

The alternative ground for regarding the decision of the district court as final and appealable is based on Rule 4(a)(2) of the Federal Rules of Appellate Procedure. The rule provides that a notice of appeal filed after the court announces its decision but before the judgment is entered shall be treated as if filed when the judgment was entered. In other words, once the decision is announced, a premature notice of appeal lingers until the final decision is entered. FirstTier Mortgage Co. v. Investors Mortgage Ins. Co., 498 U.S. 269 (1991); Otis v. City of Chicago, 29 F.3d 1159, 1166 (7th Cir. 1994) (en banc). Here a decision was announced and a notice of appeal filed (the first notice). It took effect when, the last defendant having been dismissed, the decision became final.

So the appeal was timely and we can proceed to the merits.

GPI had flopped in marketing its food-packaging system and by 1993 had run up debts of $3 million and was broke. It engaged Martin to help find investors. After an initial search turned up nothing, Martin told GPI that Allen (Martin's employer, remember) would consider investing $2 million of its own money in GPI if another investor could be found who would make a comparable investment. The presence of the other investor would reduce the risk to Allen not only by augmenting GPI's assets but also by validating Allen's judgment that GPI might be salvageable, because it would show that someone else was also willing to bet a substantial sum of money on GPI's being salvageable. To further reduce its risk Allen decided to off-load half its projected $2 million investment on other investors.

Martin located a company named Hobart Corporation that was prepared to manufacture $2 million worth of GPI packaging machines in return for equity in the company. Negotiations with Hobart proved arduous, however. There were two sticking points: the amount of equity that Hobart would receive and the obtaining of releases from GPI's creditors. Hobart may have been concerned that unless the creditors released GPI the company would fail and Hobart wouldn't be able to sell the packaging systems that it manufactured. Or it may have feared that the creditors would assert liens in the systems. All that is clear is that Hobart insisted on releases. They were also important to the other investors whom Allen wanted to bring into the deal, the ones who would contribute half of Allen's offered $2 million.

Martin told Garwood and McNamara (GPI's principals) that he would see that the deal went through "come hell or high water." Eventually, however, Allen decided not to invest, the deal collapsed, and GPI was forced to declare bankruptcy. The reason for Allen's change of heart was that the investors who it thought had agreed to put up half of "Allen's" $2 million had gotten cold feet. When Allen withdrew from the deal, no contract had been signed and no agreement had been reached on how much stock either Allen or Hobart would receive in exchange for their contributions to GPI. Nor had releases been obtained from the creditors.

GPI's principal claim on appeal, and the only one we need to discuss (the others fall with it), is that Martin's unequivocal promise to see the deal through to completion bound Allen by the doctrine of promissory estoppel, which makes a promise that induces reasonable reliance legally enforceable. Brown v. Branch, 758 N.E.2d 48, 52 (Ind. 2001); First National Bank of Logansport v. Logan Mfg. Co., 577 N.E.2d 949, 954 (Ind. 1991); Consolidation Services, Inc. v. KeyBank National Ass'n, 185 F.3d 817, 822 (7th Cir. 1999) (Indiana law); Restatement (Second) of Contracts § 90(1) (1981); 1 E. Allan Farnsworth, Farnsworth on Contracts § 2.19 (3d ed. 2004). If noncontractual promises were never enforced, reliance on their being enforceable would never be reasonable, so let us consider why the law might want to allow people to rely on promises that do not create actual contracts and whether the answer can help GPI.

The simplest answer to the "why" question is that the doctrine merely allows reliance to be substituted for consideration as the basis for making a promise enforceable. First National Bank of Logansport v. Logan Mfg. Co., supra, 577 N.E.2d at 954; Workman v. United Parcel Service, Inc., 234 F.3d 998, 1001 (7th Cir. 2000); Consolidation Services, Inc. v. KeyBank National Ass'n, supra, 185 F.3d at 822; Porter v. Commissioner, 60 F.2d 673, 675 (2d Cir. 1932) (L. Hand, J.). On this view promissory estoppel is really just a doctrine of contract law. The most persuasive reason for the requirement of consideration in the law of contracts is that in a system in which oral contracts are enforceable — and by juries, to boot — the requirement provides some evidence that there really was a promise that was intended to be relied on as a real commitment. Gibson v. Neighborhood Health Clinics, Inc., 121 F.3d 1126, 1131 (7th Cir. 1997); Scholes v. Lehmann, 56 F.3d 750, 756 (7th Cir. 1995); Krell v. Codman, 28 N.E. 578 (Mass. 1891) (Holmes, J.); Lon L. Fuller, "Consideration and Form," 41 Colum. L. Rev. 799, 799-801 (1941). Actual reliance, in the sense of a costly change of position that cannot be recouped if the reliance turns out to have been misplaced, is substitute evidence that there may well have been such a promise. Consolidation Services, Inc. v. KeyBank National Ass'n, supra, 185 F.3d at 822; Yontz v. BMER Interprises, Inc., 632 N.E.2d 527, 530 (Ohio App. 1993). The inference is especially plausible in a commercial setting, because most businesspeople would be reluctant to incur costs in reliance on a promise that they believed the promisor didn't consider himself legally bound to perform.

In other words, reasonable reliance is seen as nearly as good a reason for thinking there really was a promise as bargained-for reliance is. In many such cases, it is true, no promise was intended, or intended to be legally enforceable; in those cases the application of the doctrine penalizes the defendant for inducing the plaintiff to incur costs of reliance. The penalty is withheld if the reliance was unreasonable; for then the plaintiff's wound was self-inflicted — he should have known better than to rely.

A relevant though puzzling difference between breach of contract and promissory estoppel as grounds for legal relief is that while the promise relied on to trigger an estoppel must be definite in the sense of being clearly a promise and not just a statement of intentions, Security Bank Trust Co. v. Bogard, 494 N.E.2d 965, 968-69 (Ind.App. 1986); Wood v. Mid-Valley Inc., 942 F.2d 425, 428 (7th Cir. 1991) (Indiana law); Major Mat Co. v. Monsanto Co., 969 F.2d 579, 582-83 (7th Cir. 1992), its terms need not be as clear as a contractual promise would have to be in order to be enforceable. E.g., Janke Construction Co. v. Vulcan Materials Co., 527 F.2d 772, 777 (7th Cir. 1976) (Wisconsin law); Hawkins Construction Co. v. Reiman Corp., 511 N.W.2d 113, 117 (Neb. 1994); Neiss v. Ehlers, 899 P.2d 700, 707 (Ore.App. 1995). Indiana may go furthest in this direction: "Even though there were insufficient terms for the enforcement of an express oral contract, and unfulfilled pre-existing conditions prohibiting recovery for breach of a written contract . . ., we are not precluded from finding a promise under these circumstances. Indeed, it is precisely under such circumstances, where a promise is made but which is not enforceable as a `contract,' that the doctrine of promissory estoppel is recognized." First National Bank of Logansport v. Logan Mfg. Co., supra, 577 N.E.2d at 955.

The reason for this difference between breach of contract and promissory estoppel is unclear. A stab at an explanation is found in Rosnick v. Dinsmore, 457 N.W.2d 793, 800 (Neb. 1990), where the court said that "promissory estoppel only provides for damages as justice requires and does not attempt to provide the plaintiff damages based upon the benefit of the bargain. The usual measure of damages under a theory of promissory estoppel is the loss incurred by the promisee in reasonable reliance on the promise, or `reliance damages.' Reliance damages are relatively easy to determine, whereas the determination of `expectation' or `benefit of the bargain' damages available in a contract action requires more detailed proof of the terms of the contract." The only problem with this explanation is that its premise is mistaken; if the promise giving rise to an estoppel is clear, the plaintiff will usually be awarded its value, which would be the equivalent of the expectation measure of damages in an ordinary breach of contract case. Goldstick v. ICM Realty, 788 F.2d 456, 463-64 (7th Cir. 1986); Restatement, supra, § 90 comment d. The rationale in both cases is that the benefit of the contract to the promisee is a good proxy for the opportunities that he forewent in making the contract. Walters v. Marathon Oil Co., 642 F.2d 1098, 1100-01 (7th Cir. 1981); L.L. Fuller William R. Perdue, Jr., "The Reliance Interest in Contract Damages: 1," 46 Yale L.J. 52, 60 (1936) ("physicians with an extensive practice often charge their patients the full office call fee for broken appointments. Such a charge looks on the face of things like a claim to the promised fee; it seems to be based on the `expectation interest'. Yet the physician making the charge will quite justifiably regard it as compensation for the loss of the opportunity to gain a similar fee from a different patient"). Of course, if the promise is unclear, damages will be limited to expenses incurred in reasonable reliance on the vague promise, First National Bank of Logansport v. Logan Mfg. Co., supra, 577 N.E.2d at 952, 955-56, but that would be equally true in a breach of contract case in which the promise that the defendant had broken was unclear.

But even though the court is "not precluded from finding a promise" by its vagueness, id. at 955, the vaguer the alleged promise the less likely it is to be found to be a promise. Mays v. Trump Indiana, Inc., 255 F.3d 351, 358-59 (7th Cir. 2001) (Indiana law); All-Tech Telecom, Inc. v. Amway Corp., 174 F.3d 862, 868-69 (7th Cir. 1999); Restatement, supra, § 33(3) and comment f. And if it is really vague, the promisee would be imprudent to rely on it — he wouldn't know whether reliance was worthwhile. The broader principle, which the requirement that the promise be definite and at least minimally clear instantiates, is that the promisee's reliance must be reasonable; if it is not, then not only is he the gratuitous author of his own disappointment, but probably there wasn't really a promise, or at least a promise intended or likely to induce reliance. The "promise" would have been in the nature of a hope or possibly a prediction rather than a commitment to do something within the "promisor's" power to do ("I promise it will rain tomorrow"); and the "promisee" would, if sensible, understand this. He would rely or not as he chose but he would know that he would have to bear the cost of any disappointment.

We note, returning to the facts of this case, that there was costly reliance by GPI, which forewent other opportunities for salvation, and by Garwood and McNamara, who moved from Indiana to Ohio to be near Hobart's plant where they expected their food-packaging system to be manufactured, and who forgave their personal loans to GPI and incurred other costs as well. The reliance was on statements by Martin, of which "come hell or high water" was the high water mark but is by no means an isolated example. If GPI's evidence is credited, as it must be in the procedural posture of the case, Martin repeatedly confirmed to GPI that the deal would go through, that Allen's commitment to invest $2 million was unconditional, that the funding would be forthcoming, and so on; and these statements induced the plaintiffs to incur costs they would otherwise not have done.

But were these real promises, and likely to be understood as such? Those are two different questions. A person may say something that he intends as merely a prediction, or as a signal of his hopes or intentions, but that is reasonably understood as a promise, and if so, as we know (this is the penal or deterrent function of promissory estoppel), he is bound. Tipton County Farm Bureau Cooperative Ass'n, Inc. v. Hoover, 475 N.E.2d 38, 42 (Ind.App. 1985). But what is a reasonable, and indeed actual, understanding will often depend on the knowledge that the promisee brings to the table. McNamara, with whom Martin primarily dealt, is a former investment banker, not a rube. He knew that in putting together a deal to salvage a failing company there is many a slip `twixt cup and lips. Unless blinded by optimism or desperation he had to know that Martin could not mean literally that the deal would go through "come hell or high water," since if Satan or a tsunami obliterated Ohio that would kill the deal. Even if Allen had dug into its pockets for the full $2 million after the investors who it had hoped would put up half the amount defected, the deal might well not have gone through because of Hobart's demands and because of the creditors. GPI acknowledges that the Internal Revenue Service, one of its largest creditors, wouldn't give a release until paid in full. Some of GPI's other creditors also intended to fight rather than to accept a pittance in exchange for a release. Nothing is more common than for a deal to rescue a failing company to fall apart because all the creditors' consent to the deal cannot be obtained — that is one of the reasons for bankruptcy law. Again these were things of which McNamara was perfectly aware.

The problem, thus, is not that Martin's promises were indefinite, which they were not if GPI's evidence is credited, but that they could not have been reasonably understood by the persons to whom they were addressed (mainly McNamara, the financial partner in GPI) to be promises rather than expressions of optimism and determination. Security Bank Trust Co. v. Bogard, supra, 494 N.E.2d at 968-69; Workman v. United Parcel Service, Inc., supra, 234 F.3d at 1001-02; Wood v. Mid-Valley Inc., supra, 942 F.2d at 428; Major Mat Co. v. Monsanto Co., supra, 969 F.2d at 582-83. To move to Ohio, to forgive personal loans, to forgo other searches for possible investors, and so forth were in the nature of gambles on the part of GPI and its principals. They may have been reasonable gambles, in the sense that the prospects for a successful salvage operation were good enough that taking immediate, even if irrevocable, steps to facilitate and take advantage of the expected happy outcome was prudent. But we often reasonably rely on things that are not promises. A farmer plants his crops in the spring in reasonable reliance that spring will be followed by summer rather than by winter. There can be reasonable reliance on statements as well as on the regularities of nature, but if the statements are not reasonably understood as legally enforceable promises there can be no action for promissory estoppel.

Suppose McNamara thought that there was a 50 percent chance that the deal would go through and believed that reliance on that prospect would cost him $100,000, but also believed that by relying he could expect either to increase the likelihood that the deal would go through or to make more money if it did by being able to start production sooner and that in either event the expected benefit of reliance would exceed $100,000. Then his reliance would be reasonable even if not induced by enforceable promises. The numbers are arbitrary but the example apt. GPI and its principals relied, and may have relied reasonably, but they didn't rely on Martin's "promises" because those were not promises reasonably understood as such by so financially sophisticated a businessman as McNarama. McInerney v. Charter Golf, Inc., 680 N.E.2d 1347, 1352-53 (Ill. 1997); Gruen Industries, Inc. v. Biller, 608 F.2d 274, 281-82 (7th Cir. 1979); Clardy Mfg. Co. v. Marine Midland Business Loans Inc., 88 F.3d 347, 358, 360-61 (5th Cir. 1996). So we see now that the essence of the doctrine of promissory estoppel is not that the plaintiff have reasonably relied on the defendant's promise, but that he have reasonably relied on its being a promise in the sense of a legal commitment, and not a mere prediction or aspiration or bit of puffery.

One last point. Ordinarily the question whether a plaintiff reasonably understood a statement to be a promise is a question of fact and so cannot be resolved in summary judgment proceedings. But if it is clear that the question can be answered in only one way, there is no occasion to submit the question to a jury. See Mason Dixon Lines, Inc. v. Glover, 975 F.2d 1298, 1303-05 (7th Cir. 1992); J.C. Wyckoff Associates, Inc. v. Standard Fire Ins. Co., 936 F.2d 1474, 1493 (6th Cir. 1991). This, we believe, is such a case.

AFFIRMED.