POSNER, Circuit Judge.

This diversity suit between Lake River Corporation and Carborundum Company requires us to consider questions of Illinois commercial law, and in particular to explore the fuzzy line between penalty clauses and liquidated-damages clauses.

Carborundum manufactures "Ferro Carbo," an abrasive power used in making steel. To serve its midwestern customers better, Carborundum made a contract with Lake River by which the latter agreed to provide distribution services in its warehouse in Illinois. Lake River would receive Ferro Carbo in bulk from Carborundum, "bag" it, and ship the bagged product to Carborundum's customers. The Ferro Carbo would remain Carborundum's property until delivered to the customers.

Carborundum insisted that Lake River install a new bagging system to handle the contract. In order to be sure of being able to recover the cost of the new system ($89,000) and make a profit of 20 percent of the contract price, Lake River insisted on the following minimum-quantity guarantee:

In consideration of the special equipment [i.e., the new bagging system] to be acquired and furnished by LAKE-RIVER for handling the product, CARBORUNDUM shall, during the initial three-year term of this Agreement, ship to LAKE-RIVER for bagging a minimum quantity of [22,500 tons]. If, at the end of the three-year term, this minimum quantity shall not have been shipped, LAKE-RIVER shall invoice CARBORUNDUM at the then prevailing rates for the difference between the quantity bagged and the minimum guaranteed.

If Carborundum had shipped the full minimum quantity that it guaranteed, it would have owed Lake River roughly $533,000 under the contract.

After the contract was signed in 1979, the demand for domestic steel, and with it the demand for Ferro Carbo, plummeted, and Carborundum failed to ship the guaranteed amount. When the contract expired late in 1982, Carborundum had shipped only 12,000 of the 22,500 tons it had guaranteed. Lake River had bagged the 12,000 tons and had billed Carborundum for this bagging, and Carborundum had paid, but by virtue of the formula in the minimum-guarantee clause Carborundum still owed Lake River $241,000 — the contract price of $533,000 if the full amount of Ferro Carbo had been shipped, minus what Carborundum had paid for the bagging of the quantity it had shipped.

When Lake River demanded payment of this amount, Carborundum refused, on the ground that the formula imposed a penalty. At the time, Lake River had in its warehouse 500 tons of bagged Ferro Carbo, having a market value of $269,000, which it refused to release unless Carborundum paid the $241,000 due under the formula. Lake River did offer to sell the bagged product and place the proceeds in escrow until its dispute with Carborundum over the enforceability of the formula was resolved, but Carborundum rejected the offer and trucked in bagged Ferro Carbo from the East to serve its customers in Illinois, at an additional cost of $31,000.

Lake River brought this suit for $241,000, which it claims as liquidated damages. Carborundum counterclaimed for the value of the bagged Ferro Carbo when Lake River impounded it and the additional cost of serving the customers affected by the impounding. The theory of the counterclaim is that the impounding was a conversion, and not as Lake River contends the assertion of a lien. The district judge, after a bench trial, gave judgment for both parties. Carborundum ended up roughly $42,000 to the good: $269,000 + $31,000-$241,000-$17,000, the last figure representing prejudgment interest on Lake River's damages. (We have rounded off all dollar figures to the nearest thousand.) Both parties have appealed.

The only issue that is not one of damages is whether Lake River had a valid lien on the bagged Ferro Carbo that it refused to ship to Carborundum's customers — that, indeed, it holds in its warehouse to this day. Although Ferro Carbo does not deteriorate with age, the domestic steel industry remains in the doldrums and the product is worth less than it was in 1982 when Lake River first withheld it. If Lake River did not have a valid lien on the product, then it converted it, and must pay Carborundum the $269,000 that the Ferro Carbo was worth back then.

It might seem that if the minimum-guarantee clause was a penalty clause and hence unenforceable, the lien could not be valid, and therefore that we should discuss the penalty issue first. But this is not correct. If the contractual specification of damages is invalid, Lake River still is entitled to any actual damages caused by Carborundum's breach of contract in failing to deliver the minimum amount of Ferro Carbo called for by the contract. The issue is whether an entitlement to damages, large or small, entitles the victim of the breach to assert a lien on goods that are in its possession though they belong to the other party.

Lake River has not been very specific about the type of lien it asserts. We think it best described as a form of artisan's lien, the "lien of the bailee, who does work upon or adds materials to chattels. . . ." Restatement of Security § 61, comment on clause (a), at p. 165 (1941). Lake River was the bailee of the Ferro Carbo that Carborundum delivered to it, and it did work on the Ferro Carbo — bagging it, and also storing it (storage is a service, too). If Carborundum had refused to pay for the services that Lake River performed on the Ferro Carbo delivered to it, then Lake River would have had a lien on the Ferro Carbo in its possession, to coerce payment. Cf. National Bank of Joliet v. Bergeron Cadillac, Inc., 66 Ill.2d 140, 143-44, 5 Ill.Dec. 588, 589, 361 N.E.2d 1116, 1117 (1977). But in fact, when Lake River impounded the bagged Ferro Carbo, Carborundum had paid in full for all bagging and storage services that Lake River had performed on Ferro Carbo shipped to it by Carborundum. The purpose of impounding was to put pressure on Carborundum to pay for services not performed, Carborundum having failed to ship the Ferro Carbo on which those services would have been performed.

Unlike a contractor who, having done the work contracted for without having been paid, may find himself in a box, owing his employees or suppliers money he does not have — money he was counting on from his customer — Lake River was the victim of a breach of a portion of the contract that remained entirely unexecuted on either side. Carborundum had not shipped the other 10,500 tons, as promised; but on the other hand Lake River had not had to bag those 10,500 tons, as it had promised. It is not as if Lake River had bagged those tons, incurring heavy costs that it expected to recoup from Carborundum, and then Carborundum had said, "Sorry, we won't pay you; go ahead and sue us."

A lien is strong medicine; it clogs up markets, as the facts of this case show. Its purpose is to provide an effective self-help remedy for one who has done work in expectation of payment and then is not paid. The vulnerable position of such a person gives rise to "the artisan's privilege of holding the balance for work done in the past." United States v. Toys of the World Club, Inc., 288 F.2d 89, 94 (2d Cir. 1961) (Friendly, J.) (emphasis added). A lien is thus a device for preventing unjust enrichment — not for forcing the other party to accede to your view of a contract dispute. "The right to retain possession of the property to enforce a possessory lien continues until such time as the charges for such materials, labor and services are paid." Bull v. Mitchell, 114 Ill.App.3d 177, 181, 70 Ill.Dec. 138, 141, 448 N.E.2d 1016, 1019 (1983); cf. Ill.Rev.Stat. ch. 82, § 40. Since here the charges were paid before the lien was asserted, the lien was no good.

Lake River tries to compare its position to that of a conventional lien creditor by pointing out that it made itself particularly vulnerable to a breach of contract by buying specialized equipment at Carborundum's insistence, to the tune of $89,000, before performance under the contract began. It says it insisted on the minimum guarantee in order to be sure of being able to amortize this equipment over a large enough output of bagging services to make the investment worthwhile. But the equipment was not completely useless for other contracts — Lake River having in fact used it for another contract; it was not the major cost of fulfilling the contract; and Lake River received almost $300,000 during the term of the contract, thus enabling it to amortize much of the cost of the special equipment. Although Lake River may have lost money on the contract (but as yet there is no proof it did), it was not in the necessitous position of a contractor who completes his performance without receiving a dime and then is told by his customer to sue for the price. The recognition of a lien in such a case is based on policies akin to those behind the rule that a contract modification procured by duress will not be enforced. See, e.g., Selmer Co. v. Blakeslee-Midwest Co., 704 F.2d 924 (7th Cir. 1983). When as a practical matter the legal remedy may be inadequate because it operates too slowly, self-help is allowed. But we can find no case recognizing a lien on facts like these, no ground for thinking that the Illinois Supreme Court would be the first court to recognize such a lien if this case were presented to it, and no reason to believe that the recognition of such a lien would be a good thing. It would impede the marketability of goods without responding to any urgent need of creditors.

Conrow v. Little, 115 N.Y. 387, 393, 22 N.E. 346, 347 (1889), on which Lake River relies heavily because the lien allowed in that case extended to "money expended in the preparation of instrumentalities," is not in point. The plaintiffs, dealers in paper, had made extensive deliveries to the defendants for which they had received no payment. See id. at 390-91, 22 N.E. at 346. If Lake River had bagged several thousand tons of Ferro Carbo without being paid anything, it would have had a lien on the Ferro Carbo; and maybe — if Conrow is good law in Illinois, a question we need not try to answer — the lien would have included not only the contract price for the Ferro Carbo that Lake River had bagged but also the unreimbursed, unsalvageable cost of the special bagging system that Lake River had installed. But that is not this case. Carborundum was fully paid up and Lake River has made no effort to show how much if any money it stood to lose because the bagging system was not fully amortized. The only purpose of the lien was to collect damages which would have been unrelated to — and certainly exceeded — the investment in the bagging system.

It is no answer that the bagging system should be presumed to have been amortized equally over the life of the contract, and therefore to have been only half amortized when Carborundum broke the contract. Amortization is an accounting device; it need not reflect cash flows. There is no evidence that when the contract was broken, Lake River was out of pocket a cent in respect of the bagging system, especially when we consider that the bagging system was still usable, and was used to fulfill another contract.

The hardest issue in the case is whether the formula in the minimum-guarantee clause imposes a penalty for breach of contract or is merely an effort to liquidate damages. Deep as the hostility to penalty clauses runs in the common law, see Loyd, Penalties and Forfeitures, 29 Harv.L.Rev. 117 (1915), we still might be inclined to question, if we thought ourselves free to do so, whether a modern court should refuse to enforce a penalty clause where the signator is a substantial corporation, well able to avoid improvident commitments. Penalty clauses provide an earnest of performance. The clause here enhanced Carborundum's credibility in promising to ship the minimum amount guaranteed by showing that it was willing to pay the full contract price even if it failed to ship anything. On the other side it can be pointed out that by raising the cost of a breach of contract to the contract breaker, a penalty clause increases the risk to his other creditors; increases (what is the same thing and more, because bankruptcy imposes "deadweight" social costs) the risk of bankruptcy; and could amplify the business cycle by increasing the number of bankruptcies in bad times, which is when contracts are most likely to be broken. But since little effort is made to prevent businessmen from assuming risks, these reasons are no better than makeweights.

A better argument is that a penalty clause may discourage efficient as well as inefficient breaches of contract. Suppose a breach would cost the promisee $12,000 in actual damages but would yield the promisor $20,000 in additional profits. Then there would be a net social gain from breach. After being fully compensated for his loss the promisee would be no worse off than if the contract had been performed, while the promisor would be better off by $8,000. But now suppose the contract contains a penalty clause under which the promisor if he breaks his promise must pay the promisee $25,000. The promisor will be discouraged from breaking the contract, since $25,000, the penalty, is greater than $20,000, the profits of the breach; and a transaction that would have increased value will be forgone.

On this view, since compensatory damages should be sufficient to deter inefficient breaches (that is, breaches that cost the victim more than the gain to the contract breaker), penal damages could have no effect other than to deter some efficient breaches. But this overlooks the earlier point that the willingness to agree to a penalty clause is a way of making the promisor and his promise credible and may therefore be essential to inducing some value-maximizing contracts to be made. It also overlooks the more important point that the parties (always assuming they are fully competent) will, in deciding whether to include a penalty clause in their contract, weigh the gains against the costs — costs that include the possibility of discouraging an efficient breach somewhere down the road — and will include the clause only if the benefits exceed those costs as well as all other costs.

On this view the refusal to enforce penalty clauses is (at best) paternalistic — and it seems odd that courts should display parental solicitude for large corporations. But however this may be, we must be on guard to avoid importing our own ideas of sound public policy into an area where our proper judicial role is more than usually deferential. The responsibility for making innovations in the common law of Illinois rests with the courts of Illinois, and not with the federal courts in Illinois. And like every other state, Illinois, untroubled by academic skepticism of the wisdom of refusing to enforce penalty clauses against sophisticated promisors, see, e.g., Goetz Scott, Liquidated Damages, Penalties and the Just Compensation Principle, 77 Colum. L.Rev. 554 (1977), continues steadfastly to insist on the distinction between penalties and liquidated damages. See, e.g., Bauer v. Sawyer, 8 Ill.2d 351, 359-61, 134 N.E.2d 329, 333-34 (1956); Stride v. 120 West Madison Bldg. Corp., 132 Ill.App.3d 601, 605-06, 87 Ill.Dec. 790, 793, 477 N.E.2d 1318, 1321 (1985); Builder's Concrete Co. v. Fred Faubel Sons, Inc., 58 Ill.App.3d 100, 107, 15 Ill.Dec. 517, 524, 373 N.E.2d 863, 869 (1978). To be valid under Illinois law a liquidation of damages must be a reasonable estimate at the time of contracting of the likely damages from breach, and the need for estimation at that time must be shown by reference to the likely difficulty of measuring the actual damages from a breach of contract after the breach occurs. If damages would be easy to determine then, or if the estimate greatly exceeds a reasonable upper estimate of what the damages are likely to be, it is a penalty. See, e.g., M.I.G. Investments, Inc. v. Marsala, 92 Ill. App.3d 400, 405-06, 47 Ill.Dec. 265, 270, 414 N.E.2d 1381, 1386 (1981).

The distinction between a penalty and liquidated damages is not an easy one to draw in practice but we are required to draw it and can give only limited weight to the district court's determination. Whether a provision for damages is a penalty clause or a liquidated-damages clause is a question of law rather than fact, Weiss v. United States Fidelity Guaranty Co., 300 Ill. 11, 16, 132 N.E. 749, 751 (1921); M.I.G. Investments, Inc. v. Marsala, supra, 92 Ill.App.3d 400, 406, 47 Ill.Dec. 265, 270, 414 N.E.2d 1381, 1386, and unlike some courts of appeals we do not treat a determination by a federal district judge of an issue of state law as if it were a finding of fact, and reverse only if persuaded that clear error has occurred, though we give his determination respectful consideration. See, e.g., Morin Bldg. Products Co. v. Baystone Construction, Inc., 717 F.2d 413, 416-17 (7th Cir. 1983); In re Air Crash Disaster Near Chicago, 701 F.2d 1189, 1195 (7th Cir. 1983); 19 Wright, Miller Cooper, Federal Practice and Procedure § 4507, at pp. 106-110 (1982).

Mindful that Illinois courts resolve doubtful cases in favor of classification as a penalty, see, e.g., Stride v. 120 West Madison Bldg. Corp., supra, 132 Ill. App.3d at 605, 87 Ill.Dec. at 793, 477 N.E.2d at 1321; Pick Fisheries, Inc. v. Burns Electronic Security Services, Inc., 35 Ill.App.3d 467, 472, 342 N.E.2d 105, 108 (1976), we conclude that the damage formula in this case is a penalty and not a liquidation of damages, because it is designed always to assure Lake River more than its actual damages. The formula — full contract price minus the amount already invoiced to Carborundum — is invariant to the gravity of the breach. When a contract specifies a single sum in damages for any and all breaches even though it is apparent that all are not of the same gravity, the specification is not a reasonable effort to estimate damages; and when in addition the fixed sum greatly exceeds the actual damages likely to be inflicted by a minor breach, its character as a penalty becomes unmistakable. See M.I.G. Investments, Inc. v. Marsala, supra, 92 Ill.App.3d at 405-06, 47 Ill.Dec. at 270, 414 N.E.2d at 1386; cf. Arduini v. Board of Educ., 93 Ill.App.3d 925, 931-33, 49 Ill.Dec. 460, 465-66, 418 N.E.2d 104, 109-10 (1981), rev'd on other grounds, 92 Ill.2d 197, 65 Ill.Dec. 281, 441 N.E.2d 73 (1982); 5 Corbin on Contracts § 1066 (1964). This case is within the gravitational field of these principles even though the minimum-guarantee clause does not fix a single sum as damages.

Suppose to begin with that the breach occurs the day after Lake River buys its new bagging system for $89,000 and before Carborundum ships any Ferro Carbo. Carborundum would owe Lake River $533,000. Since Lake River would have incurred at that point a total cost of only $89,000, its net gain from the breach would be $444,000. This is more than four times the profit of $107,000 (20 percent of the contract price of $533,000) that Lake River expected to make from the contract if it had been performed: a huge windfall.

Next suppose (as actually happened here) that breach occurs when 55 percent of the Ferro Carbo has been shipped. Lake River would already have received $293,000 from Carborundum. To see what its costs then would have been (as estimated at the time of contracting), first subtract Lake River's anticipated profit on the contract of $107,000 from the total contract price of $533,000. The difference — Lake River's total cost of performance — is $426,000. Of this, $89,000 is the cost of the new bagging system, a fixed cost. The rest ($426,000-$89,000 = $337,000) presumably consists of variable costs that are roughly proportional to the amount of Ferro Carbo bagged; there is no indication of any other fixed costs. Assume, therefore, that if Lake River bagged 55 percent of the contractually agreed quantity, it incurred in doing so 55 percent of its variable costs, or $185,000. When this is added to the cost of the new bagging system, assumed for the moment to be worthless except in connection with the contract, the total cost of performance to Lake River is $274,000. Hence a breach that occurred after 55 percent of contractual performance was complete would be expected to yield Lake River a modest profit of $19,000 ($293,000-$274,000). But now add the "liquidated damages" of $241,000 that Lake River claims, and the result is a total gain from the breach of $260,000, which is almost two and a half times the profit that Lake River expected to gain if there was no breach. And this ignores any use value or salvage value of the new bagging system, which is the property of Lake River — though admittedly it also ignores the time value of money; Lake River paid $89,000 for that system before receiving any revenue from the contract.

To complete the picture, assume that the breach had not occurred till performance was 90 percent complete. Then the "liquidated damages" clause would not be so one-sided, but it would be one-sided. Carborundum would have paid $480,000 for bagging. Against this, Lake River would have incurred its fixed cost of $89,000 plus 90 percent of its variable costs of $337,000, or $303,000. Its total costs would thus be $392,000, and its net profit $88,000. But on top of this it would be entitled to "liquidated damages" of $53,000, for a total profit of $141,000 — more than 30 percent more than its expected profit of $107,000 if there was no breach.

The reason for these results is that most of the costs to Lake River of performing the contract are saved if the contract is broken, and this saving is not reflected in the damage formula. As a result, at whatever point in the life of the contract a breach occurs, the damage formula gives Lake River more than its lost profits from the breach — dramatically more if the breach occurs at the beginning of the contract; tapering off at the end, it is true. Still, over the interval between the beginning of Lake River's performance and nearly the end, the clause could be expected to generate profits ranging from 400 percent of the expected contract profits to 130 percent of those profits. And this is on the assumption that the bagging system has no value apart from the contract. If it were worth only $20,000 to Lake River, the range would be 434 percent to 150 percent.

Lake River argues that it would never get as much as the formula suggests, because it would be required to mitigate its damages. This is a dubious argument on several grounds. First, mitigation of damages is a doctrine of the law of court-assessed damages, while the point of a liquidated-damages, clause is to substitute party assessment; and that point is blunted, and the certainty that liquidated-damages clauses are designed to give the process of assessing damages impaired, if a defendant can force the plaintiff to take less than the damages specified in the clause, on the ground that the plaintiff could have avoided some of them. It would seem therefore that the clause in this case should be read to eliminate any duty of mitigation, that what Lake River is doing is attempting to rewrite the clause to make it more reasonable, and that since actually the clause is designed to give Lake River the full damages it would incur from breach (and more) even if it made no effort to find a substitute use for the equipment that it bought to perform the contract, this is just one more piece of evidence that it is a penalty clause rather than a liquidated-damages clause. See Northwest Collectors, Inc. v. Enders, 74 Wn.2d 585, 594, 446 P.2d 200, 206 (1968).

But in any event mitigation would not mitigate the penal character of this clause. If Carborundum did not ship the guaranteed minimum quantity, the reason was likely to be — the reason was — that the steel industry had fallen on hard times and the demand for Ferro Carbo was therefore down. In these circumstances Lake River would have little prospect of finding a substitute contract that would yield it significant profits to set off against the full contract price, which is the method by which it proposes to take account of mitigation. At argument Lake River suggested that it might at least have been able to sell the new bagging equipment to someone for something, and the figure $40,000 was proposed. If the breach occurred on the first day when performance under the contract was due and Lake River promptly sold the bagging equipment for $40,000, its liquidated damages would fall to $493,000. But by the same token its costs would fall to $49,000. Its profit would still be $444,000, which as we said was more than 400 percent of its expected profit on the contract. The penal component would be unaffected.

With the penalty clause in this case compare the liquidated-damages clause in Arduini v. Board of Education, supra, which is representative of such clauses upheld in Illinois. The plaintiff was a public school teacher whose contract provided that if he resigned before the end of the school year he would be docked 4 percent of his salary. This was a modest fraction of the contract price. And the cost to the school of an untimely resignation would be difficult to measure. Since that cost would be greater the more senior and experienced the teacher was, the fact that the liquidated damages would be greater the higher the teacher's salary did not make the clause arbitrary. Even the fact that the liquidated damages were the same whether the teacher resigned at the beginning, the middle, or the end of the school year was not arbitrary, for it was unclear how the amount of actual damages would vary with the time of resignation. Although one might think that the earlier the teacher resigned the greater the damage to the school would be, the school might find it easier to hire a replacement for the whole year or a great part of it than to bring in a replacement at the last minute to grade the exams left behind by the resigning teacher. Here, in contrast, it is apparent from the face of the contract that the damages provided for by the "liquidated damages" clause are grossly disproportionate to any probable loss and penalize some breaches much more heavily than others regardless of relative cost.

We do not mean by this discussion to cast a cloud of doubt over the "take or pay" clauses that are a common feature of contracts between natural gas pipeline companies and their customers. Such clauses require the customer, in consideration of the pipeline's extending its line to his premises, to take a certain amount of gas at a specified price — and if he fails to take it to pay the full price anyway. The resemblance to the minimum-guarantee clause in the present case is obvious, but perhaps quite superficial. Neither party has mentioned take-or-pay clauses, and we can find no case where such a clause was even challenged as a penalty clause — though in one case it was argued that such a clause made the damages unreasonably low. See National Fuel Gas Distribution Corp. v. Pennsylvania Public Utility Comm'n, 76 Pa. Commw. 102, 126-27 n. 8, 464 A.2d 546, 558 n. 8 (1983). If, as appears not to be the case here but would often be the case in supplying natural gas, a supplier's fixed costs were a very large fraction of his total costs, a take-or-pay clause might well be a reasonable liquidation of damages. In the limit, if all the supplier's costs were incurred before he began supplying the customer, the contract revenues would be an excellent measure of the damages from breach. But in this case, the supplier (Lake River, viewed as a supplier of bagging services to Carborundum) incurred only a fraction of its costs before performance began, and the interruption of performance generated a considerable cost saving that is not reflected in the damage formula.

The fact that the damage formula is invalid does not deprive Lake River of a remedy. The parties did not contract explicitly with reference to the measure of damages if the agreed-on damage formula was invalidated, but all this means is that the victim of the breach is entitled to his common law damages. See, e.g., Restatement, Second, Contracts § 356, comment a (1981). In this case that would be the unpaid contract price of $241,000 minus the costs that Lake River saved by not having to complete the contract (the variable costs on the other 45 percent of the Ferro Carbo that it never had to bag). The case must be remanded to the district judge to fix these damages.

Two damage issues remain. The first concerns Carborundum's expenses of delivering bagged Ferro Carbo to its customers to replace that impounded by Lake River. The district judge gave Carborundum the full market value of the bagged Ferro Carbo. Lake River argues that it should not have to pay for Carborundum's expense of selling additional Ferro Carbo — additional in the sense that Carborundum is being given credit for the full retail value of the product that Lake River withheld. To explain, suppose that Carborundum had an order for $1,000 worth of bagged Ferro Carbo, which Lake River was supposed to deliver; and because it refused, Carborundum incurred a transportation cost of $100 to make a substitute shipment of bagged Ferro Carbo to the customer. Carborundum would still get $1,000 from the customer, and if that price covered the transportation cost it would still make a profit. In what sense, therefore, is that cost a separate item of damage, of loss? On all Ferro Carbo (related to this case) sold by Carborundum in the Midwest, Carborundum received the full market price, either from its customers in the case of Ferro Carbo actually delivered to them, or from Lake River in the case of the Ferro Carbo that Lake River refused to deliver. Having received a price designed to cover all expenses of sale, a seller cannot also get an additional damage award for any of those expenses.

If, however, the additional Ferro Carbo that Carborundum delivered to its midwestern customers in substitution for Ferro Carbo previously delivered to, and impounded by, Lake River would have been sold in the East at the same price but lower cost, Carborundum would have had an additional loss, in the form of reduced profits, for which it could recover additional damages. But it made no effort to prove such a loss. Maybe it had no unsatisfied eastern customers, and expanded rather than shifted output to fulfill its midwestern customers' demand. The damages on the counter-claim must be refigured also.

Finally, Lake River argues that Carborundum failed to mitigate its damages by accepting Lake River's offer to deliver the bagged product and place the proceeds in escrow. But a converter is not entitled to retain the proceeds of the conversion even temporarily. Lake River had an opportunity to limit its exposure by selling the bagged product on Carborundum's account and deducting what it claimed was due it on its "lien." Its failure to follow this course reinforces our conclusion that the assertion of the lien was a naked attempt to hold Carborundum hostage to Lake River's view — an erroneous view, as it has turned out — of the enforceability of the damage formula in the contract.

The judgment of the district court is affirmed in part and reversed in part, and the case is returned to that court to redetermine both parties' damages in accordance with the principles in this opinion. The parties may present additional evidence on remand, and shall bear their own costs in this court. Circuit Rule 18 shall not apply on remand.

AFFIRMED IN PART, REVERSED IN PART, AND REMANDED.