This diversity case concerns a prolonged effort by appellee American Honda Motor Co., Inc. ("Honda") to franchise a new automobile dealership in Westborough, Massachusetts. This effort has been opposed under a Massachusetts dealer protection statute by appellant Richard Lundgren, Inc. ("Lundgren"), which operates an existing local Honda dealership in Auburn, Massachusetts, about eleven miles west of the proposed new dealership. The district court put an end to this multi-act drama, granting summary judgment to Honda. We affirm.

Massachusetts, like a number of states, protects franchised car dealers against the "potentially oppressive power" of their respective manufacturers. Am. Honda Motor Co., Inc. v. Bernardi's, Inc., 432 Mass. 425, 735 N.E.2d 348, 354 (2000) ("Honda IV"). Under the statute, the dealer has a legal right to prevent an "arbitrary" decision by the manufacturer to create a competing dealership nearby. Mass. Gen. Laws ch. 93B, § 4(3)(l) (2000) (amended 2002). But the dealer can invoke the statute's protection only if the new dealership would be located within the "relevant market area" ("RMA") of the protesting dealer. Honda IV, 735 N.E.2d at 351.

"Relevant market area" might sound like an antitrust concept. Cf. Tober Foreign Motors, Inc. v. Reiter Oldsmobile, Inc., 376 Mass. 313, 381 N.E.2d 908, 918 (1978). In fact it is defined rather mechanically by the applicable Massachusetts statute in a thus-far frustrated effort by the legislature to avoid protracted litigation over this initial "standing" requirement. Of course, the statute does not concern standing in the Article III sense; rather, the legislature has simply enacted a threshold requirement (albeit a somewhat complicated one) based on the proposed dealer's proximity before the "arbitrariness" test can be invoked.

The definition was more open-ended in the original version of the statute; there, the RMA was to be defined by the courts according to equitable principles. See generally Honda IV, 735 N.E.2d at 352 n. 8 (discussing statutory history). But, in 1977, the Massachusetts legislature adopted a new "bright line" formula which governs this case. A further amendment in 2002 sought to repair the 1977 version in the wake of judicial interpretation, 2002 Mass. Legis. Serv. ch. 222, § 3 (to be codified at Mass. Gen. Laws ch. 93B) ("2002 Amendment"); but by virtue of a savings clause, the 2002 version does not apply to the present dispute, id. § 5.

The governing 1977 definition, trimmed of portions not here pertinent, reads as follows:

[T]he relevant market area of a motor vehicle dealer ... is the ... geographical area immediately surrounding its existing dealer location within which it obtained, during [a defined period] ... at least two-thirds of ... its retail service sales....1

Mass. Gen. Laws ch. 93B, § 4(3)(l) (2000) (amended 2002).

Unfortunately, the 1977 statute left open a number of questions: for example, whether a dealership location should be regarded as a point or an area; whether the surrounding RMA should be defined as a perfect circle or may be some other shape; whether customer locations should be identified by air miles or some other variable (e.g., drive time); and whether the two-thirds figure should be based on transactions, customers or revenues. All of these issues, and others, were litigated in this case.

The present litigation began in April 1998, when Honda filed an action in the federal district court seeking a declaratory judgment. Honda planned to franchise a new dealer in Westborough, Massachusetts, about eleven miles east of the existing Lundgren dealership in Auburn. Anticipating a challenge (Lundgren had earlier contested a different proposal by Honda, Richard Lundgren, Inc. v. Am. Honda Motor Co., 45 Mass.App.Ct. 410, 699 N.E.2d 11 (1998)), Honda sought a ruling that Lundgren had no standing (under the 1977 definition) to contest this new dealership as an "arbitrary" decision by Honda.

Lundgren, as this litigation shows, is anxious to mount such a contest. Both sides engaged experts on the RMA issue. At the end of this case's first journey through the district court, Lundgren stipulated that it could not prevail if the RMA had to be a perfect circle. Am. Honda Motor Co. v. Bernardi's, Inc., 113 F. Supp. 2d 54, 57 (D.Mass.1999) ("Honda I"); Am. Honda Motor Co. v. Bernardi's, Inc., 113 F. Supp. 2d 58, 59 (D.Mass.1999) ("Honda II"). On that premise, the district court ruled in Honda's favor. Honda I, 113 F.Supp.2d at 57; Honda II, 113 F.Supp.2d at 59.

On appeal, this court certified the legal question to the Massachusetts Supreme Judicial Court ("SJC"). Am. Honda Motor Co. v. Bernardi's, Inc., 198 F.3d 293, 296-97 (1st Cir.1999) ("Honda III"). That court, over a dissent supporting the district court's ruling, held that the RMA had to be circular or like a circle but did not have to be a perfect circle. Honda IV, 735 N.E.2d at 350. This court then vacated the district court judgment and remanded for further proceedings. Am. Honda Motor Co. v. Bernardi's, Inc., 235 F.3d 1 (1st Cir.2000) ("Honda V").

The case was then transferred to a different district judge. Honda filed a new motion for summary judgment, and Lundgren offered two alternative RMA calculations purporting to establish standing under the statute. On February 8, 2002, the district court granted summary judgment, again in favor of Honda. Am. Honda Motor Co. v. Lundgren, 188 F. Supp. 2d 27, 33 (D.Mass.2002) ("Honda VI").

Two of the district court's rulings, favorable to Lundgren on the present facts, are not disputed by Honda on this appeal. One is that the two-thirds requirement could be measured by the dollar volume of sales or the number of transactions but not by the number of customers. Honda VI, 188 F.Supp.2d at 30-31. The other is that the location of customers, critical to determining the boundary within which two-thirds of those customers reside, can be determined by "geocoding." Id. at 32. Geocoding, championed by Lundgren's expert, is a somewhat more precise way of locating individual customers than is the zip code distribution method urged by Honda.2

On these premises, Lundgren's expert proceeded as follows. First, customers (each assigned the dollar value of his or her repairs in the relevant period) were sequenced in order of increasing distance from the Lundgren dealership. Then the expert measured the air mile distance ("last order distance") from the Lundgren dealership to the last customer ("last order location") necessary to encompass two-thirds of the total repair dollars. Thus, a circle drawn around the dealership with a radius equal to the last order distance would necessarily include two-thirds of the repair service revenue. Honda VI, 188 F.Supp.2d at 32.

The expert then compared the last order distance to the distance between the Lundgren dealership and the site of the proposed new dealership in Westborough ("inter-dealer distance"). In determining the inter-dealer distance, the expert did not measure the distance between the respective centroids (the geometric centers) of the two dealerships but between the eastern border of the Lundgren dealership and the western border of the proposed new dealership. Because the last order distance and the inter-dealer distance were found to be equal, Lundgren's expert concluded that the proposed new dealership is within the circle representing Lundgren's RMA. Honda VI, 188 F.Supp.2d at 32.

Lundgren's expert also made a second, alternative calculation of the RMA. This time he sequenced customers not by air mile distance from Lundgren's dealership (e.g., New York City is 200 miles from Boston), but by the time needed to drive the distance (e.g., New York City is four hours from Boston). Using this metric to derive the last order necessary to constitute two-thirds of repair sales, the expert found the drive time from Lundgren to the new dealer to be 17 minutes and the drive time to the last order location to be 17.6 minutes. According to Lundgren, using such an RMA based on drive time means that at least a small portion of the proposed new dealership would fall inside the RMA.

In its decision, the district court rejected the main air mile distance calculation on several separate grounds: (1) that Lundgren had earlier stipulated that it would lose if a perfect circle were used to construct the RMA; (2) that the RMA should be spanned from the centroid of the Lundgren property and not from its boundaries; (3) that the expert had admittedly rounded off air mile distances to the nearest tenth of a mile; and (4) that even disregarding the first three problems the expert's RMA merely touched rather than overlapped with the property of the proposed new dealer. The court rejected the alternative drive time calculation on the further ground that the Massachusetts statute did not allow it. Honda VI, 188 F.Supp.2d at 32. Summary judgment for Honda was entered and Lundgren now appeals.

On appeal from summary judgment, this court ordinarily reviews issues de novo, construing the record in the light most favorable to the non-moving party, here Lundgren. Euromotion, Inc. v. BMW of N. Am., Inc., 136 F.3d 866, 869 (1st Cir. 1998). At this stage, we assume that the computations of Lundgren's expert are accurate; whether they are foreclosed by the prior stipulation and, if not, whether they satisfy the requirements of the Massachusetts statute are legal issues.

The district court's reliance on the stipulation is understandable. The stipulation was unqualified and, if given effect, might appear to resolve the case cleanly in Honda's favor. Lundgren stipulated in the earlier go-around in the district court that if the RMA were computed as a perfect circle, it would not reach the location of the proposed new dealership. Honda I, 113 F.Supp.2d at 57. On this go-around, Lundgren's main computation is based upon a perfect circle RMA. Honda VI, 188 F.Supp.2d at 32. In litigation that has already lasted several years and occupied three different courts, Lundgren's about-face certainly needed explaining.

Stipulations between parties are not "absolute," T I Fed. Credit Union v. DelBonis, 72 F.3d 921, 928 (1st Cir.1995), and a party may be relieved of a stipulation for good cause — which means, in a nutshell, that good reason must exist and that relief must not unfairly prejudice the opposing party or the interests of justice. See, e.g., New Hampshire v. Maine, 532 U.S. 742, 750-51, 121 S. Ct. 1808, 149 L. Ed. 2d 968 (2001); FDIC v. Kooyomjian, 220 F.3d 10, 14 (1st Cir.2000). Here, Lundgren argues that its earlier concession was not based on geocoding or drive time calculations and that the intervening SJC decision in Honda IV offered room for its expert's more recent calculations.

In principle, an intervening change of law might count as good cause. See Arizona v. Shamrock Foods Co., 729 F.2d 1208, 1215-16 (9th Cir.1984), cert. denied, 469 U.S. 1197, 105 S. Ct. 980, 83 L. Ed. 2d 982 (1985). In general terms, Honda IV is permissive toward new and alternative methods of making the RMA calculation. Honda IV, 735 N.E.2d at 355. Further, geocoding has greater accuracy in its favor, Honda VI, 188 F.Supp.2d at 32, and language in Honda IV — discussed more fully below — at least leaves open the possibility of calculations based on drive time. But the change of law argument is not persuasive on these facts.

Honda IV did not undo a previous bar to geocoding and drive time calculations; from the outset, Lundgren could have employed such calculations on an alternative basis and argued that they were permissible. Instead, it placed its chips on the argument that the RMA did not have to be a perfect circle, stipulating that it would lose if a perfect circle were employed. Having won the perfect circle debate, Lundgren now ignores its victory and seeks to change other parts of its original computation in ways that it could easily have urged several years ago.

However, Honda's stipulation argument appears flawed for a different reason not argued by Lundgren. Suppose that instead of drawing a perfect circle Lundgren's expert had drawn a slightly imperfect one encompassing the same customers — say, caving slightly inward where no customer now exists. Nominally, this would avoid violating the stipulation — the circle would no longer be perfect — yet the cavity would be trivial and the resulting shape would arguably be sufficiently "circular" to comply with the dictates of Honda IV. In short, Lundgren's change in position is not aimed at altering the shape of the circle; what improves Lundgren's position is simply the use of geocoding and drive time data.

Lundgren apparently did not stipulate that it would limit itself to zip code-based platting methods and air mile distance data. Perhaps this about face as to methodology could be barred on some other ground or even claimed to be implicitly at odds with the stipulation; it is certainly bad timing since the legal validity of geocoding and drive time calculations could have been certified on the prior go-around. In sum, we would readily uphold the district court's refusal to allow a departure from the stipulation, whatever the standard of review; but we are not persuaded that Lundgren's new formula is at odds with the substance of the stipulation.

Turning to the merits, we begin with the air mile distance calculation made by Lundgren's expert, placing the boundary of Lundgren's RMA on the boundary of the proposed new dealership. The district court rejected this proposed showing on several different grounds. Honda VI, 188 F.Supp.2d at 32-33. The district court ruled that measuring distance from the boundary was unsound and that some central point within the dealership (e.g., the centroid) should be used, id. at 33; Lundgren's counsel conceded at oral argument that this would cause it to lose.

Use of some interior central point might seem the better choice; otherwise the RMA could easily depend on whether the dealer chanced to have a large or small lot. Yet the 1977 statute refers simply to the area immediately surrounding the existing dealer "location" — hardly a clear-cut preference for the centroid over the boundary.3 Policy arguments are not of great force either way. Without resolving this state law issue — which may not arise again given the change in the statute — we think it is not the strongest ground for affirmance.

Rather, so far as the air mile distance calculation is concerned, we prefer the district court's alternative ruling that an RMA merely touching on the boundary of the proposed new dealership does not create standing. Honda VI, 188 F.Supp.2d at 33. The statute's operative language, as distinct from the RMA definition itself, provides that a manufacturer may not arbitrarily franchise a new dealer "within" the RMA of an existing franchisee. Mass. Gen. Laws. ch. 93B, § 4(3)(l) (2000) (amended 2002). In our view, a new dealer whose property merely touches the outer boundary of the RMA does not create standing in the existing dealer. The statute provides that the car maker may not arbitrarily franchise a new dealer "within" the RMA of an existing franchise. A new dealer who is merely adjacent or tangent to the RMA is not fairly within the statutory language or within its evident purpose of preventing the new dealer from operating inside this perimeter.

Arguably, one could read the "within" requirement as contemplating that the new dealer be either entirely or largely within the RMA but we will assume arguendo that it is enough if some appreciable portion of the new dealer's property falls inside the RMA of the existing dealer. Even so, if the only arguable "overlap" is a single tangent point, then for all practical commercial purposes the new dealer is operating outside — not within — the existing dealer's RMA. No new sales can be made, or repairs accomplished, at this one-dimensional tangent point. Nor does Lundgren himself wish to treat either dealer as a point: it is only by looking to the full property in each case, and measuring from one near border to the other, that he even achieves a tangent point.

This makes it unnecessary to consider whether the district court was correct in rejecting the expert's rounding off of his distances. Honda VI, 188 F.Supp.2d at 33. The district court was perhaps suspicious that the unexplained rounding by Lundgren's expert unduly favored Lundgren. But virtually all measurements involve some rounding off; in this case we do not know the full scope of the rounding off adjustments, the impact, or the explanation. It is enough here that, having made his calculations, the expert could do no more than show the RMA as tangent to the proposed new dealership. This, as we and the district court both read the statute, is not sufficient.

It remains to address the expert's alternative, drive time calculation. Lundgren relies on Honda IV as endorsing the use of drive time based on a brief reference in the opinion. Explaining that even a perfect circle would not eliminate factual disputes, the SJC said that experts might dispute inter alia "whether the radius follows existing topography." Honda IV, 735 N.E.2d at 353. True, drive time accounts better for topographical oddities than air mile distance, but we agree fully with the district court that the SJC comment can hardly be taken to have approved drive time calculations.

But lack of endorsement is not the same as rejection, and Honda IV itself undercut ease of administration by holding that some unknown set of circular shapes, rather than a perfect circle, can be the basis of the RMA. Honda IV, 735 N.E.2d at 350. Further, an air mile distance calculation — although easier to compute — might occasionally depart from economic reality. Imagine, for example, that the new dealership fell within a perfect circle surrounding an existing dealership but that the new location was on the other side of an unbridged river. The new dealer would hardly be an effective competitor for repairs.

Admittedly, using drive time computations could present many problems, further undercutting the apparent intent of the 1977 amendments to create an easily administered test of standing. Honda VI, 188 F.Supp.2d at 32. Unlike air mile distance, the best driving route, and even more so the driving time, may vary with weather, season and commuter patterns; and each customer becomes a separate, debatable calculation. Simplifying assumptions can be made (at the price of accuracy), but the opportunities for expense and delay in litigating such a case are apparent. Even so, we are unwilling (without more precedent) to hold that a Massachusetts court would in every circumstance deem topography irrelevant in computing an RMA.

The potential relevance of topography, however, does not automatically validate Lundgren's theory based on drive time. Drive time calculations are in tension with the statutory requirement, as interpreted by Honda IV, that the RMA must be a "circular" area — circular being described as a figure having at least the "approximate form or outline of a circle." 735 N.E.2d at 350 & n. 4. This is the reading that the SJC, employing dictionary definitions, gave to the statutory terms "circumscribe" and "surround." Id. at 352. In reaching this conclusion, the SJC rejected an alternative approach urged by Lundgren — involving the platting of zip code areas — that would have satisfied the two-thirds requirement but produced a relatively irregular shape. Id. at 354-55.

A tension exists because the same drive time may translate to different geographic distances, depending on topography. Therefore, a shape with a radius equal to the last order drive time is not ordinarily going to be, except by accident, either a geographic circle or a circular shape; yet a geographic circular shape is what is required under the terms of the statute as construed in Honda IV, 735 N.E.2d at 352. As with the zip code platted boundary rejected in Honda IV, a shape drawn by plotting a fixed drive time in all directions could be anything from an arc to a star to the boundary of a Picasso scribble.4

In this case, there is nothing in the record to show that an RMA based on this drive time radius produces anything like a circular shape. Honda raised this objection in the district court and renews it in this court, yet Lundgren has made no effort to show that its alternative calculation meets this objection. Under these circumstances, the drive time theory is not in this case a barrier to the grant of summary judgment in Honda's favor.

Affirmed.