THARP, District Judge.

In this insurance coverage dispute, Manpower, Inc., claims various losses stemming from a building collapse that left its French subsidiary unable to access its office space for more than a year. Manpower won an early victory against the Insurance Company of the State of Pennsylvania ("ISOP") when the district court initially held that Manpower was covered under the so-called "master" policy issued by ISOP for its business interruption losses and the losses of its business personal property and improvements and betterments. But the case went south for Manpower when the district court excluded Manpower's accounting expert, without whom Manpower could not establish its business interruption damages. Manpower was dealt a second blow when the district court ruled that, although the ISOP master policy provided coverage for the losses of business personal property and improvements and betterments, that coverage was not triggered because the same losses were also covered under the subsidiary's local French policy, which had to be fully exhausted before coverage under the master policy was available. Manpower now appeals the exclusion of its accounting expert — and the attendant adverse judgment on its business interruption claim — as well as the district court's interpretation of the applicable policies with respect to its property loss claims. For the reasons that follow, we reverse the district court's exclusion of the expert, and therefore its entry of judgment against Manpower on the business interruption claim. We affirm the district court's judgment for ISOP on the property loss claim, agreeing with the district court that the master policy does not provide coverage for Manpower's property losses.

BACKGROUND

Manpower, Inc., an international recruiting and staffing firm headquartered in Milwaukee, is the parent company of Right Management, Inc., located in Paris, France. On June 15, 2006, a portion of a building located at 65 rue de la Victoire, in which Right Management leased office space, collapsed, seriously damaging part of the building as well as its underground garage. Right Management's offices and furnishings were not physically damaged, but they were rendered wholly inaccessible because the building could no longer be occupied by order of the Parisian Department of Public Safety. Eventually, Right Management relocated its business, without having gained access to its old office space and the furnishings therein; it therefore incurred replacement costs when it furnished and improved its new office space. In the meantime, Right had lost income and incurred expenses from the interruption of its business operations. This dispute arises from Manpower's claims for insurance coverage for its (Right's) various losses.

Two insurance policies govern the coverage dispute: a difference in conditions policy — the "master" policy — issued by ISOP and covering Manpower's operations worldwide, and a local policy that was issued by AIG-Europe, ISOP's French affiliate, to Right Management as the insured. In simple terms, the local policy provides the first line of coverage, and the master policy fills in the gaps by providing excess coverage over the local policy's limits, or by covering specific losses that are not insured under the local policy. Therefore, coverage under the master policy is triggered only when a given loss is not covered in full by the local policy, and the insured must exhaust coverage under the local policy before making any claim under the master policy. The local policy itself comprises "general conditions," which provide basic coverage for fire and optional coverage for a number of other events, and "special conditions," which set forth the terms of whatever optional coverage is in fact purchased by the insured. As relevant here, Right obtained coverage under the local policy's special conditions for "les dommages matériels" (a term we set off in quotation marks and leave untranslated until we address the parties' dispute over the correct translation of the term and whether it has a clear meaning even if properly translated) to or affecting insured property.

Right Management first filed a claim under the local policy and was paid $250,000 pursuant to a provision covering losses caused by a lack of access by order of a civil authority (here, the public safety agency's order prohibiting occupancy of the office building). Another $250,000 was paid out under the master policy, exhausting the $500,000 sublimit under that policy's similar lack-of-access provision.

Manpower also claimed that under the master policy it was entitled to reimbursement for its business interruption losses and the loss of — not damage to — the business personal property located in the office space Right Management could no longer access, as well as the improvements and betterments it had made to that space and had to replicate in its new offices. Manpower claimed losses of about $12 million for these items. ISOP denied the claim under the master policy for these losses, and Manpower filed suit in the district court. The court's rulings, which denied each of Manpower's claims — business interruption and property losses-are discussed below.

A. Business Interruption Loss

1. Scope of Coverage.

On early cross-motions for summary judgment on the scope of coverage, the district court addressed the threshold issue of whether the master policy provides coverage for business interruption losses. There was no dispute that the local policy does not cover such losses, so the only question was whether the master policy did so. Examining the language of the master policy, the district court noted that the policy contains a business interruption provision covering "[l]oss resulting from necessary interruption of or interference with business conducted by the Insured and caused by loss, damage, or destruction... to real or personal property as covered by this policy." The amount of coverage is $15 million. Coverage for business interruption is also provided by the lack-of-access provision, up to the sublimit of $500,000, for interruptions caused by an order of civil authorities.

According to ISOP, Manpower could not access the master policy's general business interruption provision because its losses were not caused by the building collapse, but only by the local government's closing of the building. ISOP argued that the collapse did not cause "direct physical loss... or damage to" Right Management's interest in the building, because it damaged the parking structure and another area of the building, but not Right's offices, which were left intact, though inaccessible. Therefore, ISOP argued, only the lack-of-access provision, with its $500,000 sublimit, applied to the incident.

Manpower, on the other hand, argued that the collapse damaged its interests in its rental property because its interest was not limited to the portion of the building reserved for its exclusive use, but also included common areas, elevators and staircases, safety systems, and the building's foundation and support structure, which were necessary to its use of the leased space. Manpower argued that any damage to these features that caused its operations to stall or cease triggered the coverage under the master policy's general business interruption provision. ISOP conceded that there was some damage to the common areas of the building, but maintained that Right's offices were unharmed, and therefore could have been used but for the order of the civil authorities that prohibited Right from occupying them.

In its November 3, 2009, opinion, the district court sided with Manpower. Parsing the expert testimony submitted by both sides, the district court concluded that "the evidence establishes that the collapse rendered the entire [building] unstable, at least for a period of eight to ten weeks following the collapse," during which temporary measures were taken to ensure the building's stability. Therefore, at least until temporary repairs were completed, it was the collapse itself that prevented Right from using its offices, and the order of the Department of Public Safety "merely confirmed that the collapse rendered the entire building unstable." Accordingly, the district court held that the $500,000 sublimit on losses caused by lack of access by order of civil authorities did not apply, and that "Manpower is entitled to reimbursement for any business interruption losses it sustained between the collapse and the time the necessary repairs could have been, or were, completed, up to $15 million." The business-interruption coverage further entitled Manpower to "extra expense" coverage, under the master policy provision that applies to additional expenses "reasonably and necessarily incurred resulting from loss, damage, or destruction to property by any of the perils covered herein." Under the policy, extra expenses are defined as costs of operating the business "over and above the total cost that would normally have been incurred to conduct the business during the same period had no loss or damage occurred."

2. Expert Witness Eric Sullivan

The legal issue of coverage having been decided, the parties moved on to expert discovery to determine the amount of the business interruption and extra expense losses. Manpower retained a forensic accounting expert, Eric Sullivan, who opined that the total loss for these items was 7,503,576, including an estimated business interruption loss of 5,125,830. At the close of expert discovery, ISOP filed a motion in limine to exclude Sullivan's opinion, arguing that Sullivan's opinion on the business interruption loss was not the product of a reliable methodology. There was — and is — no dispute that Sullivan was "qualified as an expert by knowledge, skill, experience, training, or education." See Fed. R.Evid. 702.

In ruling on the motion in limine, the district court recognized that the master policy itself sets forth the method by which the business interruption loss should be calculated. Section 9(B)(2) of the master policy provides that the loss equals the net profit lost because of the business interruption, adjusted for any costs and expenses paid during the interruption period that would have been paid even if no loss had occurred (dubbed "continuing expenses," as distinguished from "non-continuing expenses," which are not incurred while the operations are stopped). The policy further sets forth, in section 9(B)(5), how the various inputs should be estimated: "In determining the amount of net profit, charges, and expenses covered hereunder for the purposes of ascertaining the amount of loss sustained, due consideration shall be given to the experience of the business before the date of damage or destruction and to the probable experience thereafter had no loss occurred." The district court concluded that Sullivan had followed the master policy's prescribed methodology by calculating lost revenues minus non-continuing expenses, which is the equivalent of calculating net lost profits plus continuing expenses. Sullivan's calculations were, therefore, "straightforward." But the district court further explained that whether the calculations were reliable "turns on whether Sullivan used reliable methods when selecting the numbers used in his calculations — specifically, projected total revenues and projected total expenses."

The district court answered its question in the negative. The court was particularly troubled that Sullivan had used an estimated growth rate of 7.76% to project total revenues. Sullivan selected the projected growth rate by comparing the total revenues from January to May 2006 — the five-month period preceding the collapse — to total revenues earned in the same five-month period in the year 2005. The revenues were 7.76% higher in 2006. But ISOP argued, and the district court agreed, that Sullivan's chosen growth rate was not representative of Right Management's historical performance, which included a negative average annual growth rate of 4.79% during the period of 2003 to 2009, and a mere 3.8% growth rate in the period of January 2005 to May 2006. Sullivan had explained in his deposition that, although he reviewed economic data going back to 2003, he used a shorter period from which to extrapolate the growth rate because according to Right's managers, the recent acquisition of Right by Manpower and the enactment of new policies and installation of new managers had turned the company around by the end of 2005.

The district court concluded that Sullivan's method could be described as "basic growth-rate extrapolation," which a treatise cited by court deemed "the simplest and most frequently used revenue forecasting method." Although Sullivan used an appropriate and recognized method to calculate projected revenues, the district court concluded that Sullivan's "analysis [broke] down" at his choice of growth rate — "one of the most important parts of the business-interruption calculation." Citing Sullivan's deposition testimony, the district court concluded that Sullivan "did little more than assume that the growth that Right had experienced during the five months before the collapse was the result of new management and thus would continue unabated for the next fourteen months." The district court concluded that Sullivan was not entitled to uncritically accept the word of Right's managers: "Sullivan is not an expert on business management, and thus Sullivan's conversations with Right's managers cannot be considered a reliable basis for a revenue forecast." Moreover, Sullivan had not performed an economic analysis of the factors that affected Right's revenues to determine whether the growth in early 2006 was the result of factors other than new management. In the words of the district court, the positive growth could have been "the result of industry conditions that did not last throughout the interruption period," or simply "an aberration." Only a more thorough analysis of the reasons for the growth would have supported Sullivan's choice of a projected growth rate. In response to Manpower's argument that the choice of growth rate was a factual assumption that could be tested at trial, the district court concluded that "an ordinary trier of fact lacks the expertise necessary to judge how the actions of Right's managers impacted the company's growth." Without expert testimony on the issue, a trier of fact "could only speculate as to whether there was any causal connection between the managers' actions and the increased growth."

The court concluded that if "Sullivan [had] not chosen such a short base period for calculating lost profits," it "might have found his analysis reliable." The court explained that Sullivan had access to historical data that showed a declining trend in Right's monthly revenues, but that he instead chose to treat Right as essentially a new business, the valuation of which requires examination of other indicators to make up for the lack of a track record. Since Sullivan did not examine such indicators — such as the performance of comparable firms — his selection of the growth rate was based simply on assumptions, rather than reliable principles and methods. Concluding that the revenue projection analysis was unreliable, the district court declined to evaluate Sullivan's calculation of non-continuing expenses — the other variable in the equation of the business interruption loss.

Manpower moved for reconsideration of the decision, arguing that the district court had misunderstood Sullivan's analysis in numerous ways. It further argued that the district court misapplied Rule 702 and Daubert by taking out of the jury's hands the questions whether Sullivan reasonably relied on the testimony of Right Management executives and whether a growth rate derived from a five-month period was reasonable. In the alternative, Manpower requested that the district court accept a supplement to Sullivan's report, consisting of an affidavit in which Sullivan further explained parts of his expert report. For example, he attested that he "did more than simply select the 5-month growth from one year to the next," citing a table in his report in which he sampled longer and shorter growth periods. He said he believed that the growth rate yielded by the five-month comparison was more reliable and conservative than those derived from 12-and 14-month periods (20.1% and 13.67%, respectively). He further included a table that calculated the business interruption loss using various assumed growth rates, including negative rates of growth. ISOP opposed the motion for reconsideration on procedural and substantive grounds.

The district court did not deem the supplementation of Sullivan's report procedurally improper, despite ISOP's protestations, but it nevertheless denied the motion for reconsideration. The court first reaffirmed its earlier determination that "Sullivan's use of the before-and-after method and his decision to use basic growth-rate extrapolation was reliable," but his "decision to project lost revenues based on a five-month revenue base ... and a growth rate of 7.76" was not. Again the district court noted that "it appears that Sullivan did little more than assume that the growth that Right had experienced during the five months before the collapse was the result of new management and thus would continue unabated for the next fourteen months." Taking up Manpower's argument that Sullivan had done much more, such as conferring with Right's management and calculating various growth rates, the district court concluded that "Sullivan's conversation with Right's managers was not a reliable basis for his selection of a growth rate," and that Manpower and Sullivan had failed to show "that compiling a list of possible growth rates and then choosing a rate from this list is an acceptable way to choose a growth rate in the field of forensic accounting." Rather than relying upon his judgment and his usual method for choosing growth rates over the years of his career, Sullivan should have cited "literature from the field of forensic accounting" because "it's what's normal in the field of forensic accounting that matters." The district court reiterated that Sullivan seemed to be relying merely on his intuition rather than established principles. Accordingly, the district court denied all relief to Manpower. This decision led inexorably to the court's ruling on September 6, 2011, that ISOP was entitled to summary judgment on the business interruption claim, because without Sullivan's testimony, Manpower lacked any evidence to support the existence or the amount of a business interruption loss.

B. Business Personal Property and Improvements and Betterments

In the meantime, the parties had been proceeding apace on Manpower's other claim for coverage under the master policy: for the replacement value of Right Management's business personal property and improvements and betterments in the inaccessible office space. The district court concluded that the master policy covers the loss of the business personal property and improvements and betterments, despite the absence of physical damage to these items, which ISOP had argued was required. Citing section 10 of the master policy, the all-risk provision covering "all risk of direct physical loss of or damage to" covered property, the court concluded that "Right sustained a direct, concrete and immediate loss of its business personal property and improvements and betterments due to physical damage to its office building." However, coverage under the master policy is triggered only by a difference in conditions — that is, when there is no or incomplete coverage under the local policy for that type of loss. Manpower initially contended that the gap in coverage was established as a matter of law by virtue of the fact that AIG-Europe, the issuer of the local policy, had closed its file upon paying out $250,000 for lack of access. The district court rejected this position as an "absurd" interpretation of the local policy under which DIC coverage would be available any time the insured did not obtain a desired payment from the primary insurer, or indeed, if it simply skipped the step of filing a claim under the local policy. The only question was whether the local policy by its terms covered the loss, not whether a claim had actually been filed or paid.

The court also rejected ISOP's position that a difference in conditions could be established only after a French court had interpreted the scope of the local policy in other litigation between Right and AIG-Europe that was pending in Paris. The district court concluded — and no party now disputes — that "[w]hether the DIC [master] policy is in fact broader is a question of law that can be resolved by this Court." The district court therefore instructed Manpower, which bore the burden of proof, that it could establish that the DIC policy is broader than the local policy either by "asking that [the court] compare the two policies ..., or by establishing the extent of coverage available under the [local policy] through the litigation in France." Manpower chose the first option, moving for summary judgment on the issue of whether the master policy's coverage was broader than the local policy's coverage with respect to the claim for business personal property and improvements and betterments.

To resolve that question, the district court had to determine whether the local policy — like the master policy, see pp. 800-01, supra — provided coverage when the event causes a loss of covered property, but not physical damage to it. The parties agreed that the building collapse fell within the "all risk" provision of the local policy; the question was simply whether that provision applied only to instances in which the property was physically altered by the insurable event.

As we have noted, the local policy consists of general conditions and special conditions. Article 2 of the general conditions provides — as translated from the French — that there is coverage for "les dommages matériels,1 in other words, damages that may affect the structure or substance of the item, caused by an insured event, affecting... the buildings ... as well as all their fittings and fixtures ...; furniture and personal effects ...; equipment...; [and] merchandise." Manpower argued that there was no "physical damage" to its property and therefore no coverage under the local policy (leaving it free to collect under the master policy). But the district court rejected this interpretation of the local policy. The court analogized Manpower's loss to a theft of its property — while physically intact, the property was no longer available for its use. And, the court thought it clear that the local policy covered loss of property by theft, as most all-risk policies would. The special conditions specifically excluded from coverage losses from "thefts committed without violence or threat to the person or without felonious entry," corroborating the view that ordinary theft was covered under the "physical damage" provision. The district court also consulted a dictionary (English, not French) and noted that one definition of "damage" was "loss due to injury." Reading the policy as a whole, the court concluded that "damage" was used in a broad sense that included "loss" as well as physical damage. Therefore, as the court had already determined with respect to the master policy, the local policy covered the property loss even though the collapse did not physically alter it. Since the local policy covered the loss, the DIC or master policy's coverage was not triggered.

With its claims against ISOP effectively extinguished, Manpower asked the district court to stay the case pending the outcome of the litigation in France between Right Management and AIG-Europe. The district court declined to do so, concluding that it had ruled on every open issue in the case against ISOP and that Manpower was now bound by the court's interpretation of the policies "no matter what happens in France." The court therefore entered final judgment in favor of ISOP, and Manpower now appeals.

DISCUSSION

Manpower raises two issues on appeal: the exclusion of Eric Sullivan as an expert witness (and the accompanying summary judgment for ISOP based on the absence of evidence of a business interruption loss)2 and the interpretation of the local policy in a manner that provides coverage for the property loss, foreclosing recovery for that loss under the master policy. We begin with the issues relating to the exclusion of Manpower's damages expert on its claim for a business interruption loss.

I. Exclusion of Damages Expert

We review de novo the grant of summary judgment for ISOP on the business interruption claim, but because that decision was premised entirely on the absence of expert testimony to support the amount of the loss, from a practical standpoint the only standard of review that counts is the one that applies to the decision to exclude the expert. We review de novo whether a district court followed the proper framework for evaluating expert testimony, and if it did, we review its decision to admit or exclude expert testimony for an abuse of discretion. Schultz v. Akzo Nobel Paints, LLC, 721 F.3d 426, 431 (7th Cir.2013). In this case, the parties agree — and we concur — that the district court properly applied the principles of Federal Rule of Evidence 702 and Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993). Manpower argues, however, that the district court abused its discretion when it barred Sullivan from testifying. We agree that Sullivan's opinion — although not bulletproof — is sufficiently reliable to present to a jury and that in excluding Sullivan's opinion, the district court exercised its gatekeeping role under Daubert with too much vigor.

Rule 702, governing the admissibility of expert testimony, provides:

In short, the rule requires that the trial judge ensure that any and all expert testimony or evidence admitted "is not only relevant, but reliable." Daubert, 509 U.S. at 589, 113 S.Ct. 2786. Daubert interpreted an earlier version of Rule 702, but it remains the gold standard for evaluating the reliability of expert testimony and is essentially codified in the current version of Rule 702. See Lees v. Carthage College, 714 F.3d 516, 521 (7th Cir.2013). The principles set forth in Daubert, which addressed scientific testimony, apply equally to non-scientific fields. Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137, 147-49, 119 S.Ct. 1167, 143 L.Ed.2d 238 (1999). No matter the nature of the witness's expertise, Rule 702 "establishes a standard of evidentiary reliability," "requires a valid... connection to the pertinent inquiry as a precondition to admissibility," and mandates that the testimony have "a reliable basis in the knowledge and experience of [the relevant] discipline." Id. at 149, 119 S.Ct. 1167 (quoting Daubert, 509 U.S. at 590, 592, 113 S.Ct. 2786). See also Bielskis v. Louisville Ladder, Inc., 663 F.3d 887, 894 (7th Cir.2011) (Ultimately, the expert's opinion "must be reasoned and founded on data [and] must also utilize the methods of the relevant discipline.").

Here, the district court's resolution of the question of reliability is the primary issue in dispute. District judges have "considerable leeway in deciding in a particular case how to go about determining whether particular expert testimony is reliable." Kumho Tire Co., 526 U.S. at 152, 119 S.Ct. 1167. Reliability, however, is primarily a question of the validity of the methodology employed by an expert, not the quality of the data used in applying the methodology or the conclusions produced. "The soundness of the factual underpinnings of the expert's analysis and the correctness of the expert's conclusions based on that analysis are factual matters to be determined by the trier of fact, or, where appropriate, on summary judgment." Smith v. Ford Motor Co., 215 F.3d 713, 718 (7th Cir.2000). "Rule 702's requirement that the district judge determine that the expert used reliable methods does not ordinarily extend to the reliability of the conclusions those methods produce — that is, whether the conclusions are unimpeachable." Stollings v. Ryobi Technologies, Inc., 725 F.3d 753, 765 (7th Cir. 2013). The district court usurps the role of the jury, and therefore abuses its discretion, if it unduly scrutinizes the quality of the expert's data and conclusions rather than the reliability of the methodology the expert employed. See id. at 766; Smith, 215 F.3d at 720.

Admittedly, this is not always an easy line to draw. As the Supreme Court observed in General Electric Co. v. Joiner, 522 U.S. 136, 146, 118 S.Ct. 512, 139 L.Ed.2d 508 (1997), "conclusions and methodology are not entirely distinct from one another. Trained experts commonly extrapolate from existing data." The critical inquiry is whether there is a connection between the data employed and the opinion offered; it is the opinion connected to existing data "only by the ipse dixit of the expert," id., that is properly excluded under Rule 702.

In this case, we conclude that the concerns that prompted the district court to exclude Sullivan's opinion implicated not the reliability of Sullivan's methodology but the conclusions that it generated. Sullivan "utilize[d] the methods of the relevant discipline." Bielskis, 663 F.3d at 894. The district court recognized this several times. Sullivan consulted the policy to ascertain the methodology for calculating the business interruption loss, and the district court expressly concluded that his equation was consistent with the policy's mandate. And in selecting the first variable — Right's projected revenue if the collapse had not occurred — Sullivan utilized growth-rate extrapolation, which the district court determined was sound. Indeed, the district court sought out a scholarly treatise and cited its endorsement of this methodology. Having drawn those conclusions, the district court's assessment of the reliability of the methodology ought to have ceased (or proceeded to the second variable in the business-loss equation). Instead, the district court drilled down to a third level in order to assess the quality of the data inputs Sullivan selected in employing the growth rate extrapolation methodology. What the district court took issue with was not Sullivan's growth-rate extrapolation methodology, but rather his selection of certain data from which to extrapolate. Indeed, the district court effectively acknowledged that its problem was not with Sullivan's methodology but with his data selection when it stated that "had Sullivan not chosen such a short base period for calculating lost revenues, I might have found his analysis reliable." R.132 at 7. The district court thought Sullivan should have selected different data, covering a longer period, as the base for his projection, but the selection of data inputs to employ in a model is a question separate from the reliability of the methodology reflected in the model itself.

We recently highlighted this distinction in Stollings, where we held the district court's exclusion of expert opinion to be an abuse of discretion. There, as here, the district judge had agreed that the expert correctly employed a valid methodology but found the expert's opinion unreliable only because he concluded that one of the key data inputs he used was not sufficiently reliable. See 725 F.3d at 766. In reversing the exclusion of the expert, we noted that even though the data input in question "was undoubtedly a rough estimate," "[t]he judge should have let the jury determine how the uncertainty about [the accuracy of the data input] affected the weight of [the expert's] testimony." Id. at 767.

We made the same point in Tuf Racing Products v. Am. Suzuki Motor Corp., 223 F.3d 585, 591 (7th Cir.2000). There, we rejected a challenge to the admission of a C.P.A.'s earnings projections even though those projections were the product of "financial information furnished by Tuf and assumptions given him by counsel of the effect of the termination on Tuf's sales." Id. Notwithstanding that the data the accountant used to make his projections came from the company and the company's counsel, we concluded that the district court's admission of the evidence was permissible because the methodology the accountant had employed "was well within the competence of a C.P.A." Id.3 That the expert accountant in Tuf could opine on future earnings on the basis of information supplied by counsel should make clear that the reliability of the data itself is not the object of the Daubert inquiry. The reliability of data and assumptions used in applying a methodology is tested by the adversarial process and determined by the jury; the court's role is generally limited to assessing the reliability of the methodology — the framework — of the expert's analysis.

The latitude we afford to statisticians employing regression analysis, a proven statistical methodology used in a wide variety of contexts, further illustrates the point. Regression analysis permits the comparison between an outcome (called the dependent variable) and one or more factors (called independent variables) that may be related to that outcome. As such, the choice of independent variables to include in any regression analysis is critical to the probative value of that analysis. Nevertheless, the Supreme Court and this Circuit have confirmed on a number of occasions that the selection of the variables to include in a regression analysis is normally a question that goes to the probative weight of the analysis rather than to its admissibility. See, e.g., Bazemore v. Friday, 478 U.S. 385, 400, 106 S.Ct. 3000, 92 L.Ed.2d 315 (1986) (reversing lower court's exclusion of regression analysis based on its view that the analysis did not include proper selection of variables); Cullen v. Indiana Univ. Bd. of Trustees, 338 F.3d 693, 701-02 & n. 4 (7th Cir.2003) (citing Bazemore in rejecting challenge to expert based on omission of variables in regression analysis); In re High Fructose Corn Syrup Antitrust Litigation, 295 F.3d 651, 660-61 (7th Cir.2002) (detailing arguments of counsel about omission of variables and other flaws in application of the parties' respective regression analyses and declining to exclude analyses on that basis); Adams v. Ameritech Servs., Inc., 231 F.3d 414, 423 (7th Cir.2000) (citing Bazemore in affirming use of statistical analysis based solely on correlations — in other words, on a statistical comparison that employed no regression analysis of any independent variables at all). These precedents teach that arguments about how the selection of data inputs affect the merits of the conclusions produced by an accepted methodology should normally be left to the jury. Here, the district court expressly concluded that Sullivan developed an equation consistent with the dictates of the policy and applied an accepted methodology (growth rate extrapolation). What he took issue with was Sullivan's choice of data — variables — to employ in that methodology.

This is not to say that an expert may rely on data that has no quantitative or qualitative connection to the methodology employed. In Stollings, we explained that Rule 702's requirement that expert opinions be supported by "sufficient facts or data" means "that the expert considered sufficient data to employ the methodology," observing by way of illustration that an opinion about an average gross sales price could not be reliably supported by evidence relating to sales to only one customer "because a single observation does not provide a sufficient basis for calculating an average." 725 F.3d at 766. But Sullivan did not rely on a single data point, or on a subset of data that was plainly insufficient to support application of the growth projection methodology. Nor was the data he relied upon qualitatively inadequate. As Rule 703 reflects, an expert must employ "those kinds of facts or data" on which experts in the field would reasonably rely, so Sullivan could not have based his earnings projections on changes in the size of the white rhino population in Africa, but he did not go amiss in basing them on Manpower's actual earnings over a sustained period immediately preceding the interruption of the business and on the assessment of its managers as to its prospects going forward. As we noted in Tuf's, these are the kinds of data that accountants normally rely on in calculating future earnings.

In excluding Sullivan's business interruption opinion, the district court took the view that Sullivan simply "compiled a list of possible growth rates and then picked the one that looked about right to him." R.156 at 4. We conclude, however, that Sullivan's data selection was substantially more nuanced and principled than the district court's characterization reflects. Having obtained a comprehensive data set for three years prior, for both the Paris office and for all of Right's offices in France, Sullivan calculated growth rates over various periods. In addition, he vetted that data for anomalies and excluded some of the data as aberrant. He further reviewed documents reflecting the changes that had occurred in the company's business over that period and discussed those issues with the company's current management. He reviewed the company's projections for future earnings (prepared prior to the collapse). These factors are all patently relevant to a projection of future revenues and taking them all into account, Sullivan selected a growth rate based on the company's performance in the five months immediately preceding the building collapse and used that growth rate as the input in his calculations of losses attributable to the interruption of Manpower's business. We cannot agree with the district court that this approach reflects a data selection based "on little more than Sullivan's own intuition." That the reasoning behind that choice could be challenged as incomplete or faulty does not make it any less grounded in real data.

Sullivan's opinion regarding Manpower's business interruption loss was not, then, ipse dixit; it was "reasoned and founded on data." Bielskis, 663 F.3d at 894. There is no question that Sullivan accessed and used actual data about Right Management's past performance when he calculated a projected growth rate. This fulfilled the policy's instruction to give due consideration to the experience of the business before and after the loss was consistent with the district court's finding that the "before and after" method was appropriate. Whether Sullivan selected the best data set to use, however, is a question for the jury, not the judge. Assuming a rational connection between the data and the opinion — as there was here — an expert's reliance on faulty information is a matter to be explored on cross-examination; it does not go to admissibility. Walker v. Soo Line R.R. Co., 208 F.3d 581, 589 (7th Cir.2000). "Our system relies on cross-examination to alert the jury to the difference between good data and speculation." Schultz, 721 F.3d at 432.

ISOP's further arguments in support of Sullivan's exclusion are variations on the same theme and are unpersuasive for the same reasons. For reasons already explained, its contention that Sullivan "did not follow any accounting principles" is inconsistent not only with the record, but with the district court's express findings. And ISOP's contention that the opinion would not have been "helpful" to the trier of fact within the meaning of 702 is premised on the supposed unreliability of his methods, with the added argument that he also unduly relied upon a French-speaking associate to understand Right's expenses and other reports. The latter point is not well developed, but we find nothing remarkable about a paid expert preparing a report with the assistance of staff, especially when interpretation of foreign documents is required. Sullivan placed his assumptions in the record for scrutiny according to the adversary process; no more was required once the district court signed off on his basic methodology.

In the end, each criticism that the district court had in assessing the reliability of Sullivan's testimony was a comment on the soundness of the factual underpinnings of his calculation. Indeed, one could envision the district court's rulings as a roadmap for ISOP's cross-examination of Sullivan. But the district court supplanted that adversarial process with its admissibility determination, leaving Manpower with no competent evidence to prove its business interruption damages and ensuring summary judgment for ISOP. We conclude that the district court set the bar too high and therefore abused its discretion in barring Sullivan's testimony about the business interruption loss. Accordingly, its judgment for ISOP on the business interruption claim cannot stand. See Schultz, 721 F.3d at 434 (admissibility of expert testimony and appropriateness of summary judgment collapsed into single issue).

II. The Local Policy's Coverage of Right's Losses of Property

Manpower's next contention on appeal is that the district court wrongly interpreted the local policy to provide coverage for the replacement value of its business personal property and improvements and betterments, thereby foreclosing Manpower's claim against ISOP for coverage under the master policy. We review this issue — the equivalent of contract interpretation — de novo. Miller v. St. Paul Mercury Ins. Co., 683 F.3d 871, 874 (7th Cir. 2012).

The parties are remarkably blasé about their presentation of the applicable law on this issue, which according to the terms of the local policy is the French Insurance Code.4 Manpower did not supply this Court or the district court with the applicable French statutes or cite any French legal authority in favor of its position, nor does it cite any legal authority for its dubious contention that that French civil law does not materially differ from Wisconsin common law as it pertains to the interpretation of contracts. ISOP, for its part, cites a provision of the French Civil Code that provides that contracts should be interpreted according to "the common intent of the parties rather than the literal meaning of the terms." We have previously recognized this feature of French law, as well as its practical limitations in light of other portions of the Civil Code that limit the use of parol evidence. See Bodum USA, Inc. v. La Cafetiere, Inc., 621 F.3d 624, 629 (7th Cir.2010). But we have not been directed to any evidence in the record of the parties' intent, common or otherwise, except for the language of the contract. So we, like the district court, begin and end there. But, as the proponent of coverage, Manpower's presentation leaves much to be desired.

Manpower's primary argument on appeal is that the district court misinterpreted the local policy by construing the term "les dommages matériels" in the local policy's general conditions to include loss in addition to physical damage to property. The parties did not agree on a translation of "les dommages matériels."5 Manpower prefers the translation "physical damage" to "material damage," for obvious reasons, and in the district court it supplied a certified translation of the general conditions that yields that result. R.194-1. In response, ISOP disputed Manpower's translation and asserted that, although "les dommages matériels" has been translated as "material damage" and "physical damage," there are other possible translations, including "property damage" or "damage to property." ISOP supported these alternative definitions by an affidavit from a certified translator, who attested that in the legal context in particular, "damage to property" is the most likely interpretation. See R.

The district court noted that the parties did not agree on an English translation of the general conditions, but it decided that, for purposes of the motion, it would "assume that Manpower's translation is correct." Even crediting Manpower's translation, however, the district court concluded that the policy as a whole could not be read to limit coverage under the all-risk provision solely to events that caused physical damage to property. The court reasoned that certain types of theft were excluded from coverage, suggesting that theft — which entails loss but not damage — was otherwise covered by the policy.

Our review is de novo, but Manpower does not make a compelling case for reading the contract any differently. Its scant two pages of analysis boils down the tautology that "les dommages matériels" means "physical damage," and only physical damage, because that is what it means.6 We find particularly unpersuasive Manpower's contention that because the master policy refers to "physical loss of" property in addition to "damage," the absence of "loss of" language in the local policy must mean that losses cannot be part of "damage." Manpower offers no reason why the master policy, written in English and issued to Manpower by ISOP, and the local policy, written in French and issued to Right Management by AIG-Europe, should be read as though they were a single document in which words should be given consistent meanings throughout. It seems only natural that these two policies would have linguistic differences and that the meaning of the language used in one would not necessarily inform the meaning of the language in the other.

By contrast, one would expect terms appearing in both the general and special conditions, which together make up the unified local policy, to have the same meaning. The upshot of that sensible notion is confirmation of the district court's view that the local policy covers Manpower's losses, so the master policy does not. As previously noted, notwithstanding their inability to agree on a translation of the general conditions, the parties put an agreed translation of the special conditions in the record. It translates the phrase "les dommages matériels" to mean "material damage." Manpower's insistence that the same term elsewhere in the same policy means "physical" damage is irrational.

So the local policy, in our view, covers "material" damages. Settling on that translation does not on its face resolve the question of whether the local policy provides coverage for losses beyond those involving physical damage to property, but we conclude for another reason that it does. The local policy is an "all risks" policy; it covers "all risks" except those that are expressly excluded, and nothing in the long list of excluded events applies to what occurred here. The special conditions state that the contract "guarantees the insured goods at the address indicated against All Risks, with exclusion noted"; and the "items covered" included all items "belonging to" the insured party, with exclusions that do not apply here. Therefore, Right's business personal property and improvements and betterments located within the office space were covered property. As to whether the collapse was a "covered event," again, the policy states that the insurance "covers All Risks ... regardless of the event causing the material damage or damage to the insured item," again with exclusions that do not apply to the building collapse. Notably, the list of exclusions contains a number of specific "losses" — such as "any loss or voluntary dispossession of documents or items in a fraudulent manner" and "any unexplained disappearance, difference, or loss noted during inventory" — as well as certain thefts, namely those "committed without violence or threat to the person or without felonious entry to the areas." If the all-risks provision by its terms already excluded any event that did not cause physical damage, these exclusions would have no purpose. And as ISOP points out, under the French Civil Code, Article 1161, "[a]ll clauses in agreements are interpreted in relation to one another, giving each the meaning that results from the entire agreement."

Because the local policy covers the loss of access to Right Management's furnishings, other business personal property, and improvements and betterments, Manpower cannot access coverage for these items under the master policy unless it first exhausts its coverage under the local policy. Accordingly, we affirm the district court's grant of summary judgment in favor of ISOP on this claim.

* * *

The district court's exclusion of Eric Sullivan as an expert is VACATED and its grant of summary judgment for ISOP on the business interruption loss claim is REVERSED. Because the district court did not evaluate the full scope of Sullivan's expert testimony after concluding that it was unreliable as to the projected growth rate, it may still consider the other arguments ISOP raised in its motion in limine when determining whether to allow the testimony. The district court's grant of summary judgment for ISOP on the property claim is AFFIRMED. The case is REMANDED to the district court for proceedings consistent with this opinion.