delivered the opinion of the Court.
This is a direct appeal1 from a judgment dismissing the Government’s Section 2 (a) Clayton Act2 suit in which it sought an injunction against the selling of fluid milk products by the appellees, The Borden Company and Bowman Dairy Company, at prices which discriminate between independently owned grocery stores and grocery store chains. The District Court in an unreported decision found the pricing plan of each dairy to be a prima facie violation of § 2 (a) but concluded that these discriminatory prices were legalized by the cost justification proviso of § 2 (a), which permits price differentials as long as they “make only due allowance for differences in the cost of manufacture, sale, or delivery resulting from the differing methods or quantities in which such commodities are to such purchasers sold or delivered.” To review the Government’s contention that the District Court had improperly permitted cost justifications based on the average cost of dealing with broad groups of customers unrelated in *462cost-saving factors,3 we noted probable jurisdiction, 368 U. S. 924, and directed the parties to brief and argue the case separately as to each appellee, 368 U. S. 963. However, finding the same problem at the root of the cost justifications of each appellee, we have dealt with both in this single opinion. We have concluded that the class cost justifications submitted to the District Court by the appellees did not satisfy their burden of showing that their respective discriminatory pricing plans reflected only a “due allowance” for cost differences.
By way of background, we first point out that the present appeal is merely a glimpse of protracted litigation between the parties which began in 1951 and which has not yet seen its end. The original complaint charged violations of §§ 1 and 2 of the Sherman Act and § 2 (a) of the Clayton Act. The District Court dismissed the suit, holding that there was no proof of the alleged Sherman Act violations and that no equitable relief was necessary under the Clayton Act charge because appellees were already restrained by a consent decree entered in a private antitrust case. 111 F. Supp. 562. On direct appeal we affirmed the dismissal of the Sherman Act charges but held erroneous the refusal to grant an injunction on the Clayton Act claim solely because of the existence of the private decree. 347 U. S. 514. On remand the case was reopened and on its prima facie case the Government introduced recent general price schedules and illustrated their effect on sample stores to show that each appellee was still engaged in illegal price discrimina-*463tions notwithstanding the consent decree. In defense the appellees each introduced voluminous cost studies in justification of their pricing systems. The entire case was submitted via stipulations, depositions, and briefs. There was no dispute as to the existence of price discrimination; the sole question was whether the differences in price reflected permissible allowances for variances in cost.
In view of our disposition, we need not relate the facts in detail. Both appellees are major distributors of fluid milk products in metropolitan Chicago. The sales of both dairies to retail stores during the period in question were handled under plans which gave most of their customers — the independently owned stores — percentage discounts off list price which increased with the volume of their purchases to a specified maximum while granting a few customers — the grocery store chains — a flat discount without reference to volume and substantially greater than the maximum discount available under the volume plan offered independent stores. These discounts were made effective through schedules which appeared to cover all stores; however, the schedules were modified by private letters to the grocery chains confirming their higher discounts.4 Although the two sets of discounts were never *464officially labeled “independent” and “chain” prices, they were treated, called, and regarded as such throughout the record.
To support their defense that the disparities in price between independents and chains were attributable to differences in the cost of dealing with the two types of *465customers, the appellees introduced cost studies which will be described separately because of their differing content and analytical approach.
The Borden pricing system produced two classes of customers. The two chains, A & P and Jewel, with their combined total of 254 stores constituted one class. The 1,322 independent stores, grouped in four brackets based on the volume of their purchases, made up the other. Borden’s cost justification was built on comparisons of its average cost per $100 of sales to the chains in relation to the average cost of similar sales to each of the four groups of independents. The costs considered were personnel (including routemen, clerical and sales employees), truck expenses, and losses on bad debts and returned milk. Various methods of cost allocation were utilized: Drivers’ time spent at each store was charged directly to that store; certain clerical expenses were allocated between the two general classes; costs not susceptible of either of the foregoing were charged to the various stores on a per stop, per store, or volume basis.
Bowman’s cost justification was based on differences in volume and methods of delivery. It relied heavily upon a study of the cost per minute of its routemen’s time. It determined that substantial portions of this time were devoted to three operations, none of which were ever performed for the 163 stores operated by its two major chain customers.5 These added work steps arose from the method of collection, i. e., cash on delivery and the delayed collections connected therewith, and the performance of “optional customer services.” The customer services, performed with varying frequency depending upon the circumstances, included “services that the driver may be requested to do, such as deliver the order inside, place the containers in a refrigerator, rearrange *466containers so that any product remaining unsold from yesterday will be sold first today, leave cases of products at different spots in the store, etc.” The experts conducting the study calculated as to these elements a “standard” cost per unit of product delivered: the aggregate time required to perform the services, as determined by sample time studies,-was divided by the total number of units of product delivered. In essence, the Bowman justification was merely a comparison of the cost of these services in relation to the disparity between the chain and independent prices. Although it was shown that the five sample independents in the Government's prima facie case received the added services,6 it was not shown or found that all 2,500 independents supplied by Bowman partook of them. On the basis of its studies Bowman estimated that about two-thirds of the independent stores received the “optional customer services” on a daily basis and that “most store customers pay the driver in cash daily.”
On these facts, stated here in rather summary fashion, the trial court held that appellees had met the requirements of the proviso of § 2 (a) on the theory that the general cost differences between chain stores as a class and independents as a class justified the disparities in price reflected in appellees’ schedules. In so doing the trial court itself found “the studies . . . imperfect in *467some respects . . . It noted the “seemingly arbitrary" nature of a classification resulting “in percentage discounts which do not bear a direct ratio to differences in volume of sales.” But it found “this mode of classification is not wholly arbitrary — after all, most chain stores do purchase larger volumes of milk than do most independent stores.” 7 We believe it was erroneous for the trial court to permit cost justifications based upon such classifications.
The burden, of course, was upon the appellees to prove that the illegal price discrimination, which the Government claimed and the trial court found present, was immunized by the cost justification proviso of § 2 (a). Such is the mandate of § 2 (b) as interpreted by this Court in Federal Trade Comm’n v. Morton Salt Co., 334 U. S. 37, 44-45 (1948).8 There can be no doubt that the § 2 (a) proviso as amended by the Robinson-Patman Act contemplates, both in express wording and legislative history, a showing of actual cost differences resulting from the differing methods or quantities in which the commodities in question are sold or delivered.9 The only *468question before us is how accurate this showing must be in relation to each particular purchaser.
Although the language of the proviso, with some support in the legislative history,10 is literally susceptible of a construction which would require any discrepancy in price between any two purchasers to be individually justified, the proviso has not been so construed by those charged with its enforcement. The Government candidly recognizes in its briefs filed in the instant case that “[a]s a matter of practical necessity . . . when a seller deals with a very large number of customers, he cannot be required to establish different cost-reflecting prices for each customer.” In this same vein, the practice of grouping customers for pricing purposes has long had the approval of the Federal Trade Commission.11 We ourselves have noted the “elusiveness of cost data” in a Robinson-Patman Act proceeding. Automatic Canteen Co. v. Federal Trade Comm’n, 346 U. S. 61, 68 (1953). In short, to completely renounce class pricing as justified by class accounting would be to eliminate in practical effect the cost justification proviso as to sellers having a large number of purchasers, thereby preventing such sellers from passing on economies to their customers. It seems hardly necessary to say that such a result is at war with Congress’ language and purpose.
But this is not to say that price differentials can be justified on the basis of arbitrary classifications or even *469classifications which are representative of a numerical majority of the individual members. At some point practical considerations shade into a circumvention of the proviso. A balance is struck by the use of classes for cost justification which are composed of members of such selfsameness as to make the averaging of the cost of dealing with the group a valid and reasonable indicium of the cost of dealing with any specific group member.12 High on the list of “musts” in the use of the average cost of customer groupings under the proviso of § 2 (a) is a close resemblance of the individual members of each group on the essential point or points which determine the costs considered.
In this regard we do not find the classifications submitted by the appellees to have been shown to be of sufficient homogeneity. Certainly, the cost factors considered were not necessarily encompassed within the manner in which a customer is owned. Turning first to Borden’s justification, we note that it not only failed to show that the economies relied upon were isolated within the favored class but affirmatively revealed that members of the classes utilized were substantially unlike in the cost saving aspects considered. For instance, the favorable cost comparisons between the chains and the larger independents were for the greater part controlled by the higher average volume of the chain stores in comparison to the average volume of the 80-member class to which these independents were relegated. The District Court allowed this manner of justification because “most chain stores do purchase larger volumes of milk than do most independent stores.” However, such a grouping for cost justification purposes, composed as it is of some independents *470having volumes comparable to, and in some cases larger than, that of the chain stores, created artificial disparities between the larger independents and the chain stores. It is like averaging one horse and one rabbit. As the Federal Trade Commission said in In the Matter of Champion Spark Plug Co., 50 F. T. C. 30, 43 (1953): “A cost justification based on the difference between an estimated average cost of selling to one or two large customers and an average cost of selling to all other customers cannot be accepted as a defense to a charge of price discrimination.” This volume gap between the larger independents and the chain stores was further widened by grouping together the two chains, thereby raising the average volume of the stores of the smaller of the two chains in relation to the larger independents. Nor is the vice in the Borden class justification solely in the paper volumes relied upon, for it attributed to many independents cost factors which were not true indicia of the cost of dealing with those particular consumers. To illustrate, each independent was assigned a portion of the total expenses involved in daily cash collections, although it was not shown that all independents paid cash and in fact Borden admitted only that a “large majority” did so.
Likewise the details of Bowman’s cost study show a failure in classification. Only one additional point need be made. Its justification emphasized its costs for “optional customer service” and daily cash collection with the resulting “delay to collect.” As shown by its study these elements were crucial to Bowman’s cost justification. In the study the experts charged all independents and no chain store with these costs. Yet, it was not shown that all independents received these services daily or even on some lesser basis. Bowman’s studies indicated only that a large majority of independents took these services on a daily basis. Under such circumstances *471the use of these cost factors across the board in calculating independent store costs is not a permissible justification, for it possibly allocates costs to some independents whose mode of purchasing does not give rise to them. The burden was upon the profferer of the classification to negate this possibility, and this burden has not been met here. If these factors control the cost of dealing, then their presence or absence might with more justification be the password for admission into the various price categories.13
The appellees argue in the alternative that their cost justifications can be sufficiently unscrambled to remove any taint the Court may find in them and still show a cost gap sufficient to justify the price disparity between the chains and any independent. This mass of underlying statistical data not considered by the trial court and now tied together by untried theories can best be evaluated on remand, and we therefore do not consider its sufficiency here.
In sum, the record here shows that price discriminations have been permitted on the basis of cost differences between broad customer groupings, apparently based on the nature of ownership but in any event not shown to be so homogeneous as to permit the joining together of these purchasers for cost allocations purposes. If this is the only justification for appellees’ pricing schemes, they are illegal. We do not believe that an appropriate decree would require the trial court continuously to “pass judgment on the pricing practices of these defendants.” As to the issuance of an injunction, however, the case is now *47211 years old and we have no way of knowing whether equitable relief is in order. Certainly a relevant factor in such consideration would be whether the practices described above are still being followed in any form. This the record here does not show. Such matters can only be ascertained upon the presently existing facts and the careful application of the principles we have enunciated. For that purpose the case is
Reversed and remanded.
Mr. Justice Frankfurter took no part in the consideration or decision of this case.Jurisdiction is conferred under § 2 of the Expediting Act of February 11, 1903, 32 Stat. 823, as amended, 15 U. S. C. § 29.
“Sec. 2. (a) That it shall be unlawful for any person engaged in commerce, in the course of such commerce, ... to discriminate in price between different purchasers of commodities of like grade and quality, where either or any of the purchases involved in such discrimination are in commerce, . . . and where the effect of such discrimination may be substantially to lessen competition or tend to create a monopoly in any line of commerce, or to injure, destroy, or prevent competition with any person who either grants or knowingly receives the benefit of such discrimination, or with customers of either of them: Provided, That nothing herein contained shall prevent differentials which make only due allowance for differences in the cost of manufacture, sale, or delivery resulting from the differing methods or quantities in which such commodities are to such purchasers sold or delivered . . . .” 38 Stat. 730, as amended, 49 Stat. 1526, 15 U. S. C. §13 (a).
Bowman's contention that the Government by stipulation limited itself to specific objections which do not include the present one is without foundation in the record. At the time the stipulation was proposed, the trial court made it quite clear that the Government by so stipulating was not waiving its right to argue the legal sufficiency of the proffered cost studies.
Borden in June of 1954 issued the following discount schedule to “be applied to all purchases of Borden’s fresh milk”:
Percent of Average converted units per day: discounts
0-24. 0
25- 74 . 2
75-149 . 3
150 and over. 4
At this same time, letters were sent to The Great Atlantic and Pacific Tea Compaq'’ and The Jewel Food Stores granting them flat 8%% discounts. A few of the larger independents by special arrangement were given an additional 1%% discount, thereby raising their total discount to 5%%•
*464In September 1955, Borden discontinued the above discount system and utilized a net price scheme which resulted in even greater disparities between chains and independents.
Bowman in June of 1954 operated under the following “Resale Store Discount Schedule”: _ .
, Percent of Average converted points per day: discounts
0 to 10. 3.0 to 3.4
10 to 20. 3.4 to 3.8
20 to 30. 3.8 to 4.2
30 to 40. 4.2 to 4.6
40 to 50. 4.6 to 5.0
50 to 60. 5.0 to 5.2
60 to 70. 5.2 to 5.4
70 to 80. 5.4 to 5.6
80 to 90. 5.6 to 5.8
90 to 100. 5.8 to 6.0
100 to 110. 6.0 to 6.2
110 to 120. 6.2 to 6.4
120 to 130. 6.4 to 6.6
130 to 140. 6.6 to 6.8
140 to 150. 6.8 to 7.0
This schedule was modified in August by the addition of the following discounts: _ , ,
, Percent of Average converted points per day: discounts
150 to 200. 7.0 to 8.0
Over 200. 8.0
During this same period Bowman by letter granted The Great Atlantic and Pacific Tea Company and The Kroger Company flat 11% discounts. Goldblatt Bros., also a multi-store operation, was granted a flat 8%%-
In 1955 and again in 1956 Bowman modified the brackets and percentages of its discount schedules, but not in a manner which reduced the disparity between independents and chains.
The third chain, Goldblatt Bros., also did not take these services.
The contention is made that the Government limited its prima facie case to a few stores on some routes and that therefore cost justification was only necessary as to them. This overlooks the fact that sampling has long been a recognized technique in price discrimination cases and that this offering was in support of the Government’s position, found valid by the trial court, that the entire Chicago pricing scheme of each appellee, as evidenced by its published price lists, was in violation of §2 (a). In addition, appellee's cost justifications were not limited to the Government’s sample stores.
Even the trial court was unwilling to give its “stamp of approval to all pricing policies and practices revealed by the evidence.” But it concluded that to enjoin such practices would lead to regulation and would require the court continually “to pass judgment on the pricing practices of these defendants,” a matter which might better be handled by proceedings before the Federal Trade Commission.
Sec. 2 (b). “Upon proof being made, at any hearing on a complaint under this section, that there has been discrimination in price or services or facilities furnished, the burden of rebutting the prima-facie case thus made by showing justification shall be upon the person charged with a violation of this section 49 Stat. 1526, 15 U. S. C. §13 (b).
For a collection and discussion of the pertinent legislative history as well as the cases and treatises on the § 2 (a) proviso, see Rowe, Price Discrimination Under the Robinson-Patman Act, c. 10 (1962).-
For instance, the Chairman of the Conference on the Bill reported to the House: “The differential granted a particular customer must be traceable to some difference between him and other particular customers, either in the quantities purchased by them or in the methods by which they are purchased or their delivery taken.” 80 Cong. Rec. 9417 (1936).
For a discussion of the Commission’s position in this regard, see Rowe, op. cit., supra, note 9, § 10.6.
Advisory Committee on Cost Justification, Report to the United States Federal Trade Commission (1956), p. 8.
Another suspect feature is that classifications based on services received by independents were apparently frozen — making it impossible for them to obtain larger discounts by electing not to receive the cost-determinative services — with no justifiable business reason offered in support of the practice.