The Massachusetts Institute of Technology (MIT) appealed to a single justice, pursuant to G. L. c. 25, § 5, from an order of the Department of Public Utilities (department) that authorizes the Cambridge Electric Light Company (company) to impose a monthly customer transition charge (CTC) on MIT, following MIT’s construction of its own cogeneration facility and its departure as a full-service customer from the company.2 The single justice reserved and reported the case to the full court.3
The CTC was authorized by the department to permit *858recovery of the company’s so-called “stranded costs”4 from MIT because of MIT’s size in relation to the company, and its decision to self-generate.5 According to MIT, the CTC authorized by the department in this case is the first such tariff imposed on a cogeneration facility anywhere in the country; the CTC would require MIT to pay an additional monthly charge exceeding $110,000, resulting in an annual charge of $1.3 million to MIT.6
On appeal, MIT argues that in approving the CTC, the department failed to apply its own PURPA regulations, 220 Code Mass. Regs. §§ 8.00-8.07 (1993). MIT next claims that the department failed to make the necessary subsidiary findings in support of its calculation of the stranded costs that formed the basis of the CTC. Third, MIT asserts that the department’s allocation of 75% of the company’s stranded costs allegedly attributable to MIT is arbitrary and capricious. Finally, MIT argues that this court must reverse the department’s order because imposing the CTC on MIT results in the inequitable retroactive application of new decisional law.
We conclude that, contrary to the claim of MIT, the imposition of a customer transition charge, as such, does not violate State PURPA regulations, 220 Code Mass. Regs. §§ 8.00-8.07. However, because the department’s subsidiary findings, see *859G. L. c. 30A, § 11 (8),7 are insufficient for us to give any meaningful review to the calculation of the stranded costs that formed the basis of the CTC, and are similarly insufficient for us to determine whether its decision to permit the company to recover 75% of the costs attributable to MIT is arbitrary, we decline to affirm the department’s order, and remand for further proceedings. See G. L. c. 30A, § 14 (7). We are also unable to conclude on the basis of the department’s inadequate decision whether the imposition of the approved CTC on MIT violates principles prohibiting the inequitable retroactive application of new decisional law.
I
In 1985, MIT began investigating the option of constructing its own electricity-generating, cogeneration facility.8 In an attempt to keep MIT as a customer of the company, in 1992 the company proposed several options to MIT, including a buy-sell agreement, discounted rates, cogeneration deferral payments, and other potential ratemaking concessions.9 After MIT and the company failed to reach any agreement with respect to MIT remaining as an all-requirements customer, MIT began construction of its multimillion dollar, twenty-megawatt cogeneration facility in 1993.10 Operations commenced on September 16, *8601995. The cogeneration facility satisfied all of the criteria for, and was designated, a qualifying cogeneration facility (qualifying facility or QF) under the Public Utility Regulatory Policies Act of 1978 (PURPA), 16 U.S.C. § 824-824k (1994).11 Although MIT’s cogeneration facility can generate enough power to meet MIT’s normal electricity requirements, MIT still must purchase auxiliary services from the company to meet its needs when the facility experiences an outage, and to cover peak periods of usage.12
In May, 1994, prior to the completion of the construction of its facility, MIT filed a petition with the department to establish “just and reasonable” rates for three types of auxiliary services MIT would require from the company after completion of its cogeneration facility: standby service, maintenance service, and supplemental service.13 In response, on March 15, 1995, the company filed for approval its proposed rates for these services; *861it also proposed an additional monthly customer transition charge (CTC) of $7.49 per kilovolt amperes (kVa) to be assessed on all applicable14 customers on a monthly basis. The department consolidated MIT’s request and the company’s proposed CTC and its proposed auxiliary rates into a single case, D.P.U. 94-101/95-36.
The CTC proposed by the company was intended to recover what it claimed were the stranded costs that would result from MIT’s departure as a major customer of the company.15 The company claimed that, absent approval of the CTC, there was a likelihood of substantial cost shifting to residential and other small customers because of the loss of a customer the size of MIT relative to the company’s system. Although the CTC for which the company sought approval was designed to apply to any of its departing customers with demands above a specific level, to date MIT is the only customer to which the CTC is applicable; the remaining applicable customers have not indicated that they would or might depart as customers of the company.
In the proceedings before the department, the company explained the basis on which it had calculated the monthly CTC charge for which it sought authorization. It began by defining a load at-risk class, the class of customers whose departure from *862its system the company claimed would result in a significant loss of revenue to it. (The company determined that this class consisted of seven customers with an average monthly demand level of 2,000 kVa or greater.)16 Next, the company calculated the total net stranded cost that it claimed it would incur if every member of the load at-risk class were to depart as a full-service customer from the company. To calculate this cost, it first calculated the gross annual revenue associated with the class (nearly $21 million).17 From this amount the company subtracted an amount equal to the “avoidable variable expenses” that would not be incurred if all of the load at-risk customers left the system (almost $6 million). The company claimed that this resulted in an annual revenue requirement figure of approximately $15 million.
Next, the company mitigated what it claimed was the “gross amount of stranded costs” for the load at-risk class by the amount of (1) expected revenues from continued services provided to the customers should they depart (approximately $ 4.5 million); (2) potential revenues from reselling the generating capacity available to be marketed as a result of the custom*863ers’ departure (approximately $3 million); and (3) the revenues associated with anticipated load growth on the company’s system (approximately $1.6 million). Subtracting these mitigating factors, the company claimed “net stranded cost” of approximately $6 million. The company then divided the net stranded cost amount by the 66,806,000 kVa purchased annually by the entire load at-risk class of seven customers, and then divided the resulting amount by twelve to arrive at a monthly charge of $7.49 kVa. It sought authorization from the department to impose this monthly CTC charge on MIT.
Regarding the CTC proposed by the company, MIT filed a motion to dismiss or for partial summary judgment, which was denied.18 MIT also challenged the calculation of the CTC proposed by the company on several grounds: that the CTC was premature, was based on incorrect assumptions concerning the company’s capacity and costs, was based on incorrect calculations, was premised on faulty economics, was applied incorrectly to certain types of customers, and was contrary to established policy. The Attorney General, intervening as of right pursuant to G. L. c. 12, § 11E, also challenged the CTC proposed by the company. Contrary to MIT’s position, he argued that the department has authority under G. L. c. 164, § 94, to approve a CTC tariff. Nevertheless, he urged the department to reject the company’s proposed CTC because, as the department later described his position, “the company had been on notice since at least 1985 that MIT was actively considering self-generation and that the Company did not undertake any meaningful effort to mitigate the foreseeable consequences of losing MIT as a full requirements customer.” According to the Attorney General, the company had failed to demonstrate that *864the costs for which it sought reimbursement through the CTC were prudently incurred.19
The city of Cambridge (city) also submitted a brief opposing the CTC sought by the company. It requested that the department abstain from ruling on the company’s proposed auxiliary rates and its proposed CTC until the department had ruled on its own investigation into the restructuring of the electric industry.20 The city argued, in the alternative, that the company’s failure to prepare for the emergence of cogenerators in its customer base had created the stranded costs that the company now sought to recover through the CTC; the city urged the department to disallow the company from recovering these costs from its remaining customers.
On September 28, 1995, the department issued its order (D.P.U. 94-101/95-36), in which it approved in relevant part each of the company’s proposed auxiliary rates. In regard to the company’s proposed CTC, the department concluded that a CTC was allowable and did not violate PURPA because it did not single out QFs, such as MIT’s cogenerating facility, for discriminatory treatment. It then determined that the company could recover from MIT 75% of the CTC for which it had sought approval. It concluded that although MIT’s cogeneration plant was “representative of the business risk to which electric utilities have been exposed since at least 1978 when PURPA was enacted,” the company “could well have regarded the MIT project as one of any number of plans by customers that do not go beyond the design or financing stage.” The department noted that the load at-risk class, consisting of the company’s seven largest customers including MIT, was “broad” and “overstated” the risk to the company of all seven customers departing the *865company’s system at the same time.21 The department nevertheless determined that this “overstatement” did not “invalidate the class as conceived.” After “balancing] the interests of the load at risk class and the interests of [the city] and its other ratepayers by apportioning the stranded costs claimed by the Company,” .the department determined that a CTC of 75% of the company’s proposed charge was “just and reasonable under the unique circumstances of this case,” and authorized a monthly CTC charge to MIT of $5.62 kVa, 75% of the $7.49 kVa that the company had proposed.22
II
We address first MIT’s argument that the CTC authorized by the department violates the department’s PURPA regulations, 220 Code Mass. Regs. §§ 8.00-8.07. MIT asserts that the regulations require utilities to sell power to QFs such as its co-generating facility based on rate schedules that are nondiscriminatory as to a customer’s sources of power, and based on the characteristics of the load served. The proposed CTC, says MIT, expressly discriminates against customers who have a source of power other than the company. We conclude that a customer transition charge for the recovery of verifiable, prudently incurred stranded costs, as such, on departing customers does not violate the department’s PURPA regulations.23
The State PURPA regulations, the purpose of which is “to *866establish rules ... for determining rates, terms, and conditions for the sale of electricity by utilities to . . . qualifying cogenerators,” 220 Code Mass. Regs. § 8.01(1), are silent on the issue of recovering stranded costs from QFs; the regulations neither expressly authorize, nor expressly prohibit, the recovery of these costs through the means of a customer transition charge. MIT argues that the company’s CTC violates 220 Code Mass. Regs. § 8.06, which specifically governs rates for supplementary, back-up, maintenance, and interruptible power to QFs pursuant to 18 C.F.R. § 292.305(b) (1997). We disagree.
First, because the CTC is not a rate for supplementary, back-up, maintenance, or interruptible power, that regulation neither governs, nor is violated by, the company’s proposed CTC.24 Second, in light of the changing regulatory regime of the electricity industry, the recovery by utilities of prudent and verifiable stranded costs is consistent with sound public policy, as has increasingly been recognized in recent years. On April 24, 1996, the Federal Energy Regulatory Commission (FERC) issued Order No. 888, 61 Fed. Reg. 21,540 (1996)25 in which it said that “recovery of legitimate, prudent and verifiable stranded costs should be allowed,” id. at 21,540, and specifically suggested that States consider imposing exit fees on departing customers as possible mechanisms to recover these costs.26 While FERC concluded that “it is appropriate to leave it to state regulatory authorities to deal with any stranded costs occasioned by retail wheeling,”27 id. at 21,647, we find the FERC’s policy to allow the recovery of stranded costs through a mechanism like an exit *867fee instructive.28 The department also has recognized that, in light of the changing regulatory landscape, the recovery of prudently incurred stranded costs by the utility is in the public interest. See D.P.U. 95-30, at 29-31 (1995).
Accordingly, we agree with the department that the imposition of a customer service charge that applies to a broad class of customers, and that does not single out QFs for unlawful discriminatory treatment, does not violate the State’s PURPA regulations. We concur that the recovery of prudent and verifiable stranded costs incurred by utility companies, as appropriately authorized, is in the public interest and consistent with PURPA. We are not, however, persuaded by the department’s decision or on the basis of the record before us that the stranded costs incurred by the company for which it seeks recovery from MIT necessarily were prudently incurred, as we describe below. Accordingly, we decline to affirm the department’s order.
Ill
We turn to consider MIT’s challenges to the calculation of the CTC itself, and its challenge to the department’s approval of 75% of the CTC sought by the company. Our standard of review of petitions under G. L. c. 25, § 5, is well settled: a petition that raises no constitutional questions requires us to review the department’s finding to determine only whether there is an error of law. See Costello v. Department of Pub. Utils., 391 Mass. 527, 532 (1984), and cases cited. The burden of proof is on the appealing party to show that the order appealed from is invalid, and we have observed that this burden is heavy. Id. at 533. Wolf v. Department of Pub. Utils., 407 Mass. 363, 367 (1990). Moreover, we give deference to the department’s expertise and experience in areas where the Legislature has delegated to it decision-making authority, pursuant to G. L. c. 30A, § 14. We shall uphold an agency’s decision unless it is *868based on an error of law, unsupported by substantial evidence, unwarranted by facts found on the record as submitted, arbitrary and capricious, an abuse of discretion, or otherwise not in accordance with law. G. L. c. 30A, § 14 (7). Moreover, we have acknowledged that the department has broad authority to determine ratemaking matters in the public interest. See Wolf, supra at 369; Commonwealth Elec. Co. v. Department of Pub. Utils., 397 Mass. 361, 369 (1986), cert, denied, 481 U.S. 1036 (1987); Lowell Gas Light Co. v. Department of Pub. Utils., 319 Mass. 46, 52 (1946); Boston v. Edison Elec. Luminating Co., 242 Mass. 305, 309 (1922).
In order to make possible our determination of these questions of law, we “carefully review the department’s findings for error.” Costello v. Department of Pub. Utils., supra at 533. We have observed in that respect that G. L. c. 30A, § 11 (8), requires the decision of the department to “be accompanied by a statement of reasons . . . including determination of eách issue of fact or law necessary to the decision.” See id.; New York Cent. R.R. v. Department of Pub. Utils., 347 Mass. 586, 593 (1964). A purpose of that statutory provision is to reqúire the department to give a “ ‘guide to its reasons’ so that this court may ‘exercise . . . [its] function of appellate review.’ There is thus a ‘duty to make adequate subsidiary findings.’ ” Hamilton v. Department of Pub. Utils., 346 Mass. 130, 137 (1963); quoting Leen v. Assessors of Boston, 345 Mass. 494, 502 (1963). Without an adequate statement of reasons, “we are .unable to determine whether an appellant has met his burden of proof that a decision of the department is improper.” Costello, supra at 533. Because this case concerns an issue of first impression, the need for an adequate statement of reasons by the department is even more pressing.29
We find the department’s decision inadequate in several material respects. First, the department has failed to make the *869subsidiary findings necessary for us to evaluate whether MIT has met its burden of proving that the company’s calculations of its stranded costs are erroneous and that the department’s decision to accept those calculations is erroneous.30 The department gives no explanation as to why it accepted the company’s methodology and rejected MIT’s challenges to the company’s calculations; MIT challenged the stranded cost calculations on numerous grounds, none of which is addressed.31 By way of example, the company claimed that $6 million in stranded costs should be collected through a CTC. The methodology the company applied to calculate this amount was based on the hypothetical annual net stranded costs that would be incurred if the entire load at-risk class were to depart as full-service customers at the same time.32 In so doing it overstated the actual stranded costs incurred as a result of MIT’s departure. In its decision, the department recognized that “the risk to [the company] of all these [load at-risk] customers’ departing [the company’s] system is overstated,” but determined, without explanation, that the risk was “not enough to invalidate the class as conceived.” We recognize that, in light of the evolving market, some reasonable approximation may be required. But *870the department makes no findings to support or elucidate its conclusion that the risk of all seven customers departing is a reasonable approximation, or why the company’s methodology of calculating the stranded costs by dividing the hypothetical costs equally among all potential departing customers is reasonable. We note, for example, that one member of the load at-risk class recently signed a seven-year contract with the company. The department apparently did not allow for any adjustment of the company’s calculation of its actual stranded costs in light of this reality.
MIT offered testimony challenging other aspects of the company’s stranded costs calculation. For example, MIT questioned whether the company understated the revenues it would receive from auxiliary service provided to departing customers, claimed that the company had used inappropriately low market prices for sales of its electricity generation, and underestimated its load growth. The result could be that the mitigation measures the company assumed in its calculation of stranded costs is significantly underestimated. We are not in a position to determine whether any of these or other claims by MIT are valid because the department failed to address them in its order. The department commented that some of the figures used to calculate the CTC “involve assumptions by the Company that introduce a margin of error into the Company’s calculation,” and concluded that “[ujnfortunately, these uncertainties cannot be precisely quantified in order to adjust the CTC,” and that the quantification of these uncertainties is “impractical at this time of transition toward restructuring.”33 While we recognize that some uncertainties cannot be precisely quantified, we do require more than a conclusory statement to that effect.
We have said that in cases such as this, where the evidence is conflicting, the administrative agency is “charged with the responsibility of making findings of fact and is in the best position to judge the credibility of the witnesses.” Costello, supra at *871536, quoting School Comm, of Chicopee v. Massachusetts Comm’n Against Discrimination, 361 Mass. 352, 354-355 (1972). But we have insisted that the agency make subsidiary findings of fact on all issues relevant and material to the ultimate issue to be decided, and that it “set forth the manner in which it reasoned from the subsidiary facts so found to the ultimate decision reached.” Id. The department’s conclusions concerning the calculation of stranded costs do not meet these standards, leaving us no alternative but to speculate whether MIT’s claims, have any merit. This we decline to do. See Costello, supra at 536, quoting Northeast Airlines, Inc. v. Civil Aeronautics Bd., 331 F.2d 579, 586 (1st Cir. 1964) (“[c]ourts ought not to have to speculate as to the basis for an administrative agency’s conclusion”).34
Second, the department has failed to explain why the allocation to MIT of 75% of the stranded costs as calculated by the company is reasonable.35 The apparent arbitrariness of this allocation is underscored by the concurring opinion in which one commissioner suggested, also with no findings to support her conclusion, that a 60% allocation would be more appropriate. General Laws c. 30A, § 11 (8), requires more from the department than merely stating in conclusive terms that such an allocation is “just and reasonable under the unique circumstances of this case.”36 We are also concerned that, assuming the calculation of stranded costs is supportable, assuming that the *872stranded costs were prudently incurred, and assuming further that such costs are attributable to MIT’s departure, the allocation the department approved — that 75% of the stranded costs be charged to MIT — may inappropriately leave the remainder of any such costs to be borne by the company’s other ratepayers. The department’s decision is ambiguous as to whether the remaining 25% is to be allocated to the company’s other ratepayers or to be borne by the company’s shareholders.37 Any customer transition charge approved by the department must ensure that utility companies be allowed to recover only verifiable, prudent, and reasonable stranded costs, and that these costs be recovered from the parties to whom they are attributable — mainly the departing customers — and not the utilities’ remaining ratepayers.38 Before we are in a position to sustain an allocation of 75% of the stranded costs to MIT, we require a reasonable and more complete explanation from the department of its percentage allocation.39 Merely listing the factors it considered in reaching its determination is not adequate. The *873department does not explain how it balanced those factors, and on the record before us it appears that it gave little or no weight to at least two factors — the goal of encouraging competition and MIT’s interest in lower rates — also without explanation. Because we are unable to conduct a meaningful review of the department’s decision to allocate 75% of the stranded costs to MIT, we decline to “supply a reasoned basis for the [department’s] action that the [department] itself has not given.” Costello v. Department of Pub. Utils., supra at 536, quoting Bowman Transp., Inc. v. Arkansas-Best Freight Sys., 419 U.S. 281, 285-286 (1974).
Third, we are concerned that there is inadequate explanation in the department’s decision for us to determine whether the stranded costs for which the company now seeks relief were prudently incurred. We note that the Attorney General raised this concern before the department,40 as did the city.41 The department concluded that:
“Although it is not beyond dispute that [the company] did all it could to mitigate the loss of MIT in a timely way, we recognize that throughout the period when MIT was refining its self-generation planning, the Company also had an obligation to serve both all-requirements and partial-requirements customers. While MIT assessed the future course most advantageous to it, [the company] had to obtain resources available to serve all its customers. MIT was just such a customer, and given the narrow reserve capacity margins that existed in the late 1980’s, the Company might well have been criticized had it proposed to treat MIT’s plans since 1985 as absolving it of its obligation to serve that load.”
*874We cannot accept this conclusion of the department on the basis of the record before us.42 The department has emphasized, in connection with the changing regulatory scheme governing the electric industry, that it has sought to promote the “aggressive mitigation of stranded costs.” D.P.U. 96-100, at 55. The department has observed that “utilities are in a position to avoid entering into or incurring new commitments and costs that might be stranded by the emergence of customer choice,” and has “direct[ed] utilities to take immediate steps to avoid the creation of such costs.” D.P.U. 95-30, at 32.
The department’s concern with the reasonableness and prudence of stranded costs reflects the Federal policy on this issue. In Order No. 888, FERC addressed a request by interested parties to eliminate the term “prudent” from the definition of stranded costs. FERC declined to do so: “we will retain the requirement that stranded costs be legitimate, prudent and verifiable.’ ” 61 Fed. Reg. 21,540, 21,664 (1996). FERC also noted that
“Prudence of costs, depending upon the facts in a specific case, may include different things: e.g., prudence in operation and maintenance of a plant; prudence in continuing to own a plant when cheaper alternatives become available; prudence in entering into purchased power contracts, or continuing such contracts when buyouts or buy-downs of the contracts would result in savings. The Commission therefore cannot make a blanket assumption that all claimed stranded costs will have been prudently incurred.”
Id. In this case, we, too, decline to make any “blanket assumption” that the costs the company now wishes to recover through its CTC were prudently incurred. The department acknowledged that it is “not beyond dispute” that the company acted prudently in incurring these costs.43 To allow a utility to recover imprudent or unreasonable costs would impede the very policies identified *875by the department as worthy of promotion, the interests of customers in obtaining lower cost power and the public policy interests in encouraging competition as a means to increase efficiency.
For all these reasons we remand this matter to the department for further findings. The department should at a minimum provide us with a statement of reasons, including subsidiary findings, why it accepted the company’s calculation of stranded costs in light of the conflicting testimony (and why it rejected MIT’s challenges to each of the component parts of the calculation); why an allocation of 75% of those costs to MIT was not arbitrary; why it concluded the stranded costs for which the company seeks reimbursement through the CTC were prudently incurred. (and why it rejected claims by MIT, the Attorney General, and the city that the costs were not prudently incurred); and what, if any, steps the company undertook to “aggressively mitigate” its costs, particularly in light of the advance notice that it had of MIT’s departure. It should, in short, provide us with a decision on the basis of which we will be able to conduct our appellate review. Costello v. Department of Pub. Utils., 391 Mass. 527, 537 (1984); School Comm. of Chicopee v. Massachusetts Comm’n Against Discrimination, 361 Mass. 352, 355 (1972).
IV
Last, MIT argues that the application of the CTC violates principles prohibiting the inequitable retroactive application of new decisional law. The department rejected this argument, reasoning that, in D.P.U. 95-30, it had announced the policy that “utilities should have a reasonable opportunity to recover, through an appropriately designed mechanism, net, non-mitigatable, stranded costs associated with commitments previously incurred pursuant to their legal obligation to serve.”
In Massachusetts Elec. Co. v. Department of Pub. Utils., 383 Mass. 675 (1981), we observed that “[ejvery administrative *876adjudication which involves the application of a new standard of conduct has a retroactive effect and must therefore be subject to judicial scrutiny,” and “when the administrative agency does choose adjudication over rulemaking, judicial review must balance the adverse effects of retroactivity ‘against the mischief of producing a result which is contrary to a statutory design or to legal and equitable principles. If that mischief is greater than the ill effect of the retroactive application of a new standard, it is not the type of retroactivity which is condemned by law.’ ” Id. at 679, quoting SEC v. Chenery Corp., 332 U.S. 194, 203 (1947). In Massachusetts Elec. Co., supra at 681, we upheld the department’s action — the dismissal of a utility’s rate filing — because (1) the policy to be enforced had been expressed by the department in earlier cases; (2) furtherance of that policy was within the department’s powers; and (3) the dismissal did not run afoul of the Chenery balancing test.
While we recognize that the CTC at issue here was the first such charge authorized by the department to recover stranded costs from a departing customer, we take into consideration the fact that the department has announced its policy to allow utilities to recover stranded costs. See D.P.U. 95-30 (1995). Further, we agree with the department that approving such a charge is within the broad supervisory power granted to it by the Legislature in G. L. c. 164, § 94.44 However, on the basis of the record before us, we cannot determine whether what was in this instance the retroactive imposition of the CTC would “run afoul of the Chenery balancing test.” Id. We observe, in particular, that the department does not address whether it is equitable to impose the CTC authorized here on MIT when MIT undertook the construction of its cogeneration facility years prior to the department’s announcement in August, 1995, *877that stranded costs could be recovered by a utility company.45 In a sense this consideration is the mirror image of the company’s claim that it would be inequitable to allow MIT to exit the customer base when the company relied on it as a continued exclusive customer. As noted previously, MIT’s construction of its cogeneration facility was consistent with national policy, and it may well be — we cannot discern from this record — that MIT could not reasonably have anticipated that stranded costs of the magnitude authorized here would be imposed on it.46 We also note that the department’s decision does not address MIT’s claim that the financial burdens of the authorized CTC all but eliminate the benefits of cogeneration.47 Whether or not MIT reasonably proceeded with the construction of its cogeneration facility in reliance on existing rules of the department and *878whether MlT has unreasonably been burdened with financial consequences that it could not have anticipated and that are not equitable cannot be discerned from the department’s decision.48 We are thus unable to determine whether the application of the CTC to MIT in the circumstances of this case “is contrary to a statutory design or to legal and equitable principles.” Massachusetts Elec. Co., supra at 679, quoting Chenery, supra at 203. The department must subject any stranded costs sought to be recovered by the company to careful scrutiny and to explain in a careful and reasonable manner why the imposition of such costs does not violate the strictures of Chenery, supra.
The case is remanded to the county court where, a judgment should be entered vacating the department’s order and remanding the matter to the department for further findings consistent with this opinion.
So ordered.
A “cogeneration facility” is one that produces both electric energy and steam or some other form of useful energy, such as heat. 16 U.S.C. § 796(18)(A). See Federal Energy Regulatory Comm’n v. Mississippi, 456 U.S. 742, 750 n.11 (1982). The MIT facility produces two forms of energy sequentially, electricity and steam. National energy policy encourages the construction and operation of cogeneration facilities such as MIT’s. See Public Utility Regulatory Policies Act of 1978 (PURPA), 16 U.S.C. §§ 824-824k (1994); Energy Regulatory Comm’n v. Mississippi, supra at 750 (PURPA “seeks to encourage the development of cogeneration and small power production facilities. Congress believed that increased use of these sources of energy would reduce the demand for traditional fossil fuels”).
MTT also pursued three collateral attacks on the customer transition charge (CTC) authorized by the department. First, on. January 5, 1996, MIT petitioned die Federal Energy Regulatory Commission (FERC) to initiate an enforcement action against the department pursuant to 16 U.S.C. § 824a-3(h), claiming that the CTC was discriminatory and in violation of Federal PURPA statutes. FERC denied MIT’s petition, holding that this court is a more appropriate forum for determining whether the department, a State agency, properly applied its own PURPA regulations. Massachusetts Inst, of Tech., 74 Fed. Energy Reg. Comm’n Rep. (CCH) par. 61,221 (1996).
Next, MIT filed an action in the United States District Court for the District of Massachusetts seeking an enforcement action against the department. On August 27, 1996, a judge of that court dismissed the case, determining that the court did not have jurisdiction over MIT’s claim under 16 U.S.C. § 824a-3(h). Massachusetts Inst. of Tech. v. Massachusetts Dep’t of Pub. Utils., 941 F. Supp. 233, 238 (D. Mass. 1996). MIT did not appeal from that judgment.
MIT also filed a petition with the department on February 22,1996, concerning the company’s supplemental service rate approved in D.P.U. 94-101/95-36. The department granted the company’s motion for summary judgment on *858September 24, 1996. Cambridge Elec. Light Co., D.P.U. 96-33 (1996). MIT did not appeal from that order.
“Stranded costs” refer to the costs prudently incurred by utilities, recovery of which may be jeopardized by the transition from existing regulation and exclusive retail franchises to a more competitive workplace in which customers are permitted to purchase electricity from a source other than their host utility. See Baumol, Stranded Costs, 18 Harv. J.L. & Pub. Pol’y 835, 835 (1995) (stranded costs refer to “those costs that the utilities currently are permitted to recover through their rates but whose recovery may be impeded or prevented by the advent of competition in the industry”). Stranded costs include (a) the imbedded costs from existing generating facilities; (b) the amount by which the cost of existing contractual commitments for purchased power exceeds the market value of such contracts; and (c) various regulatory assets. Electric Indus. Restructuring, D.P.U. 95-30, at 31-32 (1995).
Prior to 1995, when MIT’s cogeneration facility commenced operations, MIT was one of the company’s two largest customers.
As a result of the CTC, in combination with three auxiliary service rates also approved by the department described below, MIT would be required to pay as much as 35% of its previous utility bill, even though it is taking less than 10% of the amount of service that it formerly took from the company, all but eliminating the benefits of its cogeneration facility.
General Laws c. 30A, § 11 (8), provides that “[e]very agency decision . . . shall be accompanied by a statement of reasons for the decision, including determination of each issue of fact or law necessary to the decision.”
In its order denying MIT’s petition for enforcement under § 10 of PURPA, FERC found that MIT “began considering self-generation in 1985 in response to the company’s rate increases.” 74 Fed. Energy Reg. Comm’n Rep. (CCH) par. 61,221, at 61,748 (1996). The company submitted evidence to the department that it was aware of MIT’s investigation of cogeneration options from ten to twelve years before this matter came before the department; MIT had shared with the company feasibility studies prepared for MIT as early as 1985.
During the discussions between the company and MIT, the company claims that it discussed with MIT its goal to minimize the impact of the potential loss of MIT’s load on the company’s remaining customers, and its attempt to avoid stranded costs. According to the company, it informed MIT that if an equitable arrangement with MIT could not be reached, the company “could” be forced to seek the recovery of stranded costs.
It is not clear from the record exactly when MIT’s plans to begin construction of its cogeneration facility solidified, nor is it clear when the company knew about those plans. According to the department, the earliest occasion on which the company indicated that it anticipated the loss of MIT as an all-*860requirements customer in its contingency planning is in its 1991 integrated resource management (IRM) proceeding, Commonwealth Elec. Co., D.P.U. 91-234 (1992). The department does not address whether the company should have anticipated the loss of MIT before that date, nor does it discuss what steps, if any, the company took in anticipation of MIT’s departure, either in 1991 or previously. Because the company may recover only prudently incurred stranded costs, the negotiations between MIT and the company prior to MIT’s departure, and the company’s response to that departure, are singularly relevant.
A “qualifying cogeneration facility” is “a cogeneration facility which (i) the [FERC] determines, by rule, meets such requirements (including requirements respecting minimum size, fuel use, and fuel efficiency) as the [FERC] may, by rule, prescribe; and (ii) is owned by a person not primarily engaged in the generation or sale of electric power (other than electric power solely from cogeneration facilities or small power production facilities).” 16 U.S.C. § 796(18)0).
MIT must currently purchase these services from the company because MET is in the company’s franchise area. However, as the electricity market experiences deregulation, MIT may in the future be able to purchase auxiliary services from a provider other than the company. The department has issued a ruling requiring the State’s utilities to propose choice-of-supplier plans and separate power transmission by January, 1998. See D.P.U. 96-100 (1996).
Standby service provides a back-up supply of power when a customer’s alternative source of power experiences a forced outage. Maintenance service provides back-up power when the customer’s alternative source of power is withdrawn from service for scheduled maintenance. Supplemental service supplements the output of the customer’s alternative source of power where the alternative source of power serves less than the customer’s maximum electrical load. Under PURPA, electric utilities are required to supply such auxiliary services to QFs. See 220 Code Mass. Regs. § 8.06 (1993); 18 C.F.R. § 292.305(6) (1997).
See note 15, infra.
The CTC is charged not as a function of the customer’s usage, but as a function of the load (the amount of electrical power delivered or required at any specified point or points on the system) no longer served as a result of a customer discontinuing full-requirements service. It is of limited duration. The CTC sought by the company was to be applicable to any customer who discontinued all or a portion of full-requirements service from the company and who meets the following criteria:
(a) the customer has an average billing demand of 2,000 kVa, or greater, for the most recent calendar year;
(b) the customer obtains electric service from a source other than the company;
(c) the customer remains in the company’s service area.
The company proposed that the CTC be effective on a customer’s departure through December 31, 2000, and that the rate of $7.49 per kVa change only if one of the following conditions occur: (1) a significant change in the current surplus of capacity in the market; (2) a major industry restructuring that affects Massachusetts; or (3) a general rate case filing by the company.
A witness for the company explained how the load at-risk class was defined:
“The analysis to determine the load at risk began with reviewing the 1994 average billing determinants for all accounts within [the company’s] system. Using this data, I analyzed the usage and revenue levels of the accounts in order to determine whether there was a distinct point on the usage and revenue continuum which would provide a threshold for determining loads whose loss would have a significant impact on the [c]ompany. Based on this analysis,. I determined that at an average monthly demand level of 2,000 kVa the customer load levels increased rapidly while the number of customers decreased rapidly.”
Using this definition, the load at-risk class consists of seven customers, including MIT. In 1994, these seven customers purchased 66,806,000 kVa per year for total revenues of approximately $24 million, approximately 19% of the company’s total annual revenues. The class was defined solely by the usage and revenue levels of the accounts; whether any of the customers would depart or were likely to depart from the company was not a consideration.
The company arrived at this amount by using its cost-of-service study, approved in Cambridge Elec. Light Co., D.P.U. 92-250 (1992), as a “base case scenario” for the calculation of the stranded costs. From this, the company recalculated the costs of service by disaggregating the amount of load at risk into a separate class resulting in an allocated revenue requirement for the load at-risk class.
MIT argued that the department lacked authority under G. L. c. 164, § 94, to approve the CTC because the CTC is not a “rate, price, or charge for the sale or distribution of electricity or for its consumption,” that the proposed CTC was inconsistent with and preempted by PUREA, and that the CTC was not justified because the company had failed to identify any investment or costs that were incurred to serve MIT or any other large customer. MIT also argued that the department’s regulations do not authorize charging stranded costs to QFs; that the CTC constituted an improper tying arrangement in violation of antitrust laws; and that applying the CTC to MIT would result in retroactive rulemaking, and was thus prohibited. The department rejected every argument.
The Attorney General argued specifically that the company “has not demonstrated in this proceeding that the costs in question were incurred. prudently, i.e., that their incurrence was reasonable in light [of] the lack of a basis to assume MIT would remain a total requirements customer and that all reasonable steps were taken to mitigate any costs that were incurred prudently.” On appeal to this court the Attorney General represents the department.
Subsequently, on August 16, 1995, the department issued Order 95-30 which sets forth its own policy in light of the restructuring of the electric industry. The department concluded that “responsible policy must provide electric utilities a reasonable opportunity to recover net, non-mitigatable stranded costs” during this period of transition. Id. at ii.
There was evidence that the company was not aware of any plans by any large customer, other than MIT, to generate its own electricity or otherwise abandon full-service status. To the contrary, the company had recently secured a seven-year agreement to retain Harvard University, a member of the load at-risk class, as a customer.
One commissioner issued a concurring opinion, in which she stated that, “[i]f the departing customer is asked to bear a disproportionate share of costs the utility claims to have incurred to serve him, he may not be able to exercise his choice [to remain as a full-service customer or to pursue alternatives to his traditional utility suppliers] effectively.” The commissioner expressed concern “that the imposition of 75% of the CTC charge on MIT will undermine the development of the competitive generation market that will ultimately produce benefits for all customers.” She recommended that the CTC imposed on MIT be reduced to 60%.
While we conclude that a customer transition charge does not necessarily violate PURPA regulations, we remand this matter to the department for further findings, for the reasons discussed below. If we subsequently determine that the company acted imprudently in incurring costs that are attributed to providing services to MIT, a CTC as applied to MIT may violate PURPA.
As we note above, in the same proceeding in which the department approved the CTC, it also approved the company’s proposed standby, maintenance, and supplemental rates. These rates are clearly governed by 220 Code Mass. Regs. § 8.06. MIT does not appeal from these rates.
The full title of Order No. 888 reads, Promoting Wholesale Competition Through Open Access Non-Discriminatory Transmission Services by Public Utilities; Recovery of Stranded Costs by Public Utilities and Transmitting Utilities.
On May 1, 1996, the department issued its own extensive set of proposed rules: 220 Code Mass. Regs. §§ 11:00 et seq. proposed rules governing the restructuring of the electric utility industry. Cf. D.P.U. 96-100, promulgating such rules (1996).
“Wheeling” is defined as “any transmission of electricity by a utility to another utility from sources other than its own generation or purchased power sources.” 220 Code Mass. Regs. § 8.03(5).
As noted above, MIT petitioned FERC for an enforcement action against the department, claiming that the CTC was inconsistent with PURPA’s goals and was discriminatory against QFs in violation of PURPA. In holding that MIT’s petition was not within FERC’s jurisdiction under 16 U.S.C. § 210(h), FERC indicated that the CTC does not discriminate against QFs in violation of PURPA: “MIT claims that the [CTC] discriminates against QFs, allegedly in violation of PURPA. We conclude that it does not” (footnote omitted). Massachusetts Inst, of Tech., 74 Fed. Energy Reg. Comm’n Rep. 61,221, at 61,750 (1996).
The department noted that the company’s “method of determining stranded costs is not supported by precedent, as this is an issue of first impression.” It also observed that in D.P.U. 95-30, at 46-47 (in which the department stated that utilities should have a reasonable opportunity to recover net, nonmitigable stranded costs), it “did not endorse any specific methods for calculating stranded costs; rather we anticipated proposals for such methods evolving from negotiations among all appropriate representatives.” This is the first occasion on which we review the application by the department of the restructuring principles that it established in D.P.U. 95-30. The department applied those principles even though it acknowledges that MIT’s departure from the *869company “was not the result of the Department’s electric industry restructuring initiative.”
In its brief to this court the department recognizes that it did not “recite a specific finding as to every single subcalculation” and argues that G. L. c. 30A, § 11 (8), may not require such an exercise. While we may agree that § 11 (8) may not require a specific finding as to every single subcalculation in every case, here there are few subcalculations, none of which have been addressed. The statute requires more than the department provided. The department summarized at length the company’s position, then summarized MIT’s objections and comments, as well as those of other participants in the proceedings, including the Attorney General and the city, and concluded by summarizing the company’s responses to those, all without comment or guidance to us. Its “Analysis and Findings” that follow are brief, sweeping, and are not supported by subsidiary findings. In its brief to this court, the department argues that the evidence “amply supported” its approval of the company’s calculations, referring us for the first time to numerous parts of the record. The barrage of footnotes in its brief is not an adequate substitute for the requisite subsidiary findings in its decision.
The department did observe that it “also has concerns about the method of calculating the CTC,” but did not attempt to explain why it resolved its concerns wholly in favor of the company.
We acknowledge that the company’s CTC is intended to apply to any member of the load at-risk class that chooses to depart as a full requirements customer of the company. In reality, however, the only customer that will be subject,to the CTC at least at this time, the record indicates, is M3T.
In a similar vein, the department concluded, without elaboration, that the mitigating factors that the company had used to calculate its net stranded costs “introduce a margin of error into [the company’s] calculation,” noting that “[c]ontinued service revenue, market value, and load growth mitigation are all difficult to project accurately.” Again, the department does not explain why it rejected MIT’s challenges to these calculations or why it accepted the company’s proposed cost estimate. We are not able, therefore, to determine whether the department has reached a decision that should be sustained.
A witness on behalf of the company testified that “[t]he methodology selected should be consistent with the overriding objectives of equity and economic efficiency,” and that “from both an equity and efficiency standpoint, the most appropriate way to collect [stranded] costs is to internalize these costs to those customers whose actions caused them to be stranded in the first place. . . . [Requiring customers responsible for such costs to pay them will serve to minimize the impacts on remaining customers who may not have the same alternatives.” On the basis of the record before us, we are skeptical that the proposed CTC satisfies the company’s own criteria.
The department determined that “after balancing the (1) Company’s interests in recovering prudently incurred costs, (2) ratepayers’ interests given the economic impact associated with the departure of a large customer of the Company, (3) load at risk class members’ (including MIT) interests in obtaining lower cost power, and (4) public policy interests in encouraging competition as a means to increase efficiency ... a CTC of 75 percent of the net stranded costs as calculated by [the company] (or $5.62 per kVa) is just and reasonable under the unique circumstances of this case.”
The department argues that MIT will pay only its share of the stranded costs caused by the departure of the load at-risk class, but it does not explain why the company’s calculations, based as they were on so many assumptions, *872were “reasonable” or why 75% of MIT’s purported “share” of the stranded costs caused by the load at-risk class is appropriate.
Both the Attorney General and the city had argued before the department that any stranded costs should not be borne by the company’s remaining customers but by the company and MIT. The department stated in its brief that it has “not decided the extent to which the remaining ratepayers, as opposed to [the company] itself, should bear stranded costs that are not recoverable through the CTC.”
Conversely, any stranded costs that are not verifiable, prudent, and reasonable should not be recoverable from the departing customer. A witness on behalf of the company testified that utility company shareholders should not bear the costs of stranded costs. We agree that the interests of a utility’s shareholders may be protected, but only to the extent that the recovery of net stranded costs is limited to verifiable and prudent expenditures. On remand, the department should ensure that any authorization to recover stranded costs is consistent with the holding that “the function of the department is the protection of public interests and not the promotion of private interests.” Lowell Gas Light Co. v. Department of Pub. Utils., 319 Mass. 46, 52 (1946). See Commonwealth Elec. Co. v. Department of Pub. Utils., 397 Mass. 361, 369 (1986), and cases cited.
In its brief to this court the department argues that it approved the CTC on MIT because the departure of MIT from the company “could cause a significant increase in the electric rates of the remaining customers [of the company].” There is no finding by the department that this will, or is likely, to occur. We have upheld the apportionment of costs among the affected parties in order to minimize the impact on consumers and have recognized that allocation of these costs is within the department’s discretion. Attorney Gen. v. Department of Pub. Utils., 390 Mass. 208, 231 (1983). Nevertheless, the *873department must explain the rationale for its allocation before we can approve it. Here the department merely concludes that imposition of 75% of the net stranded costs is reasonable “under the unique circumstances of this case” and that this apportionment on MIT “is in the broad public interest.” Without more, that is insufficient.
See note 19, supra.
The city argued before the department that the company did not take any measures to mitigate the financial consequences of MIT’s cogeneration facility, that it did not begin planning for MIT’s departure until 1993, when the department suggested that the company consider doing so, and that, between 1985 and 1992, when the company knew of MIT’s cogeneration plans, the company continued to enter into multi-year purchase contracts, the shortest such contract being for a term of fifteen years.
There are no subsidiary findings supporting the ultimate finding, and the department appears to have accepted the company’s argument, without addressing the specific challenges to its claim made by MIT, the Attorney General, or the city.
The department appears to have concluded, erroneously, that the company’s “good faith” negotiations with MIT (in an attempt to keep MIT as a full-service customer) absolved the company of any other efforts to mitigate *875its stranded costs: “while questions have been raised about whether [the company] should have anticipated MIT’s departure sooner and whether it should have acted earlier and more aggressively to reduce the impact on ratepayers of such an event, the Department finds no clear evidence of imprudence on [the company’s] part. The record does not permit us to conclude that [the company] did not act in good faith in its negotiations to avoid losing MIT as a customer.”
MIT had argued before the department that it lacked jurisdiction under G. L. c. 164, § 94, to approve the CTC because it did not directly relate to the sale or consumption of electricity. The department concluded that it did have jurisdiction, reasoning that the Legislature, through G. L. c. 164, § 94, had granted the department broad supervision and authority over electric companies, and citing similar charges approved and upheld charges not directly related to the sale or consumption of electricity. See, e.g., Monsanto Co. v. Department of Pub. Utils., 412 Mass. 25, 28 (1992) (upholding energy conservation charges which utilities required for the costs of demand-side management activities). MIT does not raise this claim on appeal.
In its decision the department did note that in D.P.U. 95-30 its focus was on stranded costs that “may arise as the result of Department initiatives to promote competition in the generation sector of the electric utility industry by giving customers access to other sources of electricity supply in addition to the electric company in whose territory these customers were located,” that the company’s authorization for the CTC tariff imposed on MIT “differs somewhat” from that model, and that the company’s filing for authorization of the CTC predated D.P.U. 95-30. The department did not specifically address the issue of retroactivity except to state summarily that “[i]n fairness, the Company should have a reasonable opportunity to recover its investments, including those undertaken to serve MIT,” citing Federal Power Comm’n v. Hope Natural Gas Co., 320 U.S. 591 (1944); Bluefield Water Works & Improvement Co. v. Public Serv. Comm’n of W. Va., 262 U.S. 679 (1923).
The Energy Policy Act of 1992, Pub. L. 102-486, 106 Stat. 2776 (1992), sanctioned wholesale power competition by, inter alia, instructing FERC to order utilities to provide access to the nation’s transmission grid to third-party suppliers. The department has recognized that this dramatically new economic and regulatory framework threatens to reduce the “opportunity that currently exists for electric utility companies to recover their past investments in generating plants.” It is noteworthy, however, that MIT’s cogeneration facility is of a kind long sanctioned by national and local energy policy and that MIT’s commitment to build its facility predated any efforts by the department to respond to the restructuring of the electric industry at large.
The department argues that MIT could not reasonably have relied upon “a particular resolution” of the question of stranded cost recovery applied to it, because the department previously had not addressed a situation “in which a large customer of a small utility discontinued its all-requirements purchases from that utility but remained in that utility’s service territory.” If the department is to rely on its later articulated principles as described in D.P.U. 95-30, it is incumbent on it to articulate how it assessed MIT’s interests in obtaining lower cost power and the public policy interests in encouraging competition as a means to increase efficiency, and that these interests were not made subservient to the company’s interests in recovering its stranded costs.
The department argues that in the context of the larger developments that are taking place in the electric industry, the transitional recovery of stranded costs is in the public interest and that the department had the responsibility to balance the relevant interests in this case in the exercise of its broad ratemaking discretion. We agree. But this does not absolve the department from considering whether the general principles regarding the recovery of stranded costs as articulated in D.P.U. 95-30, and as applied specifically to this cogeneration facility brought on-line prior to 1995, run afoul of equitable principles.