Three utility companies (utilities) challenge orders entered against them by the Department of Public Utilities (department) that impose monetary penalties for failing to “restore service to [their] customers in a safe and reasonably prompt manner,” in violation of 220 Code Mass. Regs. § 19.03(3) (2010), after electrical outages arising from Tropical Storm Irene (Irene) on August 28, 2011, and a snowstorm two months later on October 29 (October snowstorm). The utilities — Massachusetts Electric Company and Nantucket Electric Company, each doing business as National Grid (collectively, National Grid); NSTAR Electric Company (NSTAR); and Western Massachusetts Electric *555Company (WMEC) — claim on appeal that (1) the department made an error of law in failing to apply the prudence standard when assessing the utilities’ storm performances; (2) the department’s findings were not supported by substantial evidence; and (3) the department’s penalty calculations lacked the necessary subsidiary findings and constituted an abuse of discretion.
We affirm in part and reverse in part. We conclude that the department applied the appropriate reasonableness standard in finding that the utilities violated their duty to restore service in a safe and reasonably prompt manner. We also conclude that the department’s over-all findings regarding National Grid and WMEC were supported by substantial evidence, as were its findings regarding the deficiencies of NSTAR’s communication with municipal officials and the general public, but that its finding that NSTAR failed timely to respond to priority two and three wires-down calls was not supported by substantial evidence. We, therefore, vacate the penalties the department imposed on NSTAR that were based in part on this unsubstantiated finding, and remand to the department for the imposition of penalties that reflect the more limited scope of its factually supported findings on this subject. Finally, as to the remaining penalties, we conclude that the department made the necessary subsidiary findings and, with two exceptions, did not abuse its discretion in its imposition of monetary penalties. As to those exceptions, we reverse and vacate the monetary penalties imposed against National Grid for its damage assessment performance during the last two days of the Irene restoration period and for its acquisition and deployment of resources during the last two days of the October snowstorm restoration period, because there was not substantial evidence supporting a violation on those days.
Background. 1. Statutory background. The department is tasked with the “general supervision of all . . . electric companies,” G. L. c. 164, § 76, and, consistent with that authority, has evaluated utilities’ performance in restoring power during and after major storms for at least the last three decades. See, e.g., Fitchburg Gas & Elec. Light Co., D.P.U. 09-01-A (2009) (Unitil); Eastern Edison Co., D.P.U. 85-232 (1986). In fulfilling its oversight responsibilities, the department has long declared that Massachusetts utilities have an obligation to restore service in a safe and timely manner when electric service has been disrupted by a major storm. See, e.g., Unitil, supra (2008 winter storm); Western Mass. Elec. Co., D.P.U. 95-86 (1995) (severe wind storm); Easte*556rn Edison Co., supra (Hurricane Gloria). After a December, 2008, winter storm in which all of Fitchburg Gas and Electric Light Company’s customers lost power, some for up to two weeks, the department reviewed the utility’s response and concluded that the utility’s poor performance in restoring service to customers in an effective and timely manner warranted monetary penalties, but noted that it lacked the power to impose them. Unitil, supra at 20, 91, 181.
That changed on November 12, 2009, just ten days after the issuance of the Unitil decision, when the Legislature enacted an “Act Relative to Public Utility Companies,” St. 2009, c. 133 (act). The act directed the department to “promulgate rules and regulations to establish standards of acceptable performance for emergency preparation and restoration of service for electric . . . companies doing business in the commonwealth,” and provided that the department “shall levy a penalty” against any company found to have violated those standards. G. L. c. 164, § 1J. The department was authorized to impose a penalty of up to $250,000 per violation per day, provided that the total for “any related series of violations” did not exceed $20 million. Id. The act also required utility companies annually to submit emergency response plans (ERPs), subject to the department’s review and approval, “designed for the reasonably prompt restoration of service in the case of an emergency event.”4 G. L. c. 164, § 85B. The ERPs were required to include the following:5 (1) the names of “management staff responsible for company operations during an emergency”; (2) “a communications system with customers during an emergency extended beyond normal business hours and conditions”; (3) contact with those who have a medical need “for essential electricity,” including but not limited to elderly and physically disabled individuals; (4) designation of staff assigned to communicate with local officials and relevant regulatory agencies; (5) provisions designed to ensure the safety of a company’s employees and contractors; (6) “procedures for deploying com*557pany and mutual aid crews to work assignment areas”; and (7) identification of and procedures for obtaining “additional supplies and equipment needed during an emergency.” G. L. c. 164, § 85B (a) (l)-(7), inserted by St. 2009, c. 133, § 5.6 See 220 Code Mass. Regs. § 19.05(1), (3) (2010).
After passage of the act, the department promulgated regulations providing performance standards for emergency preparation, restoration of service, and reporting. 220 Code Mass. Regs. § 19.03 (2010). Section 19.03(2) establishes a utility company’s duty to prepare for an emergency event: “Each Company shall ensure that it is adequately and sufficiently prepared to restore service to its customers in a safe and reasonably prompt manner during an Emergency Event.” Section 19.03(3) establishes a utility company’s duty to restore service during an emergency event:
“Each Company shall restore service to its customers in a safe and reasonably prompt manner during all Service Interruptions and outages. During an Emergency Event, this shall include at a minimum, but not be limited to, implementing all applicable components of the Company’s ERP related to restoration of service.”
The department may open an investigation into a utility company’s performance during an emergency event on its own initiative or at the request of the Attorney General or an affected municipality. 220 Code Mass. Regs. § 19.05(1) (2010). Where the department finds that a company violated any of these performance standards, it may impose a penalty within the parameters set forth in G. L. c. 164, § 1J.7
2. Factual background. On August 28, 2011, Irene hit the New England area. At least four inches of rain fell on the region, with wind gusts reaching sixty-seven miles per hour. NSTAR customers experienced widespread power outages through September 2; National Grid customers suffered such outages through Septem*558her 3. Two months after Irene, an October snowstorm dropped one foot of snow throughout much of the Commonwealth and up to thirty-two inches in western Massachusetts. As with Irene, the snowstorm resulted in widespread outages beginning on October 29 and lasting until November 3 for NSTAR customers and November 6 for National Grid and WMEC customers.
The department opened separate investigations regarding the performance of each of these utility companies, with the investigation of NSTAR and National Grid focused on both Irene and the October snowstorm, and the investigation of WMEC focused solely on the October snowstorm. The department’s investigations included dozens of records requests, submission of written testimony by various witnesses, sixteen public hearings throughout the utilities’ territories, and several days of evidentiary hearings. On December 11, 2012, the department issued its decisions (orders) in each of the proceedings, rejecting the majority of the Attorney General’s allegations but finding that each of these three utilities had — to varying degrees — violated its obligation to “restore service to its customers in a safe and reasonably prompt manner” during these emergency events, in violation of 220 Code Mass. Regs. § 19.03(3).
For these violations, the department levied a penalty of $18,725 million against National Grid, $4,075 million against NSTAR, and $2 million against WMEC. The utilities each filed an appeal in the county court pursuant to G. L. c. 25, § 5.8 A single justice reserved and reported the cases to the full court, and they were joined for oral argument.
Discussion. Our standard of review when considering an appeal pursuant to 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.... 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. . . . 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 [the department’s] decision *559unless it is based 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).”
Bay State Gas Co. v. Department of Pub. Utils., 459 Mass. 807, 813-814 (2011), quoting DSCI Corp. v. Department of Telecomm. & Energy, 449 Mass. 597, 603 (2007).
1. Prudence versus reasonableness standard. We first address the utilities’ contention that the department erred as a matter of law in failing to apply the proper standard in its evaluation of each company’s storm performances. The utilities argue that the department should have applied the prudence standard; the department contends that it appropriately applied the regulatory standard of whether the utility restored service to its customers “in a safe and reasonably prompt manner,” which we shall characterize as a reasonableness standard.
Under the prudence standard, “the department determines whether a utility’s actions, based on all that it knew or should have known at the time, were reasonable and prudent in light of the circumstances which then existed.” Fitchburg Gas & Elec. Light Co. v. Department of Pub. Utils., 460 Mass. 800, 802-803 (2011). Although this sounds like a reasonableness standard, it differs from the reasonableness standard applied by the department in two fundamental ways.
First, under the prudence standard, it is not “appropriate for the department merely to substitute its own judgment for the judgments made by the management of the utility.” Id. at 803. See Attorney Gen. v. Department of Pub. Utils., 390 Mass. 208, 229 (1983) (“On the issue of prudence, the department had no authority to substitute its judgment for the reasonably exercised prerogatives of [the utility’s] business managers”). The reasonableness standard applied by the department does not grant such deference to the judgment of utility management.
Second, a utility satisfies the prudence standard where it acts in conformance with “fair and prevailing utility practice.” See Boston Gas Co. v. Department of Pub. Utils., 359 Mass. 292, 301 (1971). Under the reasonableness standard, a practice followed by every utility may still be unreasonable where it fails adequately to restore service following a storm in a safe and reasonably prompt manner. Cf. Restatement (Third) of Torts § 13(a) *560(2010) (compliance with industry standards may be “evidence that the actor’s conduct is not negligent but does not preclude a finding of negligence”).
We agree with the department that it correctly applied the reasonableness standard. The prudence standard is limited to rate setting, where utilities may “charge rates which are compensatory of the full cost incurred by efficient management, [but] may not recover costs which are excessive, unwarranted, or incurred in bad faith.” Fitchburg Gas & Elec. Light Co., 460 Mass. at 802, quoting Boston Gas Co. v. Department of Pub. Utils., 387 Mass. 531, 539 (1982). See Fitchburg Gas & Elec. Light Co., supra (“For supply costs to be recovered, the expenditures must have been reasonably and prudently incurred”); Hingham v. Department of Telecomm. & Energy, 433 Mass. 198, 202 (2001); Boston Gas Co. v. Department of Pub. Utils., 367 Mass. 92, 97 (1975) (noting “long accepted and often repeated principle that a regulated public utility ... is entitled to charge rates which afford it the opportunity to meet its cost of service, including a fair and reasonable return on honestly and prudently invested capital”).
There are several reasons why prudence would be an inappropriate standard to evaluate a utility’s storm performance. The Legislature is familiar with the prudence standard and knows how to direct the department to apply it in the regulation of public utilities. See, e.g., G. L. c. 164, § 94G (d) (rate recovery permitted where “department finds by clear and convincing evidence after a public hearing that [certain] increased expenditures were [not] incurred as a result of company imprudence”). General Laws c. 164, § 1J, makes no reference to the prudence standard.
Nor is there any indication that the Legislature intended the department to apply such a standard with respect to the restoration of electricity service after an emergency event. The absence of any such indication is not surprising given that the act was adopted in the wake of the 2008 winter storm, where the utility’s restoration performance had been deficient in several respects, and the Legislature’s intent was to empower the department in the future to hold utility companies accountable for failing reasonably to restore power after outages arising from major storms. When debating this legislation, members of the Legislature described the power outages following the 2008 winter storm as a “nightmare,” and charged that Unitil was not prepared for the storm, performed poorly responding to the outages, and was not *561“accessible to its customers and local officials.”9 State House News Service (House Sess.), June 3, 2009. According to Representative Robert Rice, the bill would “put[ ] a knife over the heads of the utilities]” and “give[ ] [the department] the muscle and teeth that was previously lacking.” Id. Representative Stephen DiNatale suggested that the “legislation [would] address ... the glaring deficiencies” in the utility’s performance. Id. And Representative Barry Finegold said it would “ensure if something like this happened] again, [the department] [would] have to respond.” Id.
Applying the prudence standard would run counter to the Legislature’s purpose where it clearly intended that the department adopt and enforce its own standards of what constitutes reasonable storm performance, not the utility industry’s standards. The prudence standard would subordinate the department’s own experience and expertise to the prevailing practices in the industry it is regulating, and require deference to the prerogatives of utility management. That the department, in establishing such standards, was informed by industry practice, historical practices, and the companies’ ERPs does not suggest that it intended to apply a prudence, rather than a reasonableness, standard, especially where the department was also informed by its own considerable precedents in evaluating storm responses.
Moreover, applying the prudence standard ignores the inherent differences between analyzing whether a company is eligible for rate recovery and whether it restored power after an emergency event in a safe and reasonably prompt manner. The first analysis determines whether ratepayers or shareholders will bear the burden of paying for certain investments and expenditures; the other determines whether a company fulfilled its obligation to consumers and the general public to restore service where a major storm or other event produces massive power outages. Where a company during an emergency event must respond to priority “wires-down” calls or restore service to critical facilities, among other *562responsibilities, the consequences of any deficiency in the company’s performance are potentially catastrophic. In such a context, it was logical for the department to impose a higher standard on the utilities than simply determining whether, from a business perspective, the companies acted prudently.
The utilities also argue that the reasonableness standard used by the department is “vague” and “unascertainable,” and that the Legislature could not have intended such a standard to be used by the department.10 Reasonableness is plainly a general standard, but it needs to be because storm performance evaluations are inherently fact-specific. See Brookline v. Commissioner of the Dep’t of Envt’l Quality Eng’g, 387 Mass. 372, 378 (1982), quoting Boyce Motor Lines, Inc. v. United States, 342 U.S. 337, 340 (1952) (“the practical necessities of discharging the business of government inevitably limit the specificity with which [a regulatory agency] can spell out prohibitions”). The department’s standards, however, do not arise in a vacuum; the department’s prior storm performance evaluations help establish the standard for reasonable conduct in restoring power after an emergency event. See, e.g., Unitil, D.RU. 09-01-A (2009); Western Mass. Elec. Co., D.P.U. 95-86 (1995); In re Hurricane Bob, D.P.U. 91-228 (1992); Eastern Edison Co., D.P.U. 85-232 (1986). Several of these evaluations involved issues similar to the ones present here. For example, in previous storm performance evaluations, the department has emphasized the importance of (1) acquiring additional damage assessors, crews, and other resources before a storm hits, see Unitil, supra at 47, 69; In re Hurricane Bob, supra at 22; (2) performing a thorough damage assessment and providing estimated restoration times based in part on those assessments, see Unitil, supra at 68-72; Western Mass. Elec. Co., supra at 43; and (3) communicating effectively with local officials and the general public, among other restoration responsibilities. See Unitil, supra at 9-17; Eastern Edison Co., supra at 16-22.
In summary, the Legislature intended to enable the department to hold public utilities accountable for their storm performance, a purpose that would have been frustrated had the department’s standard of care been defined entirely by industry practice or a company’s ERR rather than by the department itself. Conse*563quently, the department did not make an error of law in applying the reasonableness standard to determine whether the utilities restored power in a safe and reasonably prompt manner.
2. Substantial evidence. The utilities also argue that the department’s findings were not supported by substantial evidence. Substantial evidence is “such evidence as a reasonable mind might accept as adequate to support a conclusion.” Boston Gas Co. v. Assessors of Boston, 458 Mass. 715, 721 (2011), quoting Tennessee Gas Pipeline Co. v. Assessors of Agawam, 428 Mass. 261, 262 (1998). Our consideration of the substantiality of the evidence is not limited to the evidence relied on by the department; it also “must take into account whatever in the record fairly detracts from its weight.” Boston Gas Co., supra, quoting New Boston Garden Corp. v. Assessors of Boston, 383 Mass. 456, 466 (1981). Nevertheless, “[w]here there is substantial evidence to support the [department’s] decision, we defer to the [department’s] judgment as to what evidence to accept,” Boston Gas Co., supra, quoting General Elec. Co. v. Assessors of Lynn, 393 Mass. 591, 608 (1984), and give “due weight to the experience, technical competence, and specialized knowledge of the [department], as well as to the discretionary authority conferred upon it.” G. L. c. 30A, § 14 (7) (g). But, “if ‘the evidence points to no felt or appreciable probability of the conclusion or points to an overwhelming probability of the contrary,’ ” we will set aside the department’s finding. Boston Gas Co., supra at 721-722, quoting Tennessee Gas Pipeline Co., supra.
a. National Grid. The department found National Grid’s storm performances to be substandard in six areas: (1) acquisition and deployment of company resources; (2) damage assessment; (3) response to priority wires-down calls; (4) performance of the outage management system; (5) communication with public officials; and (6) communication with the general public.11 We address whether there was substantial evidence to support each finding in turn.
i. Acquisition and deployment of company resources. With respect to the acquisition and deployment of personnel during *564Irene, the department found that National Grid failed to secure the number of crews it anticipated needing in its ERP for a level IV or V emergency event. When Irene threatened the region, National Grid prepared for a level V event, and Irene turned out to be one, causing over 478,000 customer outages.12 National Grid’s ERP anticipated that, in a level IV or V event, it would need all internal line crews, 500 or more contractor or foreign utility overhead line crews, and 500 or more tree crews. On August 28, the beginning of the restoration period, National Grid had secured only 233 contractor line crews and 286 tree crews, far fewer than the minimum it needed. Although National Grid had made a request on August 26 for 200 additional distribution line crews through the mutual aid process,13 it ultimately secured only thirty-seven line crews through this process, and those crews did not arrive until September 2, when the restoration process was already well underway.
National Grid classified the October snowstorm as a level III event on October 28, then a level IV event on October 29, and finally a level V event on October 30.14 But when the storm ended on October 30, it had mobilized fewer contractor line crews — thirty-two — than the sixty its ERP anticipated would be needed for a level III event. By that date, it had also mobilized 232 tree crews, more than the sixty for a level III event, but fewer than the 500 needed for a level V event. It eventually mobilized more than the number of contractor line crews and nearly the number of tree crews required by its ERP during the October snowstorm, but not until late in the restoration process.
The department recognized that the crew numbers in the ERP were not mandatory but were “an indication of the numbers of crews that should be obtained as part of a reasonable response effort and a deviation from these numbers requires justification.” For both Irene and the October snowstorm, the department found that National Grid did not provide an adequate justification for its failure to secure the necessary additional crews earlier in the restoration process.
*565National Grid argued that the crew allocation estimates in its ERP were “subject to market availability,” and that, despite its best efforts, additional crews were not available due to the “broad geographic impact” of Irene and the severity of the October snowstorm. The department acknowledged that “the [cjompany attempted to secure more resources from the mutual aid process and from contractors before the storm hit.” But the department concluded that National Grid had not moved early enough to acquire the necessary crews, and that it was unreasonable for the company to “expect that mutual aid crews [would] arrive at the beginning of the restoration effort rather than towards the end.”
The department also found that those crews the company did acquire were not deployed or redeployed in a reasonable manner. For example, the department found that during Irene the company deployed crews to the North Shore district while removing crews from the Southeast district even though the ratio of crews to outages was far greater in the latter than the former at that time. National Grid argues that the ratio of deployed crews to outages is a “contrived metric” in that it does not account for other factors that influence the deployment of company resources, including the need to address priority wires-down calls. Because the Legislature delegated the authority to adopt performance standards to the department, we defer to its expertise regarding whether the ratio of outages to crews is a relevant storm performance metric. See Bay State Gas Co. v. Department of Pub. Utils., 459 Mass. 807, 813-814 (2011), quoting DSCI Corp. v. Department of Telecomm. & Energy, 449 Mass. 597, 603 (2007) (“we give deference to the department’s expertise and experience in areas where the Legislature has delegated to it decision-making authority”). We conclude that there was substantial evidence to support the department’s finding that National Grid’s acquisition and deployment of crews was unreasonable during both storm events.15
ii. Damage assessment. The department also found that National Grid failed to conduct an effective damage assessment following both storms. Both the company’s ERP and the department’s ERP guidelines require that phase I surveys — which assess the functioning of a company’s main power lines — be completed within twenty-four hours of an emergency event, and *566that phase II surveys, which involve more localized analysis of outages in specific residential neighborhoods, be completed within forty-eight hours. The department found that after Irene, National Grid took at least two days to complete its phase I damage assessments, and did not even begin phase II assessments until after the restoration period was already forty-eight hours old. Following the October snowstorm, National Grid completed phase I assessments within thirty-six hours, but failed to complete its phase II assessments. The department determined that National Grid’s deviations from its ERP and the department’s guidelines were unjustified. In fact, the department found that National Grid failed to “pre-position” damage assessors, reassigned some damage assessors to perform other restoration functions, and used a damage assessment procedure that was manually intensive and inefficient.
National Grid argued that it was only required to begin, not complete, its damage assessments within the applicable twenty-four and forty-eight hour time periods. The department rejected this argument, stating that “[t]he language in both the ERP [guidelines and the [cjompany’s ERP indicates that the [pjhase I and [pjhase II Surveys for damage assessment are expected to be completed within the [twenty-four]- and [forty-eight]-hour time-frames.” Having examined the language in both the ERP guidelines and the company’s ERP, and giving due deference to the department’s interpretation of that language, we agree that a damage assessment was required to be completed, not merely commenced, within those time frames. Therefore, we conclude that there was substantial evidence to support the department’s finding that National Grid’s delay in conducting damage assessments violated its obligation to act reasonably to restore service after an emergency event.16
iii. Response to priority wires-down calls. Utility companies receive reports of downed wires from customers as well as municipal and public safety officials. The latter reports are given priority, and officials are trained to categorize them based on the seriousness of the situation. A priority one call means the reporting official has classified the situation as “life threatening” or a source of “imminent danger.” A priority two call occurs where a downed wire is impeding emergency operations, and a priority *567three call means the downed wire poses a nonthreatening electrical hazard. In an emergency event, the company deploys the next available crew or nearest trained resource to respond to a priority one call and the company provides the crew’s estimated time of arrival (ETA) to the reporting official. Priority two and three calls are logged in the company’s outage management system (OMS), which prioritizes the calls, assigns a crew to respond to the situation, and provides the official with an ETA.
The department found that National Grid took an average of 22.6 hours to dispatch crews to address priority wires-down calls during Irene; during the snowstorm, that average jumped to forty-six hours.17 National Grid disputes these figures. It argues that the department’s calculations are not based on when the first wires-down personnel arrived on the scene, but on the arrival of the last crew, because the company’s OMS overrode data on when company personnel first responded to a wires-down event with the time a later crew arrived. Therefore, where the company sent personnel to relieve emergency personnel from guarding a potentially live wire, and a repair crew arrived hours later to remedy the situation, the OMS would reflect the time of arrival of the repair crew, not the relief crew, thereby erroneously inflating the company’s response times.
We agree with the department’s rejection of National Grid’s argument that the department cannot rely on the company’s data regarding the initial response time for priority calls because the company failed to preserve the initial response time in its OMS, thereby rendering its data unreliable. The department was well warranted in noting that the company’s “inability to provide reliable evidence ... raises questions as to the [cjompany’s entire wires-down process.” The department found that, during Irene, “the [c]ompany’s files were missing both dispatch and arrival times for nearly 8,700 wires-down calls,” and “the [cjompany was unable to provide any records of ETA callbacks for [eighty-six] percent of the 1,148 priority wires-down calls.” During the October snowstorm, the company “did not provide records of ETAs for approximately [eighty-eight] percent of priority wires-down calls.”
Where National Grid’s own data demonstrates unreasonably *568slow response times to priority wires-down calls, and where the company’s defense to its delay is that the data it kept on response times was unreliable, we conclude that there was substantial evidence to support the department’s finding that the company’s wires-down response was unreasonable.
iv. Performance of OMS. The department found that National Grid’s OMS “did not effectively process and manage outage information” during Irene, that its performance improved only incrementally during the October snowstorm, and that its subpar performance during these storms contributed to the company’s failure to restore service in a safe and reasonable manner. Beginning at 10 a.m. on August 28, the day Irene reached Massachusetts, “[t]he [cjompany acknowledged 15.5 hours of slowness that users experienced,” a slowdown caused by the inability of the call analyzer to process calls at the rate it received them. To remedy this problem, the company needed to take OMS offline for approximately thirty minutes, during which no user was able to access the system. Because the OMS operated so slowly, National Grid had to track calls and orders manually. In short, National Grid had an OMS that had not been “volume tested in advance,” and proved to be unable to manage the number of calls that foreseeably would be received during a Level V emergency storm. The OMS performed better during the October snowstorm because of processing power and Web servers that were added after Irene, but users still experienced several periods of slow operation arising from call and work order volumes, which required the company to resort to manual tracking. In addition, there were periods during the snowstorm in which the system was not updating and “[c]ould not create, close, or assign new orders to reflect work assignments.”
The company contends that the OMS slowdown did not delay its restoration efforts and that there is no evidence to support the department’s assertion that its failures in this regard had a “cascading effect ... on the remaining restoration phases.” Where OMS receives outage information reported by customers and uses that information to guide its restoration efforts and crew deployment, and to generate estimated times of restoration (ETRs), we conclude that it was reasonable for the department to infer that the system’s slowdown (and the company’s need at times to process the information manually) unreasonably hindered its res*569toration efforts.18
v. Communication with public officials and the general public.19 The department found that National Grid failed effectively to communicate with public officials in the municipalities it served. Municipal officials from various cities and towns complained that during Irene, National Grid provided them with information that was vague, untimely, and sometimes inaccurate, and often failed to respond to information these officials had provided to the company. The poor communication impeded their ability to focus emergency management efforts in their communities during the tropical storm. National Grid’s communication issues continued during the October snowstorm, with local officials in various municipalities reporting difficulty coordinating with National Grid and obtaining credible and honest information. Local officials in numerous towns complained that they received inaccurate and untimely ETRs from National Grid, and that the company identified nursing homes, sewage treatment plants, and other key municipal facilities as “critical facilities” in its ERP but failed to make them restoration priorities.20 Specifically, the company failed to communicate with local officials regarding the restoration of power at nonhospital critical facilities even after town officials had asked National Grid to do so.
Further, during Irene the company was forced to shut down its outage and accident reporting notification protocol system (ORP), a subset of the OMS dedicated to informing municipalities about major service interruptions, because municipalities complained that the notices the company transmitted by facsimile were unhelpful and overwhelming. During the October snowstorm, the *570company had the opposite problem as the ORP did not provide notices for 239 qualifying events.
In addition, the department found that National Grid provided municipal liaisons to only forty-one of the 170 communities impacted by Irene; during the October snowstorm, that number jumped to ninety-three out of 158 affected communities. Although the company’s ERP did not require it to provide a municipal liaison to every impacted community, National Grid acknowledged the benefit of municipal liaisons, and the department properly could consider the company’s failure to provide more of them as indicative of a lack of effective communication. Even in localities where there were liaisons, communication still suffered because many liaisons did not know enough about the locality’s municipal government and its electric system to be helpful.
National Grid’s communication failures extended to its dealings with the general public, particularly regarding its provision of ETRs. During Irene, the department found that National Grid’s ETRs were inaccurate, in part because the company appeared to rely on global ETRs covering a large area rather than local ETRs focused on specific communities. National Grid acknowledged that, during the October snowstorm, the initial ETR it provided to “many customers” was not accurate, and that it failed to update the ETRs until after that estimated time “had come and gone.”
National Grid argues that the department improperly based its findings regarding communication failures on complaints made by municipal officials and members of the public that were inherently subjective and unreliable. The company suggests that these individuals were not unhappy with the messenger but with the message, because “not everyone accepted the enormity of the storm and the fact that the sheer volume of damage required a reasonable amount of time to restore power for everyone.” But there is no evidence to suggest the department afforded improper weight to the complaints lodged by municipal officials or members of the public, or that the complaints were consistently unreliable. Moreover, the complaints were not the only source of information the department considered in finding that National Grid had violated its duty to communicate effectively with municipalities and the public. The company’s inadequate provision of municipal liaisons and the liaisons’ lack of familiarity with the locality, the poor performance of its ORP notification system, and its practice of updating ETRs only after they had not been met were just some of the other deficiencies the department relied on *571in concluding that National Grid had violated its communication duties. We conclude that there was substantial evidence to support the department’s findings regarding the inadequacy of National Grid’s communication with public officials and the public, and its contribution to the company’s failure to restore service in a safe and reasonable manner.
b. NSTAR. The department determined that NSTAR had failed to restore service in a safe and reasonably prompt manner in three respects: (1) response to priority two and three wires-down calls; (2) communication with municipal officials; and (3) communication with the general public.21
i. Response to priority two and three wires-down calls. The department found that, with respect to priority one calls, NSTAR’s response was “reasonable and timely” during both Irene and the October snowstorm.22 But it also found that, with respect to all other priority calls, the company’s response times “were often too long” and did “not reflect the urgency such calls require[d].” NSTAR responded to eighty per cent of priority two calls and seventy-five per cent of priority three calls within twenty-four hours during Irene, and to ninety-four per cent of priority two calls and ninety-one per cent of priority three calls within twenty-four hours during the October snowstorm. Nevertheless, the department found that, in some instances, the time NSTAR took to respond to priority two calls (up to seventy-four hours for Irene and up to sixty-five hours for the October snowstorm) “suggests” that it violated a requirement in its ERP that it respond to priority two calls with the next available trained resource.
The record, however, does not support that suggestion, because *572NSTAR’s ERP provided that it would respond only to priority one calls with the next available trained resource; it did not commit in a level V event to provide the next available trained resource on a priority two or three call.23 The department noted that fire or police personnel needed to stand by downed wires during all priority events until the company responded, and found that such personnel sometimes waited for more than twenty-four hours during each storm for the company to respond to priority two and three calls. But it rejected the argument that twenty-four hours is the outside limit for a reasonable response time to all priority two and three calls, and that any priority wires-down response over twenty-four hours violates the company’s obligation to restore power in a safe and reasonably prompt manner. Where the department found that NSTAR timely had responded to priority one calls and had responded to the vast majority of priority two and three calls in less than twenty-four hours, where there was no commitment in the ERP to provide the next available resource to priority two and three calls, and where the department recognized that a response longer than twenty-four hours may be reasonable during a level V event, we conclude that reliance on a few outlying response times does not constitute substantial evidence supporting the department’s finding that “NSTAR failed to respond to [priority] wires-down calls in a timely manner.”
The department also found that NSTAR “failed to communicate effectively with municipal officials regarding [priority] calls,” and that this failure “compromised public safety.” This particular finding regarding priority calls was supported by substantial evidence. Based on the testimony of municipal officials, the department found that the company at times failed to coordinate its priority wires-down response with municipalities, and failed to provide estimates to local officials regarding when company personnel would arrive on the scene and relieve municipal personnel. Several municipal officials testified to the difficulty they had in obtaining arrival time estimates from NSTAR despite specific requests for them, and that fire fighters and police officers spent long periods of time guarding downed wires while waiting for company personnel to arrive.
*573The department assessed a penalty of $2 million against NSTAR regarding its response to priority calls: $250,000 per day for each of the four days it took to respond to priority wires-down calls for Irene, and for each of the four days it took to respond to priority calls for the October snowstorm. Where we conclude that only one of the department’s two key findings regarding the response to priority calls — the company’s failure to effectively coordinate with and communicate response times to municipal officials — was supported by substantial evidence, we vacate these penalties and remand to the department for the imposition of penalties that reflect the more limited scope of its factually supported findings on this subject.
ii. Communication with municipal officials. The department found that NSTAR’s “communication with municipal officials was at times ineffective, resulting in inaccurate or incomplete information being disseminated,” and that the ineffective communication had “a negative impact on the restoration efforts.” The department heard testimony and received letters from numerous municipal officials asserting that they could not reach the company, and when they did, they sometimes received no information or inaccurate information. Many municipal officials were often unable to obtain ETRs, and when they did, the ETRs either changed or were inaccurate.
NSTAR contends that the department improperly relied on written and oral complaints from municipal officials that were not subject to cross-examination. We disagree. In an adjudicatory proceeding under G. L. c. 30A, an agency “need not observe the rules of evidence observed by courts.” G. L. c. 30A, § 11 (2). See 220 Code Mass. Regs. § 1.01(3) (2008) (“Adjudicatory Proceeding shall be as defined in [G. L.] c. 30A, § 1 [1]”). But, under the department’s regulations, it “shall follow the rules of evidence observed by courts when practicable” (emphasis added). 220 Code Mass. Regs. § 1.10(1) (2008). The department is entitled to deem it impracticable to observe the rules of evidence where it conducts public hearings as part of its investigation, and where it considers letters submitted to the department, provided such letters are made part of the record in the proceeding. 220 Code Mass. Regs. § 1.10(3) (2008) (“Any matter contained in any records, investigations, reports, and documents in the possession of the [department of which a party or the [djepartment desires to avail itself as evidence in making a decision, shall be offered and made a part of the record in the proceeding”). Therefore, not all *574complaints about a utility’s performance in the record need be subject to cross-examination. The department’s regulations also provide that “[a]ll unsworn statements appearing in the record shall not be considered as evidence on which a decision may be based.” 220 Code Mass. Regs. § 1.10(1). Even though the department referred to unsworn statements in its orders, it is clear from the other evidence in the record that its decisions did not rest on unsworn statements.24
NSTAR also contends that, with respect to the testimony of three municipal officials who were cross-examined, the department did not consider testimony elicited during that cross-examination, as evidenced by its omission from the department’s 147-page decision. We are not persuaded. The department is not obliged to reference every part of every witness’s testimony in its final decision, and did acknowledge the company’s arguments regarding the credibility of such testimony. Accordingly, we conclude that the department’s finding was supported by substantial evidence.
iii. Communication with the general public. The department found that many customers complained that they were unable to reach NSTAR, unable to obtain outage and restoration information, and unable to obtain ETRs. It also found that, during the October snowstorm, NSTAR erroneously notified customers that their service had been restored. We conclude that there was substantial evidence to support the department’s finding that NSTAR “failed to communicate effectively with customers” during both storm events.
c. WMEC. The department assessed WMEC’s performance in thirteen different categories and found it to be unreasonable in only one — responding to priority wires-down calls during the October snowstorm. Specifically, the department found WMEC took an average of twenty-two hours to respond to priority one calls, 35.4 hours for priority two calls, and 41.8 hours for priority three calls, and that these average response times were “unreasonably long.”25 The department also highlighted several particularly egregious examples of WMEC’s failure to respond to dan*575gerous wires-down situations in a timely manner, including one priority call reporting that a power line had melted through a sidewalk and posed a danger of setting fire to a nearby apartment building, to which WMEC took a full day to respond, and another priority call where a power line had fallen on a motor vehicle and, despite assurances from WMEC that a crew was on its way, the company never responded. The department concluded that WMEC’s “unreasonably long response times” for resolving priority calls raised “significant public safety concerns,” and did not comply with the standard to restore service in a safe and reasonably prompt manner. We conclude that the department’s finding was supported by substantial evidence.
3. Penalties. For the violations discussed above, the department fined National Grid $18,725 million,26 NSTAR $4,075 million,27 and WMEC $2 million.28 The utilities challenge these penalties *576on the grounds that the department did not discuss the penalty factors set forth in 220 Code Mass. Regs. § 19.05(2) (2010), and did not explain adequately the amount or duration of the penalties it imposed.
Where an administrative agency imposes a penalty it is authorized to enforce, “neither a trial court nor an appellate court is free to substitute its own discretion as to the matter; nor can the reviewing court interfere with the imposition of a penalty by an administrative tribunal because in the court’s own evaluation of the circumstances the penalty appears to be too harsh.” Vaspourakan, Ltd. v. Alcoholic Beverages Control Comm’n, 401 Mass. 347, 355 (1987), quoting Levy v. Board of Registration & Discipline in Med., 378 Mass. 519, 529 (1979). We “will interfere with the agency’s discretion in this area ‘only ... in the most extraordinary of circumstances.’ ” Vaspourakan, supra, quoting Levy, supra at 528-529.
The Legislature afforded the department broad discretion regarding the amount of any penalty imposed on a utility company for violating its storm performance standards, provided that the penalty did not exceed $250,000 per violation per day, or $20 million for all related violations. G. L. c. 164, § 1J. In imposing such penalties, the department considers, among other factors: (a) the gravity of the violation; (b) the appropriateness of the penalty to the size of the company; (c) the company’s good faith in attempting to comply with the department’s standards; and (d) the degree of control that the company had over the circumstances that led to the violation. 220 Code Mass. Regs. § 19.05(2).
As an initial matter, although its orders must be detailed enough to permit meaningful judicial review, Costello v. Department of Pub. Utils., 391 Mass. 527, 538 (1984), the department is not required specifically to discuss the penalty factors in its order so long as the order demonstrates that the department has duly considered them. Here, there is ample evidence that the department did so.
The department repeatedly framed its discussion of the violations both in terms of their importance to public health and safety and the extent of a utility’s violations. With respect to a company’s size, the department noted each company’s number of customers and communities served, and the record reflected the amount of money each utility spent on storm restoration, among other expenditures, as well as the number of people each company employed, in order to place the amount of the penalty im*577posed within the context of the company’s over-all size. For example, the penalty imposed on National Grid represented less than sixteen per cent of what the company spent on storm restoration, and was roughly nine times the salary of the company’s highest paid executive.
As to a utility’s good faith, the department recognized the size and severity of both storms, and the significant challenges the storms presented, including the large number of downed wires and the damage to company equipment and systems. See, e.g., Western Mass. Elec. Co., D.P.U. 11-119-C, at 70 (2012) (noting company received more wires-down calls during first day of snowstorm than it had received over previous twelve months). In each order, the department balanced those considerations against the nature and extent of the violations. See, e.g., NSTAR Elec. Co., D.P.U. 11-85-B/11-119-B, at 119 (2012) (imposing $100,000 penalty per day — $150,000 less than daily maximum — for communication failures during October snowstorm in light of company’s efforts to improve communications after problems experienced during Irene).
Finally, the department’s discussion regarding a company’s violations focused on the company’s degree of control over the violation. For example, the department noted that National Grid’s ineffective damage assessment was based on choices made by company management, and that, even though the company knew its OMS could not detect nested outages prior to Irene, it declined to modify or upgrade the system before either storm.
Accordingly, although the department did not individually discuss each of the factors set forth in 220 Code Mass. Regs. § 19.05(2), in the penalty sections of its orders, it did consider the factors in imposing the penalties. We therefore decline to disrupt the amount of the penalties imposed on this basis.
We now turn to the utilities’ second argument: that in imposing penalties each day until power was fully restored in their respective areas, the department simply assumed that violations had occurred on those days rather than identifying specific violations that had occurred on each day for which the utilities were penalized.
In reviewing whether the violations at issue here were supported by substantial evidence, we must consider not just whether a departmental standard was violated, but also how long the violation occurred. G. L. c. 164, § 1J (“department shall levy a penalty ... for each violation for each day that the violation of the *578department’s standards persists”). The department need not make detailed findings as to each day on which a violation occurred so long as there is substantial evidence a violation persisted for the duration of the period for which a penalty was imposed. See Costello, 391 Mass. at 538 (“Although an agency must make all findings necessary to its decision [G. L. c. 30A, § 11 (8)], it need not make detailed findings of all evidence presented to it, as long as its findings are sufficiently specific to allow us to review the department’s decisions”).
For the penalties imposed against NSTAR and WMEC, and most of those imposed against National Grid, the evidence was sufficient to support a finding that the applicable violations persisted for the duration of the time period for which they were imposed. There are a few instances, however, in which the department’s findings that a violation continued for a certain period of time fall short. We discuss those violations below, and overturn the portion of the penalties for those violations where the evidence was insufficient for us to infer that a violation persisted for the duration of the penalty period.
With respect to National Grid’s acquisition and deployment of resources, the evidence was sufficient to support the department’s findings for Irene, for which National Grid never acquired the number of crews set forth in its ERP and did not justify its deviation from those figures. During the last two days of the restoration period of the October snowstorm, however, the company had acquired more crews than provided for in its ERP. Although a company’s ERP establishes only its baseline storm performance obligations — such that a company may be in compliance with its ERP but still violate the department’s standards — the department must offer at least some evidence to suggest a particular violation occurred where the company has otherwise complied with the applicable section of its ERP, especially where it relied on the shortfall from the ERP crew number in earlier finding a violation. Nor are we aware of any evidence in the record demonstrating that those crews were deployed in an unreasonable manner during the last two days of the restoration period.29 Accordingly, we conclude, with respect to the last two days of the snowstorm restoration period, that the department’s *579finding that National Grid violated its duty to acquire sufficient resources and deploy them efficiently was not supported by substantial evidence, and we vacate the penalties — $250,000 per day — imposed for those two days.
The department’s findings regarding National Grid’s damage assessment performance were similarly inadequate to support imposing a penalty for the duration of Irene. The department found the company’s damage assessment to be deficient because its phase I and phase II assessments were not completed within the twenty-four and forty-eight hour periods, respectively, required by National Grid’s ERP and the department’s ERP guidelines. Specifically, National Grid began its phase I survey on August 28 — the first day of the restoration period — and completed it on August 30; its phase II survey did not commence until August 30 and was completed on August 31. But the department fined National Grid $200,000 per day for six days for its poor damage assessment during that storm, and not just for the four days (August 28 to August 31) between the beginning and end of the company’s damage assessment. The department could not continue to penalize National Grid for its damage assessment performance after the assessments were completed absent evidence the company’s performance in this area continued to be deficient. Because National Grid completed its damage assessment on the fourth day of the restoration period, and because the department cited no evidence that a violation persisted beyond that day, we conclude that there was not substantial evidence that this violation of the department’s standards persisted during the last two days of the Irene restoration period. Those penalties — $200,000 per day for two days — must be vacated.
Conclusion. For the reasons stated above, we remand the cases to the single justice for entry of the following orders. The department’s order regarding National Grid is affirmed, but the penalty imposed is reduced by $900,000 to $17,825 million. With respect to NSTAR, the department’s order is affirmed to the extent it imposed penalties totaling $2,075 million for NSTAR’s unreasonable conduct in communicating with municipal officials and the general public. Given our conclusion that the department’s finding that NSTAR failed timely to respond to priority two and three wires-down calls was not supported by substantial evidence, *580this finding is reversed and the $2 million in penalties that were based in part on this unsubstantiated finding is vacated. The case is remanded to the department for the imposition of penalties that reflects the more limited scope of its factually supported findings on this subject. The department’s order regarding WMEC is affirmed.
So ordered.
An emergency event occurs where a storm or other event beyond a utility company’s control causes widespread outages or service interruptions. A utility company may classify — using procedures outlined in its emergency response plan (ERP) — an event as anywhere from level I through V based on the expected number of service interruptions and their estimated duration. A level III, IV, or V event is considered an emergency event.
General Laws c. 164, § 85B (a), was subsequently amended in 2012. See St. 2012, c. 216, § 4. Because the storms at issue occurred in 2011, the revisions in the statute made in 2012 do not apply.
If a company failed “to implement its [ERP],” and that failure resulted in power outages “materially longer than they would have been but for the company’s failure,” the Department of Public Utilities (department) was authorized to prevent the company from recovering all or part of its service restoration costs through the rates it charges customers. G. L. c. 164, § 85B (d.). Rate recovery, however, is not at issue in these cases.
Any penalty levied by the department for a violation of the department’s storm performance standards is to “be credited back to the company’s customers in a manner determined by the department.” G. L. c. 164, § IK.
General Laws c. 25, § 5, provides, in pertinent part, that “[a]n appeal as to matters of law from any final decision, order or ruling of the [department] may be taken to the supreme judicial court by an aggrieved party in interest.”
The department concluded that, among other performance deficiencies, Fitch-burg Gas and Electric Light Company had failed to complete a proper damage assessment in a timely manner, to acquire sufficient repair crews during the storm, or to communicate effectively with the public or local officials. Fitchburg Gas & Elec. Light Co., D.P.U. 09-01-A (2009) (Unitil). Accordingly, the department found that the company’s poor performance in these areas “represent[ed] a failure to satisfy its obligation to provide safe and reliable electric service to its customers” and a failure to restore power “in an efficient, effective, and timely manner.” Id. at 72, 84, 102.
The three utility companies (utilities) have not challenged the facial constitutionality of G. L. c. 164, § 1J, or 220 Code Mass. Regs. § 19.03(3) (2010).
The department concluded that the storm performance of Massachusetts Electric Company and Nantucket Electric Company, each doing business as National Grid (collectively, National Grid), was reasonable with respect to several other categories, including (1) weather forecasting and event classification; (2) communication with life support customers; (3) advance planning and training; (4) vegetation management; and (5) reporting.
With a level V event classification, National Grid expected customer outages of 113,700 or more, which could extend to all 1.2 million customers in its service territory.
The mutual aid process facilitates the sharing of crews and resources among participating utility companies.
The department found these classifications to be reasonable “in the face of the rapidly escalating weather forecasts.”
We address later in the opinion whether there was substantial evidence to warrant penalties on each of the days for which penalties were assessed.
We address later in the opinion whether there was substantial evidence to warrant penalties on each of the days for which penalties were assessed.
National Grid did not provide the priority level for thirty-nine per cent of the priority calls during Tropical Storm Irene (Irene), and for thirty-seven per cent of the priority calls during the October 29, 2011, snowstorm (October snowstorm).
The department also found, and the company acknowledged, that the outage management system was incapable of detecting nested outages, which occur when smaller outages are imbedded in larger ones, meaning that a pocket of outages persists even after larger outages are repaired. Had National Grid addressed this functionality issue earlier, it would have been able to provide more accurate restoration information to its customers.
Although the department found these to be separate violations, we combine them for purposes of our analysis.
National Grid’s ERP — consistent with the department’s guidelines — defined critical facilities as: “A location or facility where the loss of electrical service would interrupt vital services to the public.” National Grid uses a tiered categorization system to determine restoration priority. Hospitals and other critical care facilities comprise the first tier. The second tier consists of fire, police, water, and sewer departments; nursing homes; and some schools that may be used as shelters. Tier three includes some schools, housing authorities, and telephone and cable providers, among others.
The department found that the performance of NSTAR Electric Company (NSTAR) was either reasonable, or that the evidence was insufficient to establish it was unreasonable, with respect to the majority of areas in which it was evaluated, including (1) weather forecasting and event classification; (2) acquisition of resources; (3) damage assessment; (4) outage management system performance and use of technology; (5) communication with State officials; (6) communication with life support customers; (7) advance planning and training; (8) vegetation management; (9) distribution automation; and (10) reporting.
The department found that, during Irene, NSTAR responded to all priority one calls that it considered life-threatening within two hours and ninety per cent of priority one calls, including those that were misclassified, within twenty-four hours. During the October snowstorm, NSTAR responded to all priority one calls that it considered life-threatening within two hours, and one hundred per cent of priority one calls, including those that were misclassified, within twenty-four hours.
In a level V event, NSTAR’s ERP provides that the company will respond to priority two and three calls based on a process that accounts for the particular circumstances and threat to public safety of each case and will prioritize the calls accordingly.
Those who offered comments at public hearings were asked to swear an oath before they testified. Commenters were informed that they were not required to take the oath, but that the department could base its decisions only on sworn testimony.
Like National Grid, Western Massachusetts Electric Company (WMEC) argued that the department’s response time calculations incorrectly measured when a downed wire was permanently repaired, not the initial arrival of standby *575guards or a field crew making temporary repairs, and included priority calls that had been misclassified by the reporting official. But as the department noted, WMEC “did not track most data related to wires-down calls,” nor did the data demonstrate how many of the priority wires-down calls “had fire or police personnel guarding the wire, or when the fire and police were relieved.” Accordingly, as was the case with National Grid, any deficiencies in the data regarding response times arose from WMEC’s failure to maintain accurate arrival and classification data in wires-down calls.
The breakdown of the penalties imposed against National Grid was as follows: (1) crew acquisition and deployment — $250,000 per day for six days ($1.5 million) for Irene, and $250,000 per day for eight days ($2 million) for the October snowstorm; (2) damage assessment — $200,000 per day for six days ($1.2 million) for Irene, and $200,000 per day for eight days ($1.6 million) for the snowstorm; (3) priority wires-down response — $250,000 per day for six days ($1.5 million) for Irene, and $250,000 per day for eight days ($2 million) for the snowstorm; (4) outage management system performance — $250,000 per day for six days ($1.5 million) for Irene, and $200,000 per day for eight days ($1.6 million) for the snowstorm; (5) communications with public officials — $250,000 per day for seven days ($1.75 million) for Irene, and $225,000 per day for nine days ($2,025 million) for the snowstorm; and (6) communications with customers — $100,00 per day for seven days ($700,000) for Irene, and $150,000 per day for nine days ($1.35 million) for the snowstorm.
The breakdown of the penalties imposed against NSTAR was as follows: (1) priority wires-down response — $250,000 per day for four days each for Irene and the October snowstorm ($2 million); (2) communications with public officials — $150,000 per day for six days ($900,000) for Irene, and $100,000 per day for five days ($500,000) for the snowstorm; and (3) communications with customers — $50,000 per day for six days ($300,000) for Irene, and $75,000 per day for five days ($375,000) for the snowstorm.
For its priority wires-down response, WMEC was fined $250,000 per day for eight days ($2 million).
The department did find that the company failed to “use[ ] an efficient restoration process in terms of. . . distributing its crews/resources . . . throughout the event.” But the specific examples highlighted in the order where the company made questionable decisions regarding resource allocation did not occur during the last two days of the restoration period. Furthermore, the data *579relied on by the department in assessing the efficiency of its crew deployment does not support an inference that the company’s performance in that regard was unreasonable during the last two days of the restoration period.