Massachusetts charged a journalist $823,425 to ask the State Police how often they fly drones. The Supervisor of Records said that was fine — because the State Police gave the request a filing date five days after it was actually sent.

The full distribution of dispositions

Across all 30,800 orders, less than half end with the Supervisor issuing any order against the agency. About a third close procedurally because the agency provided some response during the appeal — without the Supervisor reaching the merits. Of the orders that are issued, only a fraction require actual record release; the rest direct the agency to respond more thoroughly, clarify scope, or quote a fee.

The 10.1% no-outcome residual is being studied. Most of these are appeals where the disposition phrasing eluded the current classifier. They are flagged for human and secondary-classifier review — and become the next regex iteration cycle.

The takeaway in one sentence

Of the long-tail "ordered" appeals, only 1 in 10 actually requires the agency to release records. The bulk are procedural-cure orders telling the agency to respond more thoroughly, revise a fee estimate, or clarify scope. The current single-bucket "ordered" rate on the headline chart should not be read as "released" — it captures all four of these sub-categories.

This is exactly the kind of distinction working journalists make instinctively from reading individual orders, and the kind of distinction a corpus-scale classifier has to learn explicitly. The sub-categorization is the next regex iteration cycle.

What it takes to get a release

Even when the Supervisor orders the agency to release records, the records do not appear by operation of the order itself. Six things have to happen in sequence, and the requester loses leverage at every step:

1. File a request. Free, but the requester needs to know to file and where — many would-be requesters never reach this step.
2. Receive a response (or non-response). Statutory deadline is ten business days. Agencies routinely seek extensions; many requests are answered with a citation to an exemption, an extension request, or silence.
3. Appeal to the Supervisor. Free and within ninety days, but requires writing a coherent appeal letter and waiting weeks to months for adjudication. A substantial fraction of denied requesters do not appeal.
4. Receive a substantive ruling. About 44% of orders are substantive (an order against the agency), per the disposition table above. The rest are procedural closes, withdrawals, or agency-respond-mid-appeal exits.
5. Agency complies. SOR has no direct enforcement mechanism. The agency can decline to comply. If it does, the requester's only escalation is a referral to the Attorney General, which the Supervisor may decline to make.
6. Sue in Superior Court. If the agency does not comply, the remaining remedy is a civil action under G.L. c. 66 § 10A. This requires a lawyer (or considerable pro se grit), filing fees, and a multi-month timeline. Most requesters do not reach this step.

The records-released rate is whatever survives all six steps. The single-digit "release-ordered" share inside the "ordered" bucket on the chart above is a ceiling, not a floor. Real-world disclosure — records actually changing hands and being read — is materially smaller still.

This is the structural reality the data describes: a system that produces an order in roughly half of regular appeals, but where the order is itself one stage in an attritional chain. The chart on this page measures stage 4. Stages 5 and 6 happen offline, mostly outside the public docket.

Why this matters

The Public Records Law was designed to be free by default. Agencies can charge for the time it takes to segregate exempt material from responsive records, but only after a Supervisor finds the burden cannot be reasonably met without segregation. The growth of agency-side fee petitions — and their high success rate — turns a free-by-default system into one where, for a meaningful fraction of requests, the requester pays.

Time petitions deserve the same scrutiny. An extension is procedurally minor; routine extensions across high-volume custodians are how response timelines slip from ten business days to ninety, in legally compliant increments.

The agencies most often asking to charge for redaction

One agency dominates the docket: the MBTA, which files fee petitions at 1.5 times the rate of the next-most-frequent custodian. Many of these are templated security-segregation petitions under G.L. c. 66, § 10(d)(iv).

The pattern hiding beneath the volume. Across 1,568 fee petitions in this corpus, custodians invoked the "commercial purpose" prong of the statute 1,146 times. The Supervisor of Records analyzed the claim substantively in most of those — and ruled against the custodian in 1,000, expressly determining the request was not for a commercial purpose. SOR found the claim was actually met in 68 cases, almost all of them personal-injury law firms fishing for accident reports. In 74 cases SOR sidestepped the question entirely.

Here is the twist. Of the 1,000 cases where SOR rejected the commercial-purpose claim on the merits, the petition was still granted 562 times — on the alternative "good-faith fee" prong. The commercial-purpose label is theater. SOR rejects it on the merits and grants the fee anyway, because the statute permits either prong to support a grant. The label costs the custodian nothing while adding rhetorical heft that goes unchallenged in practice.

The statute defines "commercial purpose" narrowly (G.L. c. 66 § 10(d)(ix)) and explicitly excludes news, citizen oversight, academic, scientific, journalistic, public research, and education uses. A separate analysis explains how to challenge these petitions — and one CPCS public defender (Deen Haleem, Esq.) has already done it in five business days on a single emailed objection.

Total dollars petitioned for, by agency

Across 1,541 fee petitions where a dollar amount could be extracted from the order text, two agencies account for nearly two-thirds of the cumulative ask: Massachusetts State Police at $1.2 million across 92 petitions, and the City of Malden at $912,000 across 77.

The median column is the distribution's tell. Most agencies sit at $25 — the statutory baseline — for the bulk of their filings, with a handful of large outliers responsible for the total. Two exceptions: the Department of Housing and Community Development filed only two petitions but both were large, and Town of Littleton's four petitions average $5,723. These are the agencies whose dollar profile is unusual not because of volume but because of the quote sizes themselves.

Known undercount. The dollar totals above only count fees the agency filed as a formal fee petition at the Supervisor of Records. They do not count the much larger universe of fee quotes that appear in the body of regular appeals — where the agency quoted an eye-popping fee directly to the requester, the requester appealed, and the dollar amount appears in the order body but the case is classified as a regular appeal rather than a fee petition. Agencies that quote large fees informally (and watch most requesters walk away) are systematically underrepresented in this table. A re-extraction over regular-appeal body text is queued as classifier-rerun #32; the inclusive totals will replace this table when that pass lands.

The ten largest fee petitions in the corpus

The Lipton drone case (SPR20/1448) is a worked example of the timing math: the request was sent on August 5, 2020; the Supervisor's order records receipt as August 10, 2020 — a five-calendar-day discrepancy that, under either reading, places the petition near the ten-business-day filing deadline at 950 CMR 32.06(4)(g). The regulation's own consequence for late filing is automatic: under 950 CMR 32.06(4)(b), "Failure to comply with 950 CMR 32.06(4) will result in a waiver of the right to assess fees for public records." The Supervisor did not address timeliness in the order.

One hundred and sixteen out of one hundred and sixteen

Among 116 fee petitions in the corpus where the requester was identifiably a journalist, news organization, or academic researcher and the Supervisor reached the prong-1 question, the outcome was the same in every single case: "not for a commercial purpose." One hundred percent. The named requesters include reporters from The Boston Globe, WGBH, Boston 25 News, MuckRock, the ACLU of Massachusetts, and several university research programs.

Yet in the same set, 24 of the 61 Globe-requester petitions were still granted on the alternative "good-faith fee" prong. The exclusion defeats the commercial-purpose label; it does not, by itself, defeat the fee.

The procedural clock — how fast the Supervisor rules

The regulation gives the requester five business days from being served to file a written objection (950 CMR 32.06(4)). When MA legal holidays are properly subtracted from the business-day math, the Supervisor's response time is even tighter than the standard window: 99.6% of fee petitions are decided within the 5 BD requester-objection window. Only six petitions across the entire corpus actually exceed that window once the holiday math is accounted for.

The 30.8% decided within three business days is the structural barrier. A requester who learns of the petition through the mail receives the agency's copy on day two or three, must read it, must draft a written objection, and must transmit it to the Supervisor at pre@sec.state.ma.us — all before the Supervisor rules. The cohort that gets a meaningful objection on the record is small, and the cohort that has the procedural runway to even attempt one is small.

This is one piece of the broader pattern: in 179 additional fee-petition orders, the Supervisor recited the regulation governing fee petitions and did not apply it in the analysis. The framework exists. The framework is cited. The framework is not enforced.

Methodology correction (2026-05-13): an earlier version of this section showed 8.4% of petitions ruling 6–10 BD after filing. A subsequent re-audit traced that cohort to a calendar artifact — the prior business-day math counted weekdays without subtracting MA legal holidays under G.L. c. 4 § 7 cl. 18. Once Veterans Day, Thanksgiving, Columbus Day, Presidents' Day, MLK Day, Labor Day, and Memorial Day are properly excluded, the cohort collapses to six true outliers. Time petitions (712 separate cases) are tabulated separately and behave the same way under their own statutory window.

The most-appealed custodians

What we're watching for next

The headline number — share of appeals where the Supervisor issued any order — is only the start. The classifier is being extended to catch finer-grained patterns: the four sub-categories of "ordered" (release / respond-better / clarify / fee-quote), doctrinal-avoidance maneuvers, recurring stalled matters, agency petition strategies.

Each new sub-classifier is published with its precision against a labeled holdout. The full methodology is open for inspection.

Ask a question of the data →