Methodology · Corrections

Corrections

This site's only real asset is that its numbers can be checked. That is worth nothing unless the errors are published too. Every mistake found here is logged below — what we said, what was wrong with it, and what it should have said — including the ones no reader had noticed.

Found something wrong? Tell us and it will appear here.

6 August 2026Material

We said no ABA platform publishes its pricing. Several do.

We published "Not one of the twelve platforms in this guide publishes its pricing. Every single one is quote-only."

It should have said that most do not. Several publish per-seat rates on their own websites — Office Puzzle, for one, lists $19.99 per user per month.

What went wrong. The claim was almost certainly true when written and was never re-checked. Pricing transparency is exactly the kind of fact that changes without anyone announcing it, and an absolute — "not one", "every single one" — is the most fragile sentence you can write about a market. It was found while building a replacement comparison page, not by a reader.

Also. Several tool listings still carry a machine-readable flag saying the vendor does not publish pricing. Those are being re-checked one vendor at a time against each vendor's own site, and we are not flipping any of them on inference.

6 August 2026Editorial

Two vendor listings cited review sentiment. Both are gone.

Our CentralReach listing carried "Learning curve is the most commonly cited drawback in public reviews," and our Rethink listing carried "Newer entrants frequently position against its interface."

Both are unverifiable claims about named, identifiable businesses sourced to aggregate sentiment rather than to anything checkable — which is precisely what our own comparison rule forbids. They were written before that rule existed and were not caught when it was added.

They have been replaced with facts from each vendor's own published documents. The CentralReach listing now records its published contract term, notice window and price-adjustment policy, which is both safer and more useful to a buyer than a sentiment summary ever was.

5 August 2026Material

Our correction to Florida's rate was itself wrong

We published $12.19 per 15 minutes for CPT 97153 — the figure we moved to on 2 August when fixing the error described below.

It should have said $12.26 per 15 minutes, $49.04 an hour.

What went wrong. The same failure as the original, one code over. $12.19 is what Florida pays for 97152, 0362T and 0373T — three other codes on the same page. In correcting a code mix-up we introduced a second code mix-up, and then rejected the right answer: a research pass had proposed $12.26 and we ruled it out, believing it belonged to a different year. It did not. Florida prints $12.26 on both the 2025 and the 2026 schedule.

Why it took three days. Because we never reopened the PDF. The 2 August fix was made by reasoning about which figure belonged to which code rather than by reading the document again — the exact substitution this site's first rule exists to forbid. The error was found only when we returned to restore the withdrawn rows.

Also. The correction below states that $15.37 is Florida's rate for 97156. That is wrong too. $15.37 is 97155 with modifier HN, the assistant-analyst rate for protocol modification; 97156 with modifier HN is $15.24. The original error was real and the size of it was right — the code we blamed it on was not.

What changed as a result. Florida's page now carries all 23 rows from the 2026 schedule, including the 97155 and 97156 rows withdrawn in July, each read twice from the source PDF by two different extraction methods. We also compared the 2025 and 2026 schedules line by line and confirmed Florida changed no ABA rate between them.

2 August 2026Material

Florida's Medicaid rate was overstated by 26%

We published $15.37 per 15 minutes for CPT 97153, Florida's core direct-therapy code, or $61.48 an hour.

It should have said $12.26 per 15 minutes, $49.04 an hour. At the time we published $12.19, which was also wrong — see the entry above, dated 5 August.

What went wrong. A code mix-up. $15.37 is Florida's rate for 97155 with modifier HN, protocol modification by an assistant behavior analyst — a different service delivered by a different practitioner. The figure was read from the correct schedule but attached to the wrong code. It was live for roughly two weeks, on the state holding a quarter of the country's ABA organizations, in the dataset this site's credibility rests on. It was caught by re-reading the source rather than by any reader.

Also. Florida's 97155 and 97156 rows were withdrawn at the same time pending a re-read, rather than corrected on a guess. They were restored on 5 August.

2 August 2026Material

Indiana's rates had been superseded for three months

We published $17.06 per 15 minutes for 97153, transcribed correctly from IHCP bulletin BT2023169.

It should have said $16.04. Indiana cut all non-group ABA rates by 6% effective 1 April 2026 under bulletin BT202627 — three months before our own page claimed it had been checked.

What went wrong. We verified the figure against the document we cited and never asked whether that document was still current. Codes 0362T and 0373T were also in the cited bulletin but missing from our table with nothing explaining the absence, so the page implied Indiana does not pay them.

2 August 2026Withdrawn

TRICARE's rates were removed rather than corrected

We published hourly ranges for TRICARE across 97151/97156 and 97153/97155.

We now publish no figure at all. The rate file sits behind a JavaScript gate; three separate reads of the page disagreed with one another; and one suggested our ranges had 97155's and 97156's numbers attached to the wrong codes — the same failure as Florida. A newer schedule effective 1 May 2026 also supersedes the one we cited. The page now links the official document and states no rate.

A missing figure is recoverable. A wrong one is not.

2 August 2026Material

Nine citations pointed at documents containing no ABA rates

A citation that goes nowhere useful is a broken promise, and nine of ours were.

  • Arizona cited the Behavioral Health Outpatient schedule, which contains only H, S and T codes and no ABA CPT codes whatsoever.
  • Texas cited the Outpatient Behavioral Health schedule. The ABA codes are on a different schedule entirely, Autism Services.
  • Georgia cited the January 2018 ASD policy manual, which prices ABA against Category III codes the AMA deleted at the end of 2018.
  • Maryland's own fee schedule link returns a 404; the live schedule is inside a provider manual.
  • Ohio cited a package stamped "DRAFT — NOT FOR FILING" with proposed figures and a blank effective-date column.
  • West Virginia's policy link returns a 404.
  • Nevada and Washington cited billing guides that list every ABA code and no dollar amount anywhere.
  • Michigan's cited document refuses all non-browser requests.

All nine now cite documents that contain the figures.

2 August 2026Clarified

The BCBA count needed its scope explained

We published 75,574 certified BCBAs. The BACB's own headline figure is 85,587, so anyone checking would have concluded our number was wrong.

It was not wrong, but it was unexplained. 85,587 is worldwide; 75,574 is the 50 states and DC. Roughly ten thousand BCBAs are certified outside the United States. Both source notes now say this outright.

2 August 2026Material

The company directory listed one percent of the companies

We published 234 companies, those with three or more registered locations.

The registry holds 17,781. Every single-site company was invisible, which meant a reader searching for a real company they knew existed would conclude it was not registered. Two were reported to us by name.

What changed. Every registered organization is now listed, across state and city pages, and searchable. Street addresses are still withheld for companies with fewer than three locations, because those may be home addresses — that protection was the reason for the threshold and it has been kept, while the exclusion has gone.

2 August 2026Names

Companies were listed under names nobody uses

We published federal legal business names. ABAECARE LLC is known as Catalyst Behavior Solutions; SOAR HEALTH, INC. as Soar Autism Center; PANTOGRAN LLC is the Center for Autism and Related Disorders.

What changed. Twenty-four companies now lead with the trading name, each confirmed against the company's own site rather than a directory. Five separate legal entities turned out to be one company, HealthPRO Pediatrics, and were merged into one page; CARD was split across two registrations and was merged too.

What did not change. Thirteen companies where the registry's alternative name is a program brand rather than a trading name, or where two different companies share a name, were left alone. Asserting a name we could not confirm would be worse than an unfamiliar one.

2 August 2026Editorial

Two page titles were wrong, and the writing was not American

The reimbursement dataset's title said "The Data Index" rather than "The ABA Index". The statistics page was titled simply "Statistics". British spellings — organisation, programme, practise, catalogued — had spread through the company pages and the statistics page on a site written for a US audience.

All fixed, and a build gate now fails on British spellings outside quoted names and cited document titles.

1 August 2026Material

A company history section failed its own audit

An audit of 141 claims across the company pages found fewer than half were supported by the source cited for them, and ten were contradicted by their own citation. Among them:

  • Motivity's headquarters was cited to a page that states no location anywhere.
  • The NPAWorks page said a chief executive "was quoted without being named". The cited release names and quotes her.
  • Passage Health's founding year was wrong by one year, and its funding total double-counted a note that later converted.
  • CentralReach's acquisition price was labeled gross when the filing says net of cash acquired; its headquarters city was stale.
  • Artemis ABA described someone as a co-founder when the company's About page says he joined after founding.

The section was held back from publication until every claim was re-verified against its source. Verification before publication is now the standing rule.

1 August 2026Editorial

We drew conclusions the documents did not support

Company pages carried lines such as "private-equity owners generally sell within a few years, and the next owner sets the roadmap and the renewal price". That is assumption presented as analysis — unsourced, unfalsifiable, and unfair to companies that had done nothing to earn the implication.

Removed. The rule now is to state what the filings say and let the reader draw the conclusion.

Why this page exists. Anyone can claim their figures are sourced. The test is what happens when one turns out not to be. Publishing the failures is the only version of that claim worth anything — and every entry above was found by us re-reading our own sources, not by a reader complaining.

What we count as a correction. Any change to a published fact, any figure withdrawn, and any citation that did not support what it was attached to. Typos and design changes are not logged. Additions are not corrections.

See also data health for what is currently unverified, stale or missing, generated from the data itself.