Search "RBT turnover rate" and you get numbers between 30% and well over 100%, most of them without a hedge, a year, or a link. That range isn't noise — the underlying figures genuinely differ by source, by provider size, and by what is being counted. But the citations are usually missing, and a number you can't trace is a number you can't use in a board deck.
So this page is deliberately boring. Every figure below appears in the phrasing its source supports, hedges included, with the citation directly underneath it. Where a commonly-quoted number didn't survive checking, it isn't here at all. Take what's useful; the citations are the point.
A live 20-minute look at the product on roles you're hiring for now. Google Meet. No pitch deck.
One stat, its hedge, and the corroboration underneath it. The hedge is not decoration; it is the difference between the finding and a false claim.
Frontline RBT annual turnover sometimes exceeds 100%, alongside persistent BCBA shortages. The word "sometimes" is load-bearing and belongs in every restatement.
L.E.K. Consulting, "Investing in Autism Therapy: Key Trends Shaping the US ABA Therapy Sector," June 2025. Corroborated by the CentralReach 2025 Autism and IDD Care Market Report.
Typical RBT turnover runs 30–75%. Corroborating data shows large ABA providers averaged 103.3% all-employee turnover in 2024, and mid-size providers 89.3% — all-employee figures, not RBT-specific ones.
L.E.K. Consulting, June 2025. Confidence: high (three-of-three verifier corroboration against a primary source).
Two things follow from reading that carefully. First, "RBT turnover is 100%" is a misquotation — the sourced claim is that it sometimes exceeds it. Second, the 103.3% and 89.3% figures describe all employees at those providers, so quoting either as an RBT rate silently converts a workforce-wide number into a role-specific one.
The conclusion an operator can safely draw is narrower than the headline and more useful than it: at the rates these sources describe, the RBT bench is something you refill continuously, not something you fill once.
The single most misquoted number in ABA hiring content, and the reason it's misquoted.
164,015 RBT exam sittings occurred in 2025 — 109,341 first-time, 54,674 retakes. That figure counts testing events, not unique people; it is never "164,015 people got certified."
BACB Certificant Annual Report Data; BACB, "2025 RBT Examination Pass Rates for 40-Hour Training Providers," updated 05/2026.
The first-time pass rate is 75%, the retake pass rate 38% — roughly one in four first-time candidates does not pass, and retakers succeed fewer than four times in ten.
BACB Certificant Annual Report Data. Confidence: high (three-of-three, primary source is the certifying body).
BACB reports 246,109 active RBT certificants worldwide versus 81,566 BCBAs as of end of 2025 — roughly 3 RBTs for every BCBA, with the RBT population growing about 2.5x faster (25.2% vs. 10.0% year over year). "Worldwide" stays in the framing: these are net active credential holders globally, not US-employed clinicians.
BACB Certificant Annual Report Data (the certifying body's own dataset).
US employer demand for BCBAs — measured by job postings — hit 132,307 in 2025, up 28% year over year, and has risen every year since 2010. It is a count of postings requiring or preferring the credential, never an unfilled-position count.
BACB, "US Employment Demand for Behavior Analysts: 2010–2025," Lightcast deduplicated postings data, updated 01/2026.
Candidates must be 18 or older, hold high-school-level education, pass a criminal background check plus an abuse-registry check, complete 40 hours of qualified training, and pass an initial competency assessment. Two hard 90-day rules apply: the competency assessment must follow the 40-hour training and fall within 90 days of application submission, and the application itself stays active only 90 days before requiring reapplication and a new fee.
BACB RBT Handbook (current 2026 edition); BACB, "Guidance for Meeting RBT Requirements During the 2026 Transition," updated 08/2025. Cite the current 2026 documents rather than the 2021 handbook PDF — eligibility requirements are unchanged, but the 40-hour training must align with the 2026 curriculum outline and the January 1, 2026 transition supersedes the older URL.
RBTs must receive ongoing supervision for at least 5% of monthly service hours, including at least two face-to-face real-time contacts per month with at least one direct observation.
BACB RBT Handbook (current edition). BACB is a certification board rather than a regulator as such, though payers and state licensure incorporate its standards — and some payers require more than 5%, so 5% should not be presented as a universal floor with no exceptions.
There is no verified source for the cost of replacing an RBT. So this is a worksheet, not an estimate — every input is one you supply, and we label them as assumptions because that is what they are.
Every dollar figure you have seen for "cost to replace an RBT" is either modelled or invented, and most published ones don't say which. We aren't going to add another. What follows is the arithmetic itself, with the inputs left where they belong — with you.
| Input | What it is | Where the number comes from |
|---|---|---|
| R | RBTs and BTs on your payroll today | Yours. A fact, not an assumption. |
| T | Your annual turnover rate for that group, as a decimal | Yours, from your own last twelve months. Sourced context: typical runs 30–75%, and sometimes exceeds 100%. |
| H | Hours of recruiter and owner time spent per hire, end to end | An assumption you supply. No verified industry benchmark exists for this. |
| W | Loaded hourly cost of the people doing that hiring | An assumption you supply. Yours to work out from payroll. |
| B | Job-board spend per hire | Yours, from last year's invoices. A fact if you look it up, an assumption if you don't. |
| D | Billable hours lost per vacancy before the seat refills | An assumption you supply, and the one most likely to be wrong. No verified source states it. |
Replacement hires per year = R × T. That's the whole first step, and it's the step most agencies never write down. It's also where the sourced figures earn their keep: at the top of the range those sources describe — where turnover sometimes exceeds 100%, so T is at or above 1.0 — the number of people you hire in a year meets or exceeds the number of people you employ. Everything downstream doubles when T doubles, which is why the rate is worth measuring before anything is worth buying.
Annual hiring labour = R × T × H × W. Annual job-board spend = R × T × B. Annual service disruption = R × T × D, valued at whatever a billable hour is worth to you. Add the three. Whatever that total is, it is a recurring operating cost with a rate attached, not an occasional expense — and the only terms in it you can move quickly are H and B.
Two honest caveats about using this. It says nothing about the cost of a bad hire, which is real and which nobody has credibly quantified for this role. And every assumption you plug in is yours — if you present the output externally, present the inputs with it.
A compilation is only as useful as the things it declined to include.
Several figures that circulate widely in ABA content failed verification when we checked them, so they appear nowhere on this site — not with a hedge, not with an attribution, not as "some sources estimate":
That last one is worth stating on a page like this. It would be easy to end a turnover article with a figure about what our software does to it. We don't have one, and an invented one would make every sourced number above worth less.
Nothing clever. At the rates above, RBT hiring is a standing process, so the product is built as one: BT, RBT, BCBA, BCBA-D, BCaBA, QBCBA and QASPS are first-class credential types with verification tracked through its full lifecycle and an immutable audit trail; expired, mismatched, and missing credentials are flagged before you open a file; every applicant is scored 0–100 against the opening's real requirements with the reasoning shown; and the applicants you already paid for can be imported from any spreadsheet export (CSV/Excel) and bulk-scored against today's openings.
No automated credential verification. Screening is automated. Verification is a person checking a registry, with an audit trail.
No BT-to-RBT certification pipeline tracker. The 90-day windows above are your world, not a module in this product. On the roadmap; not built.
No retention or turnover analytics. This is hiring software. It will not tell you why people left.
No customer-outcome statistics. None exist, so none are published.
SherStaff (Sher-Staff LLC), "RBT turnover: what the sourced numbers actually say," sher-staff.com/aba/rbt-turnover-data/ — but cite the underlying BACB and L.E.K. sources named above wherever you can, because those are the ones a reader can check.
Because the source says "sometimes," and dropping it changes the claim. L.E.K.'s finding is that frontline RBT annual turnover sometimes exceeds 100% — not that it generally does, and not that a given agency's does.
Typical RBT turnover runs 30–75%. The corroborating all-employee figures — 103.3% at large ABA providers in 2024, 89.3% at mid-size providers — are all-employee, not RBT-specific, and we label them that way every time.
Please cite the underlying sources instead — they're named and linked in every source line, and they're what a reader should be able to check. If it's useful to point at the compilation, there's a suggested citation at the bottom of the page.
If you find an error in how a figure is phrased here, tell us and we'll fix it. That's a cheaper outcome for both of us than the figure travelling further in the wrong form.
Because the ones we tested didn't survive verification. ABA market-size dollar figures, counts of how many ABA companies exist, and "X analysts short of demand" headcounts all failed the same check, so none of them appear anywhere on this site — not with a hedge, not with an attribution, not as "some sources estimate."
The demand figure that did survive is a count of job postings, and it only ever appears here as a count of job postings.
We don't know, and neither does anyone publishing a number for it. No verified source states a replacement cost for an RBT position, so any dollar figure you've seen quoted is either modelled or invented, and usually doesn't say which.
What we can do honestly is give you the arithmetic with your own inputs in it. That's the worksheet further up this page — every variable is one you supply, and the only numbers we contribute are the sourced rates.
Google Meet. No pitch deck. If it isn't a fit, I'll tell you on the call.