Integrity & Trust
How We Verify Reviews
We hear the skepticism: “How do you know the reviewer actually worked there? What if it’s a disgruntled ex-employee?” Those are fair questions. Here is every safeguard we run on every submission — in plain English.
Every review passes through seven automated checks and manual founder review before it publishes. No review goes live without passing all of them. Reviews that raise red flags are held — not silently rejected — and reviewed by the founder personally.
1. BACB Credential Verification
Every reviewer’s BACB certification number is checked against the public BACB registry before their review counts toward the Fidelity Score™. We verify that:
- The cert number exists and is active (not lapsed or revoked)
- The credential type matches what the reviewer declared (BCBA vs. BCBA-D)
- The registry name is reasonably consistent with the account name
Reviews where the credential check fails automatically go to a manual hold queue — they are not published until the founder manually confirms the credential, requests resubmission, or rejects the review.
Reviews from BCBAs whose credentials are not yet verified are held in a separate queue and released as soon as the credential clears — without requiring the BCBA to resubmit.
2. AI Integrity Checks
Every submission is scanned by our integrity engine, which runs these checks simultaneously:
PHI detection
Scans review text for patterns that could constitute protected health information — client names, specific ages with diagnoses, street addresses, phone numbers. Reviews with potential PHI are held for manual review.
De-anonymization risk
Detects language that could identify the reviewer — 'I was the only BCBA at that location,' specific start/end dates, unique role-and-location combos, small-team identifiers. Reviewers are asked to rephrase.
Ethics language detection
Flags reviews that contain specific ethics keywords (billing fraud, BACB complaint, insurance fraud, falsification, harassment, etc.). These reviews are held for manual admin review before publication — not automatically rejected.
Duplicate detection
Verifies that no reviewer has already submitted a review for the same employer. Second submissions for the same employer are blocked and the reviewer is notified.
BACB cert reuse detection
Checks whether the BACB certification number submitted is already linked to a different account — a signal of a coordinated submission attempt.
Data consistency check
Verifies that caseload and billable hour data is internally consistent. A drift of more than 15 clients or 20 billable hours from offered to actual is flagged as a possible data entry error and reviewed manually.
3. Founder Manual Review
The founder reads every submission before it publishes. This is not a rule-based filter — it is a human judgment call on every review.
The founder looks for: reviews that sound generic or templated, language that seems designed to manipulate a score rather than describe experience, content that makes specific legal accusations without supporting clinical context, and overall plausibility given the other data points the reviewer provided (credential, tenure, caseload numbers).
Reviews that raise questions are either returned to the reviewer with a clarification request, held pending additional information, or — in rare cases — rejected with a written explanation.
4. Statistical Safeguards
Two statistical signals automatically flag reviews for additional manual review:
Dimension score outlier
If a reviewer's scores across 3 or more dimensions differ by 4+ points from the current employer average, the review is flagged as a potential outlier. This catches both genuinely extreme experiences (which are valid and informative) and possible retaliatory scoring. The founder reviews the context before publication.
Note: outlier does not mean rejected. A 100% legitimate grievance can score far from the average. The check is a signal, not a gate.
Submission velocity
If 3 or more reviews for the same employer arrive within 7 days, the batch is held and reviewed manually for signs of a coordinated campaign. This catches former colleagues who might have organized a simultaneous submission — not necessarily malicious, but worth a look.
5. One Review Per BCBA Per Employer
Each BCBA can submit exactly one review per employer. This is enforced at the application layer and at the database level. A second submission for the same employer is blocked with a clear error message.
If a BCBA’s previous review was not published (e.g., it was rejected), a second submission is held for manual review rather than automatically blocked — to give the reviewer a fair opportunity to address the issue.
6. Minimum Detail Requirements
To submit a review, a BCBA must:
- Hold an active BCBA or BCBA-D credential
- Rate at least 4 of 10 clinical dimensions
- Write a minimum of 50 characters of free-text experience
- Provide employment status (current, left voluntarily, left involuntarily)
- Agree to the attestation that the review reflects genuine firsthand experience
These requirements exist to prevent drive-by ratings and ensure every review reflects substantive professional experience, not a quick 1-star click.
7. Bias Pattern Detection
The integrity engine specifically checks for two bias patterns associated with non-genuine reviews:
Termination bias signal
Reviews submitted by BCBAs who were let go involuntarily, where all dimension scores are below 3 and the star rating is 1, are flagged for extra scrutiny. This does not block legitimate grievances — involuntary separations often produce genuine low scores — but it prompts a closer look at the review text for retaliatory tone versus substantive criticism.
High-velocity submitter
If a single BACB cert number submits 3+ reviews within 7 days, subsequent submissions are held automatically. Career-history catchup reviews (a BCBA reviewing multiple past employers in one session) are valid — but 3+ in a week is unusual enough to warrant manual confirmation.
8. Honest Limitations
No verification system is perfect. Here is what we cannot guarantee:
- We cannot verify that the BCBA worked at the specific employer they reviewed. The BACB registry confirms the credential is real — it does not confirm employment history. We rely on attestation, statistical plausibility, and manual review to catch mismatches.
- We cannot eliminate all biased reviews. A legitimate credential holder can write a biased review that passes all our checks. Our statistical and manual safeguards reduce this risk, but they cannot eliminate it.
- Minimum thresholds protect against single-reviewer distortion. A Fidelity Score is not published until an employer has received at least 5 verified reviews, which prevents one outlier from setting an employer’s score permanently.
If you believe a specific review contains false or defamatory information, employers can submit a formal dispute through their dashboard. All disputes are reviewed by the founder.
Questions about our verification process? support@verifiedaba.com. See also: Moderation Policy · Community Guidelines · Legal Protections for Reviewers