Corrections #
Jason Hreha·
Updated August 9, 2026
BehavioralStrategy.com corrects public errors in place when possible and records material changes here. Routine copy edits that do not change meaning are not listed.
Correction log #
| Date | Affected URL | Old claim | Correction | Reason | Status |
|---|---|---|---|---|---|
| 2026-08-09 | /cases/duolingo-habits/; /evidence/ledger/ |
The case attached the Duolingo learning-study evidence ID to unrelated DAU/MAU claims and described a cross-sectional course-completer study as evidence of learning gains. The ledger also treated 225 recruits as the outcome denominator and compressed different Spanish and French results into one level. | The case now uses the study only for its observed ACTFL reading and listening levels. The ledger reports outcome-specific sample sizes and levels, the observed study window, its course-completer boundary, and the authors’ Duolingo affiliations. | The cited study did not measure DAU/MAU or change from baseline. Its valid denominators and results differ by outcome and language. | Corrected |
| 2026-08-09 | /evidence/ledger/ |
The ledger described the commuter experiments only as context-bound, said system design mattered more than default prompts, and listed the effect as variable without attaching or summarizing the primary paper. | The record now reports Study 1’s small registration increase, its lack of active-use improvement, the transit-pass outcomes, the study-specific windows, the participant reuse, and the post-hoc power-based equivalence bounds. | The prior summary was not traceable to the study’s reported results and omitted material findings and design limits. | Corrected |
| 2026-08-09 | /cases/couch-to-5k/; /evidence/ledger/ |
The case and ledger treated 27.3% as a settled nine-week completion rate and derived 72.7% non-completion. | Both records now disclose that the paper’s abstract reports 27.3%, while its Results section reports 64.5% dropout and Figure 1 shows 39 of 110 completers (35.5%). They also identify the modified, single-site, instructor-supported program. | The source conflicts with itself, so it does not support one settled completion rate for this cohort or the standard self-directed NHS app program. | Corrected; source conflict remains unresolved |
| 2026-08-09 | /cases/proposify-ttfb/; /evidence/ledger/ |
The records published rounded onboarding and conversion figures without sample counts, base rates, uncertainty, or a defined calculation, and the case claimed TTFB and first-proposal improvements not reported by the source. | The records now retain only the company’s qualitative early observations, label them observational, and state that the effect is not estimable from the published account. | The company account was one month old, nonrandomized, and described its own result as potentially anomalous. It did not publish the information required for a quantitative effect claim. | Corrected |
| 2026-08-09 | /cases/ttfb-mobile-2025q2/ |
The removed page stated: “Pilot case study: reducing time to first behavior (TTFB) on mobile lifted D0-D7 task completion from 22% to 36%, a 14 point gain.” It also stated: “Baseline completion: 22% Post completion: 36% Δ-B: +14 pp”. Both statements are retracted. | The case, its two-row data file, its duplicate ledger row, and its machine-text copies were removed. The old case and legacy evidence URLs now redirect in one hop to the measurement standard. | One baseline row and one post row cannot support cohort percentages or an effectiveness claim. The case also reused an evidence ID assigned to a different claim in the canonical YAML ledger. | Corrected and redirected |
How to report an error #
Send the affected URL, the statement at issue, and any source that supports the report through the contact page. The editorial review checks the public page, its evidence record, and all generated copies before it closes a correction.
See the Editorial Policy for correction severity, disclosure, and propagation rules.