Research
Research notes
Written by the Reach120 research teamReviewed
We publish what we measure about our own product, including the parts that do not flatter it. Each note states its method before its results, names the limitations a sceptical reader would raise, and ships with a way to recompute every figure from the data in our repository. Where a measurement does not support a claim, we publish the measurement and withhold the claim.
The State of TOEFL Practice 2026An annual report on how people actually practice for the 2026-format TOEFL iBT — cohort breadth across the four sections, the rubric dimensions where marks are lost, item-skill accuracy in Reading and Listening, and the findings we refused to publish. Method, sample sizes, floors and the exact SQL for every figure.What our practice score gets wrong: a residual measured on a held-out setAn original research note on the residual bias of Reach120’s free practice score — methods, per-response data, limitations, and a script that recomputes every published figure from the committed raw runs.