Retention analytics you can defend in a hard meeting.
Any dashboard can print a retention number. What matters in a hard meeting is where the number came from — so this page shows the working: what’s measured, what’s modelled, and where the uncertainty sits. A number you can’t defend is a number you can’t use.
What counts as measured retention here
A retention number is only as honest as its delay. An end-of-course quiz measures short-term memory plus context; the score that matters is retrieval after time has passed. Our measured retention is exactly that: performance on scheduled reviews at real delays, per concept, per person — the practice engine’s normal operation doubles as a continuous measurement instrument.
Where we project — a cohort’s likely curve without reinforcement, an individual’s forecast before enough delayed data exists — the projection is labelled, its assumptions are printed, and it converges to measurement as data arrives. The dashboard never lets a model wear a measurement’s clothes.
The counterfactual on the chart
Program value is the gap between the measured curve and the projected no-practice curve. Putting both lines on one chart keeps the claim honest — and makes the value undeniable when it’s real.
Error bars that exist
Small cohorts and new content mean uncertainty; we show it. A retention estimate from twelve answers wears its confidence interval — the measurement-error page explains why we insist.
Aggregation that doesn’t launder
Team and org rollups always decompose: click any aggregate and see the concepts and cohorts inside it. No number on the dashboard is more than two clicks from its raw evidence.
The number after the number
Completion is the first number; this is the second — what each cohort still holds at 30, 60 and 90 days.
Interface shown as an illustration with representative numbers, not a screenshot — the layout is the product’s.
Interrogate the methodology.
Bring your most sceptical analyst to the demo — the fastest converts are the people who try to break the numbers.
The evidence this page stands on
Questions buyers ask
Why not just test everyone quarterly?
Formal testing at scale is expensive, resented and slow. Scheduled practice generates the same delayed-retrieval data continuously, as a by-product of the thing that also improves retention — measurement and treatment in one loop.
Can retention be compared across teams fairly?
With care: content difficulty and cohort tenure differ. The dashboard compares like-for-like where it can and flags where it can’t — a comparison with a caveat printed beats a clean-looking false one.
How early does the retention picture become reliable?
Cohort-level curves firm up within a few review cycles — typically weeks. Individual forecasts start as priors and sharpen with each delayed retrieval; the interval on the chart tells you how much to trust them.
Do you measure transfer to job performance?
We measure knowledge and its durability — the necessary condition. Linking to job outcomes needs your operational data on the other side; the export API exists precisely so your analysts can close that loop.
What stops teams gaming the metric?
Retrieval happens at scheduled delays the learner doesn’t control, questions rotate from a bank, and calibration scoring catches confident guessing. Gaming spaced retrieval is mostly indistinguishable from studying.
See it on your own content.
Bring one course. We’ll show you the retention curve your current training leaves behind — and what scheduled review does to it.
- 30 minutes, on your calendar — pick a slot here
- Run on your own content wherever possible, not a canned deck
- You see the dashboards, the learner surface and the evidence exports
- No commitment — and pilot data stays yours either way