Institution analytics answer three questions: is the platform being used, by whom, and where is it not landing. This article is what each number means and — more usefully — which ones are worth acting on.
Watch: Analytics
1 min 32 s · presented by Jason, Mandarin Temple's AI tutor
Captions are part of the picture, under Jason. The full transcript is below.
Read the transcript
- Hi, I’m Jason, Mandarin Temple’s AI tutor. Here’s how to read your school’s figures in the Learning Lab, shown in a demonstration school.
- Adoption comes first: how many learners are attached to your school, how many have practised, and how many were active in the last thirty days.
- Practised means real answers, not a single sign-in, which makes it the stricter number, and the more useful one.
- Outcomes are medians: how much your learners have mastered, and how much they still recall three weeks after a review.
- The school uses medians rather than averages, so one very active learner cannot move a figure meant for everyone.
- Your classes are listed alphabetically and never ranked, and each one is compared only with its own previous thirty days.
- Movement is the change in the share of right answers, in points, so a class that started low can still show its progress.
- A class with fewer than five learners practising shows no movement at all, and says so: too small to report.
- And results from the optional research tasks never appear here, whatever individual learners have agreed to.
- That’s your school’s Learning Lab: who is practising, how well, and how each class moves, with nobody ranked.
Open the video guide on its own page, with its key moments
The headline numbers
- Teachers — Accounts on your seats. Compare against seats paid for — an unused seat is money.
- Students — Everyone in the institution who is not staff.
- Classes — How many circles exist. Worth watching: classes tend to accumulate faster than they are retired.
Assignment completion
You get how many assignments were issued, how many student-assignments that produced, and how many are done — which gives you a completion rate across the whole institution.
The rate matters less than its spread. A school at 70% overall is usually two classes at 95% and one at 20%, and the average hides the only thing you can act on.
The per-teacher breakdown
For each teacher: how many students, how many assignments they set, and what proportion came back done. This is the most useful screen in the institution view and the easiest to misuse, so:
- Zero assignments is a training issue — A teacher who has set none has almost certainly not been shown how, or does not know it exists. That is a fifteen-minute fix and not a performance conversation.
- Low completion with high volume is a calibration issue — Someone setting a lesson a day at the wrong level. The fix is fewer, larger assignments.
- High completion with low volume may be fine — A teacher who sets one unit a week and gets it all back is doing better than one setting five and getting two.
Activity over time
The trend is worth more than any single figure, because the failure mode of a platform rollout is visible in it long before it is visible anywhere else: strong for three weeks, then a slide. A flat line at a modest level is a healthier picture than a spike that decays.
What is withheld
Group medians and cohort comparisons require at least five learners. Below that, an average plus one known score identifies the rest, so the figure is withheld rather than approximated. Plain counts are not affected — a class of three still reports as three.
Nothing anyone typed appears anywhere. The analytics are built from outcomes, and no endpoint behind them returns text a learner wrote.
What to actually do with it
- Check seats against teachers, monthly — The one number with a direct cost attached.
- Find the outlier class, not the average — Institution-wide completion is nearly meaningless. The spread is where the actionable thing is.
- Look for teachers with zero assignments — Almost always a training gap, and the cheapest intervention available to you.
- Watch the trend, not the week — Four weeks of decline is a rollout problem. One quiet week is a half-term.
Common questions
Group figures need at least five learners. Below that the average plus one known score identifies the others, so it is withheld. Counts are still shown.