Analytics & reporting

The real-time picture lives in the session's live view — the heatmap of students × problems, the automatic "this student needs you" banners, and the per-problem accuracy and help-request numbers. That's where a session turns into instruction: it shows what students worked on, where they slowed down, and which misconceptions are clustering — the inputs for tomorrow's small groups.

The walkthrough above reads a live dashboard end to end. The written version, with every control explained: manage a session while students are working.

For the after-the-fact rollups — engagement and performance across weeks, not minutes — see what each report shows.

Every problem a student works gets a support level — the rating behind the heatmap colors and the per-student averages:

• Independent — completed without tutor help (at most one hint, no wrong attempts).
• Occasional support — completed with limited tutor interaction.
• Extensive support — completed, but with many hints or retries.
• Incomplete — attempted but never answered.
• Not started.

The rating is computed from two counts: how many times the student asked the tutor for help, and how many wrong attempts they made. Wrong attempts weigh about twice as much as help requests.

How to read it

Support levels answer a different question than scores do: not "did they get it right" (with the tutor, nearly everyone eventually does) but "how much scaffolding did it take". A student living in Extensive support has a gap worth small-grouping; a class drifting from Extensive toward Independent on a skill is the growth you're looking for.

Which is also why support levels shouldn't be graded — grade on completion instead. Students never see these ratings, so asking for help stays safe.

If your school runs Thinkverse under an organization, you get an Analytics page (independent accounts see their per-session analysis and dashboard counters instead). A mode dropdown scopes everything to assigned sessions, Skill Path, self-practice, or the freeform tutor — or all of them at once.

Engagement — is it being used?

• Active students — who attempted at least one problem in the period.
• Student usage by school — week-by-week or month-by-month.
• Sessions created, Problems completed.
• Practice time with AI tutor — actual working time, not tab-open time (idle time is capped per problem).
• AI interactions — how often students asked for help.

Performance — is it working?

• Proficiency gain and Confidence gain — before-vs-after movement, measured by entry and exit tickets.
• Session performance distribution — how sessions land across support levels.
• Student performance — the per-student table.

In Skill Path mode you also get Skills completed and progress toward the weekly goal you can set per classroom.

Everything respects the filter row — school, teacher, classroom, time period — and exports: Download CSV (the dashboard view, or a per-student file) and Print PDF. For families and conferences, there's a Performance report PDF per student.

For what a single session's numbers mean while it runs, see the live session view and support levels.

We recommend grading based on completion.

The tutor is built so that every student can reach the right answer with enough scaffolding — that's the design, not a loophole. So accuracy alone is a poor grade: it mostly measures who needed help, not who learned.

Grading on completion also protects the behavior you want. If asking for support costs a student points, they'll stop asking, and you lose both the learning and the signal. Don't penalize kids for asking for support.

Use the diagnostics for the real picture instead: support levels show you how much scaffolding each student needed, and the reports show whether it's translating into growth.