Adaptive Teaching: The Complete Guide

Adaptive Teaching is the companion to adaptive learning. Students already get a personalized path. This turns a cohort’s knowledge map into a teaching plan: short lessons in prerequisite order, groups formed from readiness rather than reputation, and a checkpoint closing every unit. The engine plans, AI only narrates, and the educator approves everything.

A tutor looking at a red-and-yellow cohort map still has to decide, alone: what do I teach this group next, in what order, how do I structure a 30-minute lesson, when do I stop and review, which students need something different, and how do I know it worked?

Adaptive Teaching answers those six questions with evidence the platform already has.

The vocabulary

Familiar teacher words wrapped around one ownable concept. Worth learning before the rest of this guide, because every screen uses them.

A cohort is a group of students, full stop. It is not defined by a subject, so a cohort that is weak on both early algebra and sentence boundaries can get a single cycle that teaches both foundations, ordered by which is more foundational and lower-scored.

How the cycle works

Each pass through the cycle is one loop of the spiral, and the spiral revisits skills at increasing depth: a skill taught at recall level returns later at application, then at strategic thinking. This is Bruner’s spiral curriculum made concrete on a knowledge graph.

Diagnose. The cohort’s mastery state per skill and per depth level, drawn from practice students are already doing. Nothing extra to administer.

Plan. The engine selects skills that are shared, foundational, and teachable, then orders them by the prerequisite graph. Shared means many students lack it. Foundational means its absence blocks many skills above it. Teachable means most students have the prerequisites already.

Teach. A unit of short, structured lessons, four by default, each with practice attached.

Checkpoint. A cumulative review closing the unit, built from real practice data and read together with everything else the students did that cycle.

Synthesize and re-plan. Results reshape the next unit and arrive as a change you approve, never as a silent edit.

The economy of instruction

Group instruction time is the scarcest and most expensive resource in the system, so Adaptive Teaching spends it deliberately. When a gap is not worth group minutes, it escalates to the cheapest rung that fits rather than consuming the class.

Alerts propose a rung. Humans climb it. Nothing moves a student automatically.

Because the group plan and each student’s personal journey run on the same mastery data, they coordinate on their own. When a group lesson pushes a skill past mastery, the student’s personal practice stops serving it. When a gap never clears the bar for group time, personal practice picks it up.

Planning a Spiral

Setup is a four-step wizard.

1. Cohort and students. Who is in it.
2. Cadence and format. Sessions per week (three by default), session length (30 minutes by default), subject scope, grouping mode, maximum group size, how re-grouping is handled, and modality.
3. Horizon and units. Length in weeks (ten by default), lessons per unit plus its checkpoint (four plus one by default), and an optional target test date.
4. Preview and edit. The full proposed sequence, editable before anything is committed.

Prerequisite violations warn inline. They never block, because the educator is the authority on their own cohort. Approving activates the Spiral; nothing is scheduled or assigned to students before that.

For a cohort with no practice history, the first unit is a diagnostic unit instead of a guess. The engine never fabricates readiness for students it has not measured. They are flagged as low-evidence and never quietly averaged into the group’s numbers.

Grouping, and who names the groups

Grouping runs on per-skill readiness profiles rather than an overall score, because students with very different total scores often show equivalent readiness for a specific skill.

You name your groups. The name is yours to choose and yours to change at any time. Each new group arrives with a neutral two-word name drawn from a curated, classroom-appropriate pool, for example “Brave Falcons”, and no two groups in your organization get the same one. That is only a starting point: rename any group inline, whenever you like. Tutors and organizations often have their own conventions, whether that is house names, cohort codes, or a naming scheme a school already uses, and the system should not override them.

The skill the group is working on, for example “Working on: linear foundations”, sits underneath the name as a subtitle. It tells you what the group is for without deciding what the group is called. When a re-plan reshuffles membership, your name follows the students it shares the most members with, so renaming survives re-grouping. Badges a group earns keep the name it had at the time.

What the platform does hold to is its own data model: there is no rank or level field behind a group, so whatever you call it, the platform is never sorting students into a hierarchy underneath. Grouping is by readiness for a specific skill, nothing more.

The rules that keep this from becoming tracking:

• Re-grouping happens only at checkpoints, never mid-unit, and only when the evidence points the same way at least twice.
• A change budget caps churn, so a re-cluster moves a small share of memberships rather than reshuffling everyone. Group stability is treated as a pedagogical asset, so the default is stay-put.
• Borderline students are shown as borderline. The membership grid says a student fits either group rather than silently forcing a choice, and educator pins survive re-clustering.
• Cross-class moves are time-boxed first. A guest placement lasts one unit and has an explicit expiry and return, before anyone discusses a permanent move.

When a skill does not land

A checkpoint that shows a skill was not mastered triggers a reteach, and the reteach has rules.

It carries to the front. The skill moves to the start of the next unit, and always before any lesson that depends on it. That ordering is a structural guarantee, not a convention.

The approach must change. The first attempt demonstrably missed, so a reteach cannot be a rerun. It has to change something real: a different worked example or method, different materials, a scaffold, a smaller group, a modality change, or an engagement fix first. A second reteach must differ from the first as well.

Persistence is bounded, with a high ceiling. After a set number of genuinely different attempts, the skill leaves group rotation and routes to individual support, so group time stops paying for what group time is not fixing.

Engagement is evidence

Outcome data without engagement context misattributes failure, so the profile tracks how a student works, not just whether they were right.

These feed the alerts and the lesson plan notes. A student who is disengaged gets an engagement fix before another content reteach, because reteaching content to someone who is not participating changes nothing.

Setting expectations, then measuring them

Every Spiral predicts its outcomes before you teach. Every checkpoint measures against them. You see expected versus actual, per skill and per student, and the planner recalibrates as evidence accumulates.

The forecast is reported honestly: a predicted cohort mean, a certainty band, and the distribution, so the spread is visible rather than hidden behind an average. Coverage is stated too, meaning what share of the roster has enough data for the forecast to mean anything.

A checkpoint result reads like this: the group met expectations on four of five targets, two students landed below expectation, and both had under half their pre-work done. The reason travels with the number.

What AI does, and what it never does

For students and guardians

Students see one coherent list. Group-sourced work sits alongside personal practice, with a one-line explanation of how it fits: for example, that the group starts linear equations at medium difficulty next, having cleared the easy tier last week. Two engines run behind it; the student sees one plan.

Groups earn badges together for clearing checkpoints and building foundations. There is no punitive badge and no worst group. A missed checkpoint simply earns nothing.

Guardians get the same picture in their child’s detail view and in a weekly summary.

Common questions