Skip to content
Maestro.
Assessment, in depth

Mastery makes the grade

Maestro runs three kinds of assessment with different purposes, rules, and security requirements. Confusing them is how ordinary platforms break. Keeping them separate is how the grade comes to mean something.

Three assessment types, three jobs

It is critical not to confuse these — they have different purposes, different rules, and different security requirements. Practice helps you learn. Evidence Checks diagnose what you know. Milestones carry the marks. Each is protected in the way its job requires, and none is asked to do another's job.

Evidence flows upward into a verified mastery belief; the grade is derived from the belief — never the reverse.

The three kinds

Practice interactives

Inside the node, during learning. Sorting, matching, ordering — retryable, with immediate, per-item, reasoning-based feedback. Their job is learning, so there is no secret to protect: they are never how the system decides mastery, and they carry no grade.

Evidence Checks

At the exit of every node. A short diagnostic capturing response, reasoning, confidence, and process — converted into a diagnostic band that drives routing and updates the mastery belief. One attempt, no in-flow feedback, never graded directly.

Milestone assessments

The formal, graded, summative exams under controlled conditions. The strongest witness in the evidence chain — and the anchor the whole belief system is calibrated against. Evidence Checks prepare learners for milestones.

The one-way law

The causal direction is a design law of the platform: mastery produces the grade; the grade never produces mastery. The grade attaches to the learner's accumulated, verified mastery picture — the triangulated belief — never to raw Evidence Check outcomes.

This is why cheating a check buys nothing. The check has no mark on it, and the belief it feeds is only consolidated when independent witnesses agree — including the controlled-conditions milestone. Fake evidence either collapses before the grade is computed or is contradicted by the one witness that cannot be outsourced.

The policies that keep them separate

  • Attempts. Practice is unlimited and encouraged. An official check is one attempt, with the first attempt always stored.
  • Feedback timing. Practice gives immediate, per-item feedback. A check gives nothing in-flow; feedback and routing come after reasoning is submitted.
  • What is written. Practice writes formative-weight signals. A check writes the full evidence record — band plus signals — into the learner model.
  • Progress. Progress arrives when mastery is shown, not when time is spent or boxes are ticked. Moving on is earned at the node exit.

What this means for institutions

A grade produced this way is defensible: it traces to a verified belief, built from named evidence, anchored by controlled conditions, with integrity designed into every step that produced it. That is a chain you can stand behind at any table — accreditation, appeal, or audit.

Get started

See this machinery on a real course.