Adaptive practice
Question difficulty, scaffolding and topic order all respond to the individual. Two students in the same class can be on genuinely different paths through the same scheme of work.
The platform
Built around one question:
Built around one question:
what does this student need next?
Everything below exists to answer that quickly and honestly — and to make the answer useful to the adult standing at the front of the room.
The adaptive engine
A model of the student, not a score.
Most platforms record whether an answer was right. That is a very thin signal. A student who guesses correctly and a student who reasons correctly look identical in a percentage.
Bnm maintains a picture of which underlying sub-skills a student has actually demonstrated, updated with every response. Confidence in each sub-skill rises with independent success and decays over time if it is never revisited — because knowledge does the same.
- Sub-skill granularity. "Quadratics" is a dozen separate capabilities. We track them separately, because students fail at them separately.
- Prerequisite awareness. When a student stalls, we look one level down before assuming the topic itself is the problem.
- Guess-resistant. Response patterns that look like guessing are treated as weak evidence, not as mastery.
- Decay modelled. Mastery in October does not mean mastery in March. The schedule assumes forgetting and plans around it.
Skill map — Linear equations
Collecting like termsSecure
Inverse operationsSecure
Expanding bracketsDeveloping
Unknowns both sidesEmerging
Fractional coefficientsNot started
Illustration. Sample data.
Capabilities
What's in the box.
Misconception diagnosis
Distractors are written so that each wrong answer implies a specific error. A student picking −17 has made a different mistake from one picking 7, and we treat them differently.
Spaced retrieval
Completed topics resurface on a schedule derived from the memory research, so revision is continuous rather than a panic in the fortnight before an exam.
Class & cohort insight
Grouped misconceptions, movement over time, and a clear separation between "hasn't done the work" and "has done the work and still can't do it" — two very different problems.
Assignments & export
Set targeted work in a couple of clicks, or accept the platform's suggestion. Results export as CSV so they can drop into your existing reporting rather than sitting in a silo.
Runs anywhere
Browser-based, no install, no plugin. Built for Chromebooks, iPads, older desktops and phones — and to keep working when the school Wi-Fi does not.
Accessibility
If a student can't use it, it doesn't work.
Accessibility is not a compliance box we tick before a procurement meeting. In a mainstream classroom a meaningful proportion of students have a need that badly-built software will fail — and those are frequently the students who most need the support.
We build to WCAG 2.2 Level AA as the design target and test against it as part of our release process.
- Full keyboard operation — nothing that requires a mouse or a precise drag
- Screen-reader semantics on interactive components, including maths content
- Adjustable text size, spacing and contrast, with a dyslexia-friendly reading mode
- Reduced-motion support, and no time pressure imposed by default
- Colour never used as the only carrier of meaning
Need our accessibility statement or a VPAT for procurement? Ask us and we will send the current version.
Reading & display options
Larger textOn
Increased line spacingOn
Reduce motionOn
High contrastOff
Read question aloudOff
Illustration of the student settings panel.
Security & data protection
Children's data, handled like it matters.
Schools are asked to trust suppliers with some of the most sensitive personal data there is. Here is our side of that arrangement, stated plainly.
Held in the UK
Student data is stored and processed in the United Kingdom. Where a sub-processor is involved, they are named in our processing agreement before you sign it — not discovered afterwards.
You are the controller
Your school remains the data controller. We act as processor, on your written instructions, under a UK GDPR Article 28 agreement.
Encrypted, access-controlled, logged
Encryption in transit and at rest, role-based access, multi-factor authentication for staff accounts, and an audit trail of administrative access.
Minimal by default
We ask for the fields we need to run the service and no more. If a data point does not change what a student sees or what a teacher decides, we would rather not hold it.
Never sold, never brokered
No advertising, no data sales, no "analytics partnerships" that are data sales with better branding. We do not use identifiable student data to build models for other customers.
Yours to take back
Export your data at any time. On termination we delete it to an agreed schedule and confirm in writing when it is done.
On artificial intelligence
Where we use it, and where we deliberately don't.
"AI-powered" has stopped meaning anything useful in this sector, so here is the specific version.
We use it for: generating practice variants from teacher-authored templates, clustering similar student errors so a teacher sees one pattern rather than thirty scripts, and drafting explanations that a subject specialist then reviews before any student sees them.
We do not use it for: awarding grades that count, making decisions about a student without a teacher in the loop, or generating curriculum content that ships unreviewed. Anything a student reads has had a human check it.
Every automated flag is traceable — a teacher can open it and see the specific responses that produced it, and disagree. A system that cannot explain itself has no business making judgements about a child.
See it with your own curriculum.
A demo takes about thirty minutes, and we would rather walk through your actual scheme of work than a canned script.