Learning platforms, built for the learner who stops in week three.

Learner model, content structure and outcomes

What a course actually is in your product. A linear sequence of lessons, a tree with prerequisites, a set of competencies a learner can demonstrate in any order, or a cohort that moves together on dates. These produce genuinely different platforms and the choice is pedagogical before it is technical.

Content reuse decides whether your catalogue scales. The same lesson appearing in three courses, a question bank shared across assessments, and a module updated once rather than in every copy are all consequences of modelling content as reusable objects rather than as pages inside a course.

Outcomes belong here too. If you intend to certify, the credential has to trace to specific evidence, which means the content model must carry what each item is meant to demonstrate.

Content delivery and the conditions learners are actually in

Educational video has different requirements from entertainment video. Sessions are shorter, resumed constantly across devices, and often watched at increased speed. Resume position that survives a device change is not a nicety; a learner who loses their place twice is a learner who stops.

Bandwidth assumptions cause the most damage. Large parts of the addressable market for education are on constrained or metered connections, and a platform that only works well on fast networks has excluded the learners who most need it. Download for offline study is a completion feature rather than a convenience.

Accessibility belongs in this phase and is frequently a legal requirement rather than a preference in education. Captions, transcripts, keyboard navigation and screen-reader support are part of the build, not a later remediation.

Assessment, integrity and credentialing

Assessment integrity should be proportional to the stakes, and platforms routinely get this backwards. A practice quiz needs no supervision at all. A certification an employer will rely on needs identity verification and controls that make substitution difficult. Applying heavy invigilation everywhere costs completion; applying none devalues the credential.

We build the full range of proctoring: camera monitoring, screen recording, behavioural flagging and live invigilation, plus the review queue that lets a human confirm a flag before it becomes an accusation. Which controls you apply to which assessment is configuration rather than architecture, so you can run light supervision on practice and full invigilation on a credential from the same platform.

Credentials should be verifiable by a third party without contacting you, and should reference the evidence behind them rather than asserting a result.

Progress, engagement and the mechanics of finishing

This is the phase that decides whether the product works, and it is mostly a design problem with engineering underneath. Learners abandon at predictable points: after the first difficult concept, after a week away, and at the transition from passive watching to active assessment.

The platform can address each of those directly. A returning learner should land where they stopped with context restored rather than at a catalogue. Progress should be visible and honest. Difficulty should be detectable, so support arrives before abandonment rather than after churn.

We build streaks, reminders and gamification where they help, and we say plainly when they are being used to manufacture engagement that does not produce learning. The metric is completion and outcome, not time in application.

Safeguarding, age and institutional obligations

If your learners include children, this phase outranks everything else on the page. Age verification, parental consent, restrictions on data collection from minors and controls on any communication between adults and children are legal obligations in most jurisdictions, and several carry personal liability.

Communication features need particular care. Direct messaging between a tutor and a minor without visibility or retention is a safeguarding failure waiting to be reported, so those channels are logged, reviewable and often restricted to supervised contexts by design.

Institutional buyers add their own layer: data-processing agreements, retention limits, and integration with the systems a school or university already runs.

Live classes and cohort mechanics

Live teaching is a different product from recorded content, not a feature added to it. Scheduling across time zones, attendance that matters when a cohort moves together, breakout groups, and a recording that becomes catalogue content afterwards are all distinct workflows.

Cohort models change the economics too. A course with dates has a deadline, and deadlines are the strongest completion mechanic available - which is why cohort products routinely finish at multiples of the rate self-paced ones do.

Launch and day two

One course or one cohort, with real learners and someone watching where they stop. Drop-off points are specific to your content and are visible within the first cohort, which makes that group worth more than any amount of pre-launch usability testing.

Day two is the first completion cycle and the first credentials issued, which is when assessment and certification get their real test. Sixty days of dedicated support, six months of priority bug resolution, twelve months of updates.