Building an Online Course

There is a particular kind of disappointment that visits course creators about three weeks after launch. The sales were fine. The reviews, where they exist, are kind. And yet the completion dashboard tells the truth: most buyers watched two lessons and vanished. The course works as a product and fails as an education. After building and watching enough of these, I have come to believe the failure is almost never in the recording quality, the platform, or the price. It is in the design – or rather, the absence of design.

A course is a promise to change someone

The single most useful reframe I can offer is this: you are not selling information. Information is free, abundant, and ignored. You are selling a transformation – a move from a frustrating current state to a desired future one. “Learn spreadsheets” is a topic, and topics do not sell or stick. “Go from dreading your monthly budget to building a forecast you actually trust” is a transformation, and transformations do both.

This is not a marketing trick; it is a design constraint. Once you commit to a specific before-and-after for a specific person, every later decision gets easier and every vague decision gets exposed. You cannot wave at “beginners.” You have to name someone real – the office worker who can boil pasta but freezes at a recipe, the backend developer who deploys to a VM and wants to containerise one service. The narrower the person, the sharper everything downstream becomes.

Validate before you build, not after

The most expensive mistake in this craft is building first and asking the market later. It feels productive to start recording. It is, in fact, the riskiest possible order of operations, because you are committing weeks of work to an untested guess.

Validation is cheaper than building and infinitely more honest. Read where your audience already asks questions. Talk to five real people and write down their exact words – those words become your sales page. Then do the only test that counts: ask for money. A pre-sale, a waitlist deposit, a paid workshop. “Would you buy this?” is a useless question; “here is the link, it is $149, doors close Friday” is a real one. Interest is cheap. Intent is rare and decisive. If you cannot earn a single committed pre-order after honest effort, the answer is to change the offer, not to push on regardless.

Design backward from the outcome

When demand is real, the temptation flips from “will anyone want this?” to “I should include everything.” Scope creep is the quiet killer, because every adjacent idea feels generous. The discipline is brutal and simple: for every potential module, ask whether the learner needs it to reach the promised outcome. If it is merely interesting, it goes in a bonus list, not the curriculum.

The professional method here is backward design, drawn from Wiggins and McTighe’s Understanding by Design. You start from the result, then decide what evidence would prove the learner reached it, and only then plan the lessons. Designing the assessment before the lesson feels inverted because school taught us to teach first and test as an afterthought. But if you cannot describe what success looks like, you cannot teach toward it – you can only present material and hope. Phrase every outcome as something observable, something the learner can do: not “understand budgeting” but “build a twelve-month cash-flow forecast from a bank statement.”

Build for retrieval, not consumption

Here is where most courses quietly betray their learners. They are, in practice, a pile of videos – ninety percent watching, ten percent maybe-trying. And the research, summarised accessibly in Make It Stick, is unambiguous: learning happens through effortful retrieval and application, not passive viewing. A course that is mostly video will earn warm reviews and produce almost no change.

So design the friction in. Every module should make the learner produce something – a low-stakes quiz that teaches through its feedback, an exercise that bends the lesson to a new situation, a project they keep. Use video for demonstration and human presence, text for reference and the things people copy, quizzes for recall, projects for proof. The same alignment and feedback principles that govern good assessment design apply directly: align each item to an outcome, and give feedback the learner can act on. Flip the ratio from watch-heavy to do-heavy, and you build a course people finish.

Completion is the only honest metric

Choosing a platform and a delivery model matters – hosted versus self-hosted, drip versus all-at-once, self-paced versus cohort – but one finding dwarfs the rest: the social layer is the strongest lever on completion. Self-paced courses, alone in the dark, post completion rates in the single digits. Add a cohort, a deadline, a weekly check-in, a community where someone notices you stalled, and the numbers transform. People finish things when other people are watching and helping.

Which brings us to the metric that should haunt every creator: completion rate. It is the most honest measure of quality you have, because it captures whether your design actually carried people to the transformation you promised. Revenue tells you the sales page worked. Completion tells you the course worked.

So launch to a beta cohort first, watch where learners drop off, and read a cliff at module three as a signal that module three is broken – not that learners are lazy. Then fix it, and fix it again next cohort. The course is never done. The creators who win are not the ones who shipped the most polished version one; they are the ones who built a loop – design, deliver, measure, improve – and never let it close.