How to Create an Online Course That Sells (and Teaches)

To create an online course that sells, start with a specific problem that people already spend money or time trying to solve, promise a result you could watch a learner achieve, and get people to pay before you build. Then build the course backwards from that result, with practice in every module, and run it with a small paying group before you open it to everyone.

Most guides to selling a course stop at the first half of that sentence: the idea, the price, the sales page, the launch. Those matter, and we cover them here. But the guides that rank for this query say almost nothing about whether buyers learn anything, and several repeat claims the research doesn’t support. That’s a commercial mistake as well as a teaching one. A course keeps selling when the people who bought it got the result, because they’re the source of your reviews, your referrals and your second course’s first buyers. A course that doesn’t deliver sells once, to the people who believed the sales page.

This guide walks through seven steps, with one invented but typical example throughout: a spreadsheet trainer building a self-paced course that teaches small-business owners to build a 12-month cash-flow forecast.

1. Choose a problem people already pay to solve

A topic is not a course. “Excel for small business” is a topic; nobody can tell from it what they’ll be able to do afterwards, so it’s hard to buy and easy to postpone. A problem is something your audience already loses money, time or sleep over: “I never know whether I can afford to hire.”

Look for evidence that people already pay to solve it. They hire an accountant to do it for them, buy templates, sit through webinars, or ask the same question in forums every week. If nobody is spending anything on the problem now, a course is unlikely to be the first thing they spend on.

Then talk to five people who have the problem. Ask what they’ve tried, where they got stuck, and what it would be worth to them to have it solved. This is the analysis phase of the ADDIE model, and it’s the step creators skip most often, because they assume their expertise tells them what learners need. It tells you what’s true. It doesn’t tell you what learners are missing.

For our trainer: the interviews show that business owners can use a spreadsheet; what they lack is a structure for a forecast and the habit of updating it. That shifts the course away from spreadsheet skills and towards a template, worked examples and a monthly routine.

2. Promise a result you could watch someone achieve

Your course promise and your top-level learning objective should be the same sentence. Write it so that you could watch a learner do it and check the result: “By the end of this course, you’ll have built a 12-month cash-flow forecast for your own business and used it to test one decision, such as a new hire.” Compare “understand cash flow”, which no buyer can judge and no course can test. Our guide to writing learning objectives covers the method.

A testable promise does two jobs. It tells buyers exactly what they’re paying for, which is what a sales page needs. And it tells you what the course must contain, and what it can leave out.

Be careful with promises about money. “Build a forecast” is a result you can deliver. “Double your profit” depends on things your course doesn’t control, and income or earnings claims draw particular attention from consumer regulators.

3. Validate with money before you build

Likes, survey answers and “I’d definitely buy that” are not validation. Payment is. Before you record anything, offer the course in one of three ways:

• A pre-sale. A short page describing the result, the outline, the format and the delivery date, with a refund if you don’t deliver. If too few people buy, you’ve saved months of production.
• A paid pilot group. Five to fifteen people who pay a reduced price to take the first version, often live or with more contact from you, in return for honest feedback.
• A smaller paid offer. A single workshop or mini-course on the first step of the problem. If it sells and learners succeed, the full course has a ready audience.

Decide in advance what counts as a pass: a number of buyers, or a number of pilot places filled from people who aren’t friends or family. If you miss it, change the problem, the audience or the offer before you change your mind about the evidence.

One way to test interest in courses and features before dedicating time to them, is to create a pre-validation page where people register interest in particular courses or features. That data can then be used to shape where you spend energy and time in course building.

4. Design the final task first, then the practice

Now design the course, and start at the end. Decide how learners will show they’ve achieved the promise. For our trainer, the final task is the promise itself: learners build their own forecast from the template and submit the answer to one “what if” question, such as whether they can afford a part-time hire from March.

Work backwards from that task to the modules that get learners there, and give every module practice, not just content. This is where the evidence is strongest. In a large review of ten popular study techniques, Dunlosky and colleagues rated two as having high utility: practice testing (recalling or applying what you’ve learned, through quizzes or exercises) and distributed practice (spreading that practice out over time) (Dunlosky et al., 2013).

The same review gave re-reading and highlighting low utility ratings. For a course, that means a lesson learners only watch is weaker than one that asks them to do something with what they’ve just seen. In practice, we’d give each module:

• three to five short questions at the end that ask learners to recall or apply the key idea, with feedback that explains the right answer;
• one applied exercise on the learner’s own material (for our trainer, adding one month of real figures to the template);
• a few questions from earlier modules mixed in, so earlier ideas come back after a gap.

5. Build lean: short segments, planned before you record

Development is where most of the hours go, and where most of them are wasted. Two pieces of evidence are worth knowing before you set up a camera.

Plan videos in short segments. A study its authors described as the largest of video engagement at the time analysed 6.9 million viewing sessions across four edX courses (Guo et al., 2014).

The authors recommend segmenting videos into chunks of less than six minutes, and the edX producers they interviewed said planning before recording had the largest effect on engagement. Two cautions: the study measured how long people watched, not how much they learned, and it looked at university MOOCs rather than paid courses. You’ll see this study passed around as “videos over six minutes lose half their viewers”. It doesn’t say that. Treat it as evidence for planning short, focused segments, each followed by practice, not as a stopwatch rule.

Don’t build for “learning styles”. Several of the guides ranking for this query advise offering video, audio and text to suit different learning styles. A widely cited review found that very few studies had used a method capable of testing that idea, and concluded:

Choose formats for the content instead: a screen recording to show the forecast being built, a diagram to show how cash moves through a business, a downloadable template to practise on. Captions and transcripts are still worth providing, for accessibility and for learners who want to search or skim, but that’s a different reason from learning styles. How much to put in each segment, and how to pair visuals with narration, are covered in our library entries on cognitive load and dual coding.

Build one complete module first, with its practice and quiz, and put it in front of your pilot group before you build the rest. Problems in the design only show up once something is built, and they’re far cheaper to fix in module one than across eight modules recorded in the same style.

6. Pilot with paying learners, and measure the task

Run the first version with your pilot group, and watch three things:

1. Can they do the task? How many pilot learners complete a usable forecast and answer the “what if” question correctly? This is the evidence that your promise is true.
2. Where do they stall or stop? The practice questions most people get wrong, and the lessons where people drop out, point to specific fixes.
3. What do they say? A short survey at the end, and a follow-up a month later asking whether they’re still using what they learned.

Fix what the pilot reveals before a wider launch. The online course design checklist is a useful pass at this stage.

The pilot is also where your first honest testimonials come from, and the rules here are real. In the US, the Federal Trade Commission’s rule on consumer reviews and testimonials, issued in August 2024, bans fake reviews and testimonials, buying reviews, undisclosed reviews by insiders, and some review-suppression practices (FTC, 2024). In Australia, the ACCC’s guidance says reviews must reflect the genuine opinion of the person who experienced the product, warns against suppressing negative reviews, and says any incentive must be offered whether the review is positive or negative and be clearly disclosed (ACCC). In practice: ask every pilot learner, not just the happy ones; don’t make a discount depend on a positive review; and don’t present one learner’s exceptional result as typical. Rules differ by country and change, so check your own regulator before you publish testimonials. This isn’t legal advice.

Pilots are an important part of the course building process, and it’s important to build in time to test and validate that each course works and is logical to the user.

7. Choose a platform, set a price and launch

Platform. Choose the platform last, once you know what the course needs. Most creators compare platforms on price and marketing tools. Compare them on learning features too, on the plan you’d actually pay for:

• quizzes with feedback on each answer, question banks and retake rules, so practice can go in every module;
• assignments or file uploads, if learners submit work (our trainer needs learners to upload their forecast);
• reporting by learner and by question, so you can see who’s stuck and which questions confuse people;
• certificates and completion rules, if your buyers need proof of completion.

We haven’t yet published our reviews of Thinkific, LearnWorlds, Teachable and Kajabi, and plans change often, so check each of these in the platform’s current documentation before you commit.

Price. There’s no reliable benchmark for course prices; the figures you’ll see quoted mostly come from platforms describing their own customers. Price from the value of the result to the buyer, test it with the pilot group, and ask the people who didn’t buy what stopped them.

Sales page. Your design work has already written most of it. Lead with the promise, show the final task, list the modules with what learners will be able to do after each, include a sample lesson, and quote pilot learners in their own words. State your refund policy plainly.

Launch. Open to your list, network or audience first, with a deadline if there’s a real reason for one (a live cohort start date, for example). After launch, keep measuring the same three things you measured in the pilot, and update the course when the evidence says a module isn’t working.

Four mistakes that cost sales later

1. Selling a topic instead of a result. If a buyer can’t picture what they’ll be able to do, they won’t buy, and if they do buy, they can’t tell whether it worked.
2. Building everything before anyone has paid. A pre-sale or paid pilot is the only test of demand that means much.
3. All content, no practice. Learners who only watch tend to forget, and learners who forget don’t write good reviews.
4. Trusting popular statistics. Engagement percentages and “learning styles” advice travel fast from blog to blog. Check the original source before you design around them.

What to do next

Write your course promise as one sentence you could watch a learner achieve, then put it in front of five people who have the problem and ask them what it would be worth to them. If no one would pay, change the problem before you build anything.