How to create an online course: the complete build guide.
Seven phases in build order, from the transformation you are actually selling to the four numbers that tell you what to change before the second cohort. Written to work with any course platform; where a tool decision genuinely changes the plan, it says so.
What you will have at the end
- A transformation written as one sentence you can sell and be judged against
- Evidence that somebody will pay for it, gathered before you record anything
- A curriculum where every module ends in something the student can point at
- A recording plan that produces usable lessons rather than a perfect first one
- An access rule — open or dripped — chosen deliberately and stated in a sentence
- A price argued from the outcome, and a launch with a real deadline behind it
- Four numbers, defined precisely, that tell you what to fix before cohort two
Name the transformation, then find out if anyone will pay for it.
Most courses that fail were decided badly in week one and discovered in month four. The author picked a subject they know well, built forty lessons about it, and only then went looking for someone who wanted them.
A course is not a body of knowledge with videos attached. It is a change of state for a specific person: they could not do something, they worked through your material, and now they can. Everything downstream — the outline, the price, the sales page, the metric you judge it by — is an argument about that change. If you cannot state it, you cannot outline it, and you certainly cannot sell it.
Write the end-state sentence
One sentence, in this shape: by the end, [a specific person] can [do a specific thing], and they will know it worked because [observable evidence]. Three tests on it. Is the person specific enough that you could name two of them? Is the capability something they perform rather than something they know — “understand pricing” fails, “set a price and defend it on a call” passes. And is the evidence something the student can see without asking you?
Then write the second sentence nobody wants to write: who this is not for. It feels like turning away money and it is actually how you keep the course finishable, because every audience you refuse to serve is a set of lessons you no longer have to build.
Validate before you build
Recording is the most expensive and least reversible part of this whole process. Do it last, and only after something other than your own enthusiasm has told you the course should exist. In rough order of how much the signal is worth:
- Sell the outline. A detailed curriculum, a date the first module opens, a price, and a checkout that takes money. The strongest signal available, because it is the only one where somebody spends something. Build it after it sells, with buyers waiting and questions arriving.
- Teach module one live, once. Run the first milestone as a paid workshop or a live session. You find out which explanation lands, which prerequisite you assumed, and where the room goes quiet — all of which is cheaper to learn before it is recorded.
- Have five conversations with the person you are building for. Not a survey. Ask what they have already tried, what they gave up on, and what they would need to see to believe it is different. You are collecting their words, which become the sales page and half the lesson titles.
- Watch what they already buy. If people in this audience are paying for adjacent things and not for this, that is information. If they are paying for this and complaining about it, that is a better brief than any survey answer.
Counts as validation
- Somebody paid before the lessons existed
- You can quote three sentences of what a buyer said they wanted, in their words
- The people saying yes have the problem now, not eventually
- You know what they tried before you and why it did not work
Does not count
- A poll where people said they would be interested
- Compliments on the idea from people who are not the audience
- A waitlist that has never been asked for money or a date
- Search volume for the topic, with nobody in front of you who has the problem
The strongest of these is preselling: a detailed outline, a start date, a price and a working checkout, offered before a single lesson is recorded. How to presell an online course covers what you can honestly promise, what to say about the delivery date, and how to handle the refund question when the product does not exist yet.
Modules are milestones, not chapters.
The outline is where a course is really made. It is also the only phase where being wrong costs you an afternoon rather than a month of re-recording, which is exactly why it deserves more time than it usually gets.
Work backwards from the end-state sentence. What has to be true immediately before a student can do that thing? And before that? Keep going until you reach the state they are in on the day they buy. Those steps, in reverse, are your modules — and each one should carry a milestone sentence of its own: by the end of this module the student can do a smaller specific thing, with something to show for it.
The test is blunt. If you cannot write the milestone sentence for a module, that module is a topic rather than a step, and topics are why courses feel long and finish empty. A first course usually lands somewhere between four and eight modules — not because there is a correct number, but because fewer than four rarely spans a real transformation, and more than eight is usually two courses sharing a title.
Complete beats exhaustive
Complete means nothing is missing that the transformation requires. Exhaustive means nothing is missing that you happen to know. The second is the expert’s instinct and it is what turns a six-hour course into a twenty-hour one that nobody finishes. For every lesson, ask whether removing it would stop a student reaching the end state. If it would not, it is a resource, an appendix or a future course — all of which are fine, and none of which belong in the required path.
The full method — module math, lesson sizing, the sequencing rules and a copyable template with bracketed placeholders — is the sibling guide: how to write a course curriculum outline. Come back here when the skeleton exists on one page.
Do not record until the skeleton is stable
The skeleton is: the end-state sentence, every module milestone, every lesson title, and one line per lesson describing the single idea it teaches and the action it asks for. That document is the course. Read it top to bottom and you should be able to feel the student getting better. If you cannot, move things now, while moving them is free.
Record the version that teaches, not the version that impresses.
Production value is the thing creators obsess over and students barely notice. Clarity, audio and pace are what they notice. Optimise for the ones being judged.
- Record a throwaway lesson first. Pick something from the middle of module three and record it knowing you will delete it. It calibrates your setup, your pace and your nerves without costing you a lesson you needed.
- Batch by setup, not by module. Every camera-to-screen switch costs you fifteen minutes of fiddling. Record all the talking-head pieces, then all the screen recordings, then all the walkthroughs, regardless of where they sit in the curriculum.
- Bullets, not a script — with two exceptions. Read-aloud scripts sound read aloud. Work from the one-line idea and three bullets instead. The exceptions are the first sixty seconds of a lesson and any explanation you have historically fumbled: write those word for word.
- Do not restart the take. When you stumble, pause, breathe, and say the sentence again from the start of the sentence. Editing out three seconds is trivial; re-recording eleven minutes because minute nine went wrong is how courses stall.
- Date nothing. No “as we head into next year”, no current prices, no screen recordings of an interface that changes weekly, no references to a news event. Anything dated is a re-record you have scheduled for yourself without knowing it.
- Record module one last. You will be visibly better by the end, and module one is the part every student watches. It is also the part most likely to change once the rest exists.
Captions and transcripts are not an afterthought. They serve people who need them, people watching without sound, and anybody who wants to find the one line they half-remember — and they are far cheaper to produce at record time than a year later. Course accessibility basics covers the minimum that should ship with every lesson.
Setup, microphone choice, screen-recording settings, the batching order and what to do when a lesson needs re-recording are all in how to record course videos.
Assemble the modules, the portal and the access rule.
Three decisions turn a folder of videos into a course: how it is structured, what the student walks into, and when each part becomes available.
Structure students can navigate
Name modules for the milestone, not for the subject: “Set your price and defend it” tells a student where they are; “Pricing theory” tells them nothing about their own progress. Number lessons within modules so that “I am stuck on 3.2” is a sentence somebody can say to you. Put the resources a lesson needs on the lesson, not in a separate library the student has to go hunting through.
The portal is one question: what now?
A student arriving for the second time should see, without scrolling or thinking, the next thing they are meant to do. Resume where they left off, what percentage of the course is behind them, and one obvious way to get help. Everything else on that screen is competing with the lesson they were about to watch.
The first forty-eight hours after purchase decide the start rate, and start rate is the metric that quietly determines every other one. Student onboarding emails covers the sequence that gets somebody from receipt to first completed lesson.
Choose the access rule deliberately
Drip versus open is usually decided by accident, on the platform’s default setting. It is a real trade and worth thirty seconds of thought:
| Model | What it is, and what it costs you |
|---|---|
| Open on purchase | Every module is available the moment the student enrolls.Respects the buyer who cleared a weekend for this, and is the only honest model if the sales page implied immediate access. Costs you the pacing: a student who opens all of it at once tends to sample all of it at once, and sampling is not finishing. |
| Dripped on a schedule | Modules unlock on a fixed cadence from the enrollment date — one a week is the common shape.Paces the work and gives every student the same rhythm, which makes support predictable and gives your emails something true to say. Costs you the fast student, who wanted the whole thing and now has a reason to ask for a refund. |
| Dripped on progress | The next module unlocks when the current one is complete, whenever that happens.Keeps the prerequisite chain intact without punishing pace. Costs the most to operate, because a student stuck at a gate is invisible unless something is watching progress and telling you. |
Whichever you pick, say it on the sales page in the words you will use in the product. A student who expected everything and found a drip schedule has been surprised in the one way that reliably produces refunds. The wider version of this decision — whether the course runs as a cohort at all — is in cohort versus self-paced.
The course knows the progress. The email tool knows the person.
On a stitched stack those are two records, so “everyone who stalled in module two” is an export and a manual import rather than an audience. In Agorix, lesson activity, assessments, milestones and completion sit on the same customer record as the purchase and the event that preceded it — see how courses work.
Price the outcome, not the running time.
Hours of video is the worst available basis for a price. It rewards padding, punishes editing, and describes your effort rather than their result — and nobody has ever bought a course because it was long.
Two questions set the range. What is reaching the end state worth to the person buying — in money, time, risk avoided or a decision they can finally make? And what is their alternative: doing nothing, doing it alone, hiring someone, or a different course they are already considering? The price lives between what the outcome is worth and what the alternative costs.
Illustrative arithmetic, with your own numbers
Not a benchmark and not a forecast — just the shape of the trade, which you should redo with figures you believe. Suppose you want a first launch to produce ten thousand dollars. At two hundred dollars you need fifty buyers; at a thousand you need ten. Fifty buyers means fifty onboarding paths, fifty support relationships and a promise that has to work for a wide range of people. Ten buyers means ten conversations you could actually have. Now change the numbers to yours and ask which audience you can genuinely reach, and which version of the support load you can genuinely carry.
The model follows the value
A one-time payment fits a course that delivers a bounded transformation. Instalments keep the same price reachable and add collection work you should be ready for. A subscription only makes sense when you are genuinely adding value every month — sold against a static library it produces cancellations and resentment in about the same period. Subscription versus one-time pricing works through where the line sits.
Your refund policy is part of the price. A generous, clearly written guarantee lowers the risk of buying and raises the standard you have to meet; refund policies and guarantees covers how to write one you will still be happy to honour in month six. The full pricing argument — anchoring, tiers, launch pricing and what to do when nobody buys — is in how to price an online course.
A launch is a date, an audience and a reason to act now.
Publishing a course is not launching one. A launch has a window that opens and closes, a group of people who hear about it more than once, and something that makes waiting the worse option.
For a course, the highest-yield launch mechanism is usually a live teaching session: you teach the method well enough that a reasonable person concludes they want help executing it, then you make the offer plainly. The teaching is real and the conclusion is honest — that is the whole mechanism, and it works because the session is the course’s free sample rather than a disguised pitch.
- Pick the window and the deadline. Enrollment opens on a date and closes on a date, or the price changes, or the cohort starts and latecomers wait. Whatever the mechanism, it has to be true — an extended deadline teaches your audience that deadlines are decorative, and you only get to teach that once.
- Run the teaching session. Live, at least the first time. You need to hear the questions, find out which explanation loses the room, and watch where people leave before you automate any of it.
- Send the sequence, not the announcement. One email is an announcement. A launch is a sequence with distinct jobs per send, and buyers removed from it the moment they purchase.
- Answer objections in public. The questions people ask before buying are the questions everyone else is thinking. Put the answers where the undecided can read them rather than replying privately five times.
- Close it, and mean it. When the window closes, close it. Then write down what happened while you still remember which email did the work.
The session structure, the offer transition and the follow-up split are in selling courses with webinars; the emails themselves — open, mid-window, objection, last call and the post-close handling — are written out in the course launch email sequence. If you are running the live session for the first time, how to host a webinar is the operating guide for the session itself.
No audience yet? The launch does not change shape, only scale — launching without a list covers borrowing an audience rather than waiting to build one.
The course itself
The portal and access
Checkout and terms
The launch mechanics
The first cohort tells you what the second one should be.
A course is not finished when it is published. It is finished when the people going through it stop getting stuck in the same place — and the only way to find that place is to watch where they stop.
Read the drop-off as a location
When a run of students stops at the same lesson, that lesson is rarely the problem on its own. Usually it is the first lesson that demands something the previous ones did not prepare them for: an assumed prerequisite, an unexplained step, or simply the point where the work stops being watching and starts being doing. Fix the run-up, not just the lesson.
Two other signals are worth as much and cost less to collect. A question asked three times is a lesson you did not build — put the answer in the curriculum rather than in your inbox for a fourth time. And a student who finishes and cannot show the evidence from your end-state sentence means the course taught the subject but did not produce the change, which is a sequencing or an assessment problem rather than a content one.
Four numbers, defined precisely
Definitions travel between businesses; numbers do not. Record these four after the first cohort and compare the second cohort only to the first.
| Metric | Definition and what it judges |
|---|---|
| Enrollment rate | Enrollments ÷ visitors to the sales page, tracked per traffic source. It judges the page and the promise and nothing else — at the moment somebody enrolls, nobody has seen a single lesson. Averaged across sources it will hide the one channel that was actually working. |
| Start rate | Students who complete at least one lesson ÷ enrollments. The most neglected number on this list and usually the most fixable: it judges the first forty-eight hours — the receipt, the access email, whether the login works and whether lesson one is short enough to start. A student who never starts almost never returns on their own. |
| Completion rate | Students who finish the final required lesson ÷ enrollments, measured inside a fixed window you choose — thirty, sixty or ninety days — and stated every single time you quote the number. Without the window it is not a metric, and it will drift upward on its own as older cohorts trickle in. |
| Refund rate | Refunds ÷ purchases, split by whether the student had started. A refund from someone who never logged in is an onboarding failure; a refund from someone who finished module three is a promise failure. They have opposite fixes, and one blended number hides both. |
Then act on the diagnosis rather than the total. Completion is a system you build, not a personality trait your students have — course completion systems covers the mechanics, and certificates and assessments covers giving students a reason to reach the end and a way to prove they did. For getting the words out of students in the first place, see student feedback loops.
Finishing is a signal, not a trophy.
A student who finished has told you something specific: they did the work, they got the result, and they are now a different person than the one who bought. That is useful context for what you offer next — an advanced program, a cohort, a conversation. It is not proof that they are ready to buy again, and treating it as an automatic qualification is how a good course ends in an unwelcome pitch.
Six questions before you build.
How long should an online course be?
As long as the transformation requires and no longer. Running time is a cost you impose on the student, not a measure of what you are giving them — and a course that gets finished is worth more to both of you than a longer one abandoned in module two. Write the end-state sentence first, then include only what a reasonable person needs to reach it. The curriculum outline method is mostly a set of rules for deciding what to leave out.
Do I have to finish building the course before I sell it?
No, and finishing first is the more expensive order. Selling a detailed outline before the lessons exist tells you whether anyone will pay, and the people who do pay tell you which parts to build first. If it does not sell as an outline you have saved yourself the recording; if it does, you build with students waiting and real questions arriving. Preselling a course covers how to make that promise without over-promising a delivery date.
Should I run my course as a cohort or self-paced?
Cohorts add a deadline and a group, which is most of why they get finished; they also require you to be present on a schedule. Self-paced scales and never fills, which is why completion has to be engineered into it deliberately. Run the first version with a cohort even if the destination is self-paced — you need to watch people move through it before you automate it. The trade-offs in full: cohort versus self-paced.
What equipment do I need to record course videos?
A microphone better than the one in your laptop, a room that is not echoing, and light in front of you rather than behind you. Audio is the variable students notice and complain about; camera quality mostly is not, and screen recordings with clean narration are perfectly acceptable for most subjects. The recording guide covers the setup, the batching order and what to do about mistakes mid-take.
How much should I charge for an online course?
Price against the outcome and the alternative, not against the hours of video — length is not an argument for a higher price, and buyers do not experience it as one. Ask what solving this problem is worth to the person buying and what they would otherwise spend on it, then choose the model that matches how the value arrives. Pricing an online course works through the arguments and the payment models.
What is a good completion rate?
Published completion benchmarks average across subjects, price points, formats and audiences that have nothing to do with yours, so they cannot tell you whether your number is good — only whether it resembles an average you have no reason to care about. Define the metric precisely instead: finishers divided by enrollments, inside a fixed window you state every time you quote it. Then compare cohort two to cohort one. Definitions travel between businesses. Numbers do not.
The four phases worth their own guide.
You now have the plan. Agorix is the system that runs it.
The launch webinar, the checkout, the course, the student portal and the follow-up — on one customer record, so what a student did in module two is something your emails can act on rather than something you export.
14 days · $0 due today · Payment method required · Cancel before first charge and pay nothing