Is Selling Online Courses Profitable? 3 Years, 4 Ledgers

Editorial Note: This is a reconstructed field record, not a single real-world case study. The people and personal financial records are fictional, while the market data, platform information, and regulatory cases are drawn from public sources and used to examine the economics of selling knowledge online.

Is Selling Online Courses Profitable? Three Years, Four Ledgers

A field record from the course-and-community economy, 2023–2026

Lisbon, 2:47 a.m., March 14, 2023. Mara refreshed her dashboard for the third time. The cache hadn't cleared, so the number flickered before settling: $11,240 for the month. She made Notion templates and landing-page templates by herself. No team, no ad budget, no funnel. The only marketing she had ever done was a tutorial video she posted eight months earlier, one she privately thought was badly shot.

She assumed she was living inside a product story. Three years later, looking back, that money had come from an account she never intended to run, and the real accounting started that autumn.

Five people appear in this record. Mara, who builds templates. Jonas, who stands in front of the camera. Dev, who does the math. Priya, who buys the course. Plus Nadia, a platform data analyst who shows up twice and says very little. They did not know each other until the winter of 2024, when four of them ended up on the same community platform.

Every section below ends at the moment where one of them had to make a decision. External figures are attributed with sources and years. Modeled figures are labeled as modeled, while the personal records in the narrative are fictional.

1. Is Selling Online Courses Profitable? Where the First $10,000 Comes From

Mara's templates sold for $19. After platform fees and payment processing, she kept a little over $15. So $11,240 was roughly 600 units. She opened her traffic report: 78% of it traced back to that badly shot video. The rest scattered across three sources she could not name.

That night she went looking for where she sat in the distribution. Gumroad disclosed figures in 2020: more than 46,000 creators earning a combined $142 million, which works out to about $3,087 per creator. The version of that dataset that circulates widely online adds claims about extreme concentration at the top, but those come from platform self-reporting and secondhand retellings, and the year is old. She treated it as a shape, not a fact. The coordinate she could actually verify came from Goldman Sachs Research in 2023: the creator economy at roughly $250 billion, projected toward $480 billion by 2027; around 50 million creators worldwide, of whom only about 4% earn more than $100,000 a year, a share the firm expected to stay flat even as the market grew.

After tool subscriptions, taxes, and currency conversion, the $11,240 landed as under $8,000. She screenshotted it, filed it in a folder named 2023-03, and closed the laptop.

The next morning she did something small. She exported the buyer emails: 587 addresses. She sent the same one-line message to 20 of them — "What problem were you actually trying to solve when you bought this?" Six replied. Four of the six described the same thing. They had not wanted a template. They had wanted someone to tell them what to do next.

Her first decision. Keep selling files, or start selling a person — show her face, run an account, turn buyers into an email list. Files have no repeat purchase. A person can be reached again. She chose the second path. In 2023 it looked like a marketing choice. By 2025 it had determined the shape of everything that followed.

She had figured out the revenue. She had not yet figured out who was buying.

2. Why Product Sellers Pivot to Communities

By April 2024, Mara's template sales were down 41% year over year. The cause was not mysterious. AI tools could generate structurally similar work in minutes, and her price had slid from $19 to $9 to "free, just leave your email." She was not alone; that month at least thirty people in her seller group were describing the same slide.

That May, in London, Dev was hired to run the numbers for Jonas. Jonas worked out of Austin, selling business courses to freelancers, and had cleared six figures on a single livestream in 2022. He had two options on the table: cut the price and chase volume, or raise the price to $1,800 and run a small cohort plus a members' community.

They started by writing market size on a whiteboard, and produced three numbers that contradicted each other. Dev cited one research aggregator: the online course market at roughly $295.7 billion in 2025, reaching $1.27 trillion by 2034. Jonas pulled up another: eLearning at $495.6 billion in 2025, reaching $1.28 trillion by 2032. A third source gave $324 billion growing to $2.04 trillion by 2035. The statistical boundaries were incompatible, and none of the three could disprove the others.

Dev erased all three lines and wrote one sentence underneath: the only thing these numbers have in common is that each one is being used to support a conclusion someone already had. He switched to a denominator he could check.

The ICF 2025 Global Coaching Study, conducted by PwC across 127 countries with more than 10,000 respondents, reported 122,974 practicing coaches, up 15% from 2023, with industry revenue of $5.34 billion. Divide one by the other and you get roughly $43,424 in annual revenue per coach — revenue, not profit, and it includes enterprise work. In the same study, 60% of coaches delivered training alongside coaching, 59% expected growth the following year, and that growth came from client count and hours, not from price increases.

That same week, Nadia was looking at a different set of numbers on her platform's dashboard. Total paying communities and total gross revenue were both climbing fast. The three-month survival rate for newly created communities was falling. The aggregate curve and the individual distribution were moving in opposite directions. On her dashboard that was two lines. On the timelines of the people doing this work, it was one.

Dev's decision. Go high-ticket. The reasoning was technical: low-ticket volume required ad efficiency, and acquisition costs had already eaten the margin; high-ticket could trade human delivery for margin and needed no ad budget up front. The report looked better. It was also easier to lose control of.

3. How Paid Communities Make Money: Two Spreadsheets

In October 2024, after the same autumn launch, Dev and Jonas each opened a spreadsheet.

Dev's: 214 enrollments, $1,800 per seat, $385,200 gross. After the modeled platform fees, acquisition costs, refund reserve, and delivery labor, the remaining contribution was approximately $290,000 before taxes and other overhead.

The difference was which rows Dev and Jonas chose to count. Dev's version broke out platform fees, a refund reserve, teaching-assistant labor, and refunds likely to arrive in the next 90 days. Jonas's version treated refunds as a minority problem and deferred them to next quarter. Dev printed his sheet and taped it to the wall. The platform fee was a published rate. The rest were his own assumptions, written in pencil with assumed in the margin.

Line item	Value	Nature
Price per seat	$1,800	Pricing
Platform and payment fee	3.9% + $0.30 per transaction above the $900 threshold	Skool published rates
Customer acquisition cost	~$210 per person	Modeled
First-month churn	18%	Modeled
6-month renewal rate	62%	Modeled
Refund reserve	4% of gross	Modeled
Delivery labor per person	~$90	Modeled

Only the platform fee in that table is a published rate. The rest are pricing assumptions or modeled values, labeled so they do not read as fact. What is interesting is what the two published rows imply. Skool's Hobby plan is $9/month plus a 10% transaction fee; the Pro plan is $99/month plus 2.9% + $0.30 under $901 per transaction, and 3.9% + $0.30 above it. For transactions below the Pro plan's high-ticket threshold, the two cost curves cross at roughly $1,268 in monthly member revenue, before considering other plan features. Below that level, Hobby has the lower combined monthly cost; above it, Pro generally has the lower transaction cost. High-ticket transactions above $900 change the calculation because the Pro fee rises to 3.9%. At $9,800 a month, assuming the transactions are above the Pro plan's $900 threshold, the variable fee alone is approximately 3.9% on Pro, while the Hobby plan remains at 10%. The fixed monthly subscription and the $0.30-per-transaction charge make the true effective rate slightly different.

Dev read that crossover point to Jonas over the phone. Jonas went quiet for a few seconds, then said: "So the platform fee is already sorting us by size for us." There was no emotion in the sentence. It described a fact: small sellers pay a percentage, large sellers pay a flat fee plus a lower percentage, and the two are not starting from the same line.

On Dev's desk was a PDF, the FTC's public record in the Digital Altitude case. One allegation was stated plainly: the firm had promised consumers "one-on-one coaching by successful marketers," and those people were in fact salespeople whose job was to sell higher-priced membership tiers. The FTC said most consumers never earned the income promised, that consumer losses ran into the tens of millions, and that some individuals lost more than $50,000. One defendant, Sean Brown, settled on condition of surrendering specified assets, with a judgment of $10.8 million. The document contained no moral judgment. It contained allegations, settlement terms, and amounts. Dev left it on the desk as a reminder of where the line sat.

Dev's decision. In November, the renewal rate fell from 62% to 47%. Two options: add delivery (hire assistants, compress response times to four hours, and take the hit to margin), or add traffic (fill the top of the funnel and dilute churn with new volume). He chose traffic.

Three months later the problem had changed from "we can't sell enough" to "too many are asking for their money back."

4. Do People Finish the Courses They Buy?

Priya lived in Toronto. At 11:41 p.m. on November 17, 2024, she received a payment confirmation: $1,800, on a credit card, in installments, for Jonas's cohort. She had been laid off six weeks earlier. She had sent 74 applications and gotten 3 interviews.

Her own record, which she later typed into a notes app: week one, three lessons finished. Week two, one lesson. Week three, none. Week five, she posted a question in the group chat; nobody answered for 48 hours, so she deleted it herself. Week nine, she started to suspect she was the problem. Week twelve, she stopped opening the platform.

What she did not know was a piece of statistics. Katy Jordan, at the Open University in the UK, analyzed 29 large-scale MOOCs and found completion rates averaging 6.8%, with a range from 0.8% (Princeton's A History of the World since 1300, 83,000 enrolled) to 19.2% (EPFL's Functional Programming Principles in Scala, around 50,000 enrolled). Those were free open courses, so the figure should not be treated as a benchmark for paid courses. It does, however, show why enrollment and completion are very different measures.

The pain for Priya was not the unfinished course. It was that she had read a distribution as a personal defect. That misreading costs almost nothing to make, requires no evidence to sustain, and conveniently removes any obligation on the seller to explain.

Priya's decision. In week fourteen she requested a refund. The terms gave a 14-day window, long expired. Support offered her something else: join the partner program, refer three paying members, and the tuition gets offset. The commission rate was disclosed, and the arrangement was presented to her as a referral program. She signed.

She referred two people. One of them was her cousin, 23, freshly graduated.

5. When the Refunds Come: Chargebacks, Enforcement, and the Line

On February 6, 2025, three things arrived in Dev's inbox on the same day: a risk notice from his payment processor (chargeback rate up four weeks running), an inquiry letter from a consumer protection agency, and a refund ticket from Priya, numbered R-2025-0418.

The letter accused no one. It asked two things: what is the basis for your earnings claims, and can you produce evidence.

Dev ran a cold audit that day. He listed every earnings figure he had published in the previous 12 months: 37 claims. Nine had retainable student data behind them. Twenty-eight came from screenshots, verbal retellings, or "one student." It took him three days to take all 28 offline, including the one on his homepage with the highest conversion rate.

He wrote three lines from public records on sticky notes and put them on the edge of his monitor.

Lurn. The FTC sued in September 2023, alleging unsubstantiated earnings claims — that consumers could become "stay-at-home millionaires," or "fail 98% of the time and still make $11,453 per month." In 2024 the FTC distributed more than $2.4 million in refunds to 1,922 consumers (roughly $1,249 each, derived).
Digital Altitude. Judgment of $10.8 million, deferred on condition of surrendering specified assets. The structure the FTC identified: a promise of one-on-one coaching that was in practice an upsell.
IM Mastery Academy. The FTC and the State of Nevada alleged unsubstantiated earnings claims used to sell financial training and multi-level marketing opportunities, with consumer losses exceeding $1.2 billion since 2018. The two highest-compensated salespeople in that program took more than $36 million between them.

The third note was the one he stared at longest. It made no moral argument. It showed where the money ended up: not with students, with salespeople.

Priya's ticket moved to "partial refund, $600" on day 22. In the group chat she saw eleven other tickets with adjacent numbers. Her cousin's ticket was never refunded.

That summer, in an internal review, Nadia said something that made it into the minutes: complaint volume on the platform was rising roughly linearly with the number of communities, while the average complaint count per community was falling. The problem was spreading rather than concentrating. The sentence pointed at nobody's character. It described a distribution getting wider.

That year, three people made three different choices.

Dev contracted. He cut the $1,800 tier, kept a $297 course and a $49/month community, pulled his promises back inside his delivery capacity, and rewrote every earnings claim into a version he could document. Jonas scaled harder, putting the entire budget into content and audience, betting he would not be the one who got noticed; his follower count tripled that year. Mara took a third road: in the autumn of 2025 she stopped selling templates and stopped selling courses, and began building industry-specific templates and white-label solutions for agencies and companies — a few thousand to tens of thousands of dollars per engagement, where the client was buying two saved weeks and delivery certainty.

6. Who Actually Makes Money: Positions, Not Personalities

In May 2026, four of the five were in the same room at a forty-person meetup in Austin. The way each of them reported their numbers was itself information.

Mara reported invoices: about $180,000 in annual revenue, 11 clients, the largest accounting for 34%, 30-day payment terms, and not one dollar coming from platform-recommended traffic. Her curve was not steep. It barely depended on any single platform's rules.

Jonas reported gross: a record Q4 in 2025, a 23% renewal rate in the first half of 2026, and refunds plus chargebacks consuming roughly 60% of new gross margin. His audience was still growing. He called that an asset.

Dev reported cash flow: about $60,000 in 2026, lower than 2024, but with a standing refund reserve and every earnings claim substantiable. "I got smaller and I sleep fine," he said, then added the technical version: his CAC payback moved from 2.1 months to 5.4 months, and the churn curve finally flattened.

Priya reported net: $1,100 refunded in total, $700 of it still hers plus interest, and with her cousin's ticket included, she still believed she had simply failed to study hard enough. By 2026 she was running operations at a company and, on the side, a $15/month community with 40 members. She replied to every message within 12 hours.

Nadia was not there. Her KPI was retention, not anyone's income, and her platform's fee revenue grew over the same period.

Lay those five positions out and the people who make money fall into roughly five categories. Distribution earners own their traffic or email list and do not depend on one platform. Trust earners have enough retention to recover acquisition costs over the expected customer lifetime. Delivery earners match price to labor cost and can substantiate every promise. Fee earners are platforms and processors, who take no creative risk. Information-gap earners depend on claims or information advantages that may become harder to sustain as evidence requirements and enforcement tighten.

Which category lasts comes down to one ratio you can compute: renewal rate × price per seat, versus acquisition cost + delivery cost + refund reserve. There is no moral term in that expression, and it tends to expose an unsustainable structure within six to nine months. Dev's drop from 62% to 47% was the first time it spoke.

Dev later printed a single page and handed it to anyone who came to him asking whether to do this. The header read: Answer these ten questions before we discuss whether to build it.

In the last 90 days, how much of my traffic came from channels I control (email list, owned account, search)? Below 40% means single-platform risk.
What is my three-month renewal rate? Use members whose term expired in that window as the denominator, not total members.
Do I hold a standing refund reserve? What percentage, and on what basis?
For every earnings claim I make, can I produce retainable student data?
How are delivery promises (response time, coaching sessions, Q&A frequency) written in the contract, and what is my actual hit rate?
Do platform and processing fees rise or fall as monthly revenue grows? Where is the breakeven?
What share of revenue comes from my largest single client or channel?
If this platform changes its fee split or rules tomorrow, what happens to my cost structure?
How many months does CAC payback take? What does it become if renewal drops 15 points?
Can I deliver what I sell, inside the promised window, to the slowest student in the cohort?

These ten questions have no standard answers. They do one thing: they turn "is this profitable" from a feeling into a set of numbers you can actually run.

References

Goldman Sachs Research, 2023 The creator economy could approach half-a-trillion dollars by 2027 https://www.goldmansachs.com/insights/articles/the-creator-economy-could-approach-half-a-trillion-dollars-by-2027
Gumroad, 2021 Last Year in the Creator Economy https://gumroad.gumroad.com/p/last-year-in-the-creator-economy
International Coaching Federation, 2025 2025 ICF Global Coaching Study https://coachingfederation.org/resources/research/global-coaching-study/
Federal Trade Commission, 2018 Defendant Who Took Part in Business Coaching Scheme Agrees to Settle FTC Charges https://www.ftc.gov/news-events/news/press-releases/2018/09/defendant-who-took-part-business-coaching-scheme-agrees-settle-ftc-charges

Menu