The Revenue Record of 9 Udemy Courses — the First Took 100 Hours and Made $165, and What It Taught Produced a "¥1.5 Million Course"

Compared with blogging or YouTube, the revenue structure of Udemy instruction has extremely little published data. Revenue shares and how sales events really work are invisible until you become an instructor, so “does it pay or not” debates drift into impressions. The note by Naoto Takizawa, a course creator who makes his living producing and selling digital content, is a rare primary source that fills this gap with real data from 9 courses. You can follow, in numbers, the path from a first course that took 100 hours and ended at $165 to a successor course, built on that learning, that generated over $10,000.

Two years of instructing, in numbers

The starting point: ¥20,000 for 100 hours

In January 2024, Takizawa invested about 100 hours doing everything himself (slides, recording, editing) and published a copywriting course. The result: $165 (about ¥20,000) over more than a year. Converted to an hourly wage, roughly ¥200. Taken in isolation, this number says “quit.”

But he reflects, “Without that experience I would never have made 9 courses.” The first course’s real deliverable was not revenue but learning. At its core was the realization that “what I want to make” and “what the market wants” are different things, which, he says, fundamentally changed his course design from then on. What concretely changed: thorough market research, an overhaul of audio quality and structure, and designing the virtuous cycle in which reviews bring the next enrollments. More reviews mean more exposure in Udemy’s recommendations, and exposure brings further enrollments and reviews, compounding that runs inside the platform.

The turning point came around course three. Students began to browse, “let me look at this instructor’s other courses.” A student who takes one course peeks at the same instructor’s others. The more courses, the lower the acquisition cost per course, the same catalog compounding as the “build the shelf” strategy of Kindle publishing. Behind the 16x growth in monthly students, from 34 in the first month to 563 in about a year, is not a one-off hit but this browsing structure.

97% and 37% — the price tag on customer acquisition

Udemy’s revenue share is determined by traffic source. Sales the instructor brings in through their own referral link pay 97% to the instructor. Sales via Udemy’s internal search and recommendations drop to 37%. As Takizawa puts it, “the difference in revenue comes out to roughly 3x.” The same lecture nets about 3x more or less depending on who acquired the customer. This gap is a price list proving that “acquisition is the most expensive step in the chain.” An instructor’s real capability is the product of teaching skill and owning acquisition channels (social media, an email list).

The other gravitational force is sales events. The list price is set around ¥20,000, but 98–99% of purchases happen during sales at ¥1,500–2,000. New enrollments run about 10 per day normally and jump past 50 during the year-end sale. The list price is merely an anchor. Business math can only be done on the effective unit price of sale price × revenue share.

Interestingly, Takizawa himself does almost no external acquisition. Rather than chasing the 97%, he chooses to ride Udemy’s acquisition power even at 37%. The scale of 563 students a month is also the result of designing around the platform’s gravity.

“Build it once and it earns on autopilot” — half wrong

His current maintenance load is about 1 hour a month (answering questions, updating content, checking reviews) which he describes as “close to passive income.” At the same time, he cautions that the received wisdom of “build it once and it earns while you ignore it” is “half true, half not.”

Points to discount. Of the 9 courses, only the two extremes have concretely disclosed revenue ($165 and over $10,000). The performance of the other 7 is unknown. The portfolio’s median cannot be known. Note also what sits before that $10,000+ figure: the sunk production time of 9 courses, including the first course’s 100 hours. Recovery is an “if it takes off” story. The time invested until it does is a fixed cost. Revenue shares, sale design, and recommendations, for their part, are all variables on Udemy’s side. Dependence on sales events for 98–99% of purchases is synonymous with surrendering pricing power, and the structural risk that changes to rates or terms hit revenue directly never disappears.

One more point: the review flywheel spins in reverse, too. Low ratings early on drop a course out of recommendations, and the exposure needed for a comeback never arrives. That Takizawa put audio quality and structural overhaul at the top of his improvements reads as a response to this irreversibility.

Converting 563 students a month into yen

The published data allows a rough back-of-envelope on the order of revenue. If 98–99% of purchases happen during sales (¥1,500–2,000) and the Udemy-sourced revenue share is 37%, the take per student is ¥555–740. At 563 students a month, that is a range of ¥310,000–420,000 per month. In reality it will not land exactly there because of the 97%-share referral-link portion and price differences between courses, but the sense of scale, “over 500 students a month gets you to this level”, is consistent with the ¥481–666 per student that Isobe discloses. Udemy instruction is not a business that earns on unit price. It is a “volume business” that stacks up headcount by riding sales events and recommendations.

Conditions for replication

The piece worth taking away is the design theory. Budget the first course as the cost of market learning, not revenue. Plan multiple courses on the premise that browsing, the catalog effect, starts to kick in around course three. Estimate sales not at list price but at the effective unit price of sale price × 37%. The real data from Isobe, who runs 22 courses, discloses an effective unit price of ¥481–666 per student, and this way of estimating matches almost exactly. That two independent disclosures, Takizawa with 2 years of instructing and Isobe with 5, produce the same numbers for revenue shares (97%/37%) and sale dependence is itself corroboration that these are platform specifications, not personal impressions.

Initial velocity, on the other hand, does not carry over. The first-month figure of 34 students depends on the category’s competitive density and demand size. Takizawa is also a full-time content producer and seller. Repeating 100-hour-class production nine times, converted into a side hustler’s disposable time, becomes a 2–3 year plan.

Related reading

• 22 Udemy courses earning $2,124/month — real data from the volume side
• 24 Kindle books — the book version of “shelf compounding”