¥4.61 Million a Year, Five Years After Starting to Learn Programming. The Full Three-Year P&L and Playbook of an Indie Education App

Indie-development revenue disclosures abound, but records that line up all three figures (income, expenses, and profit) for three years, and even show the line at which the developer went independent, are rare. The developer in this case, Kengo KOSAKAI, went from zero programming experience to ¥4.61 million in annual income in five years, let go of the cram school that had been his main business, and became a full-time indie developer. This is not a flashy hypergrowth tale; it reads as a life-size ledger of “surpassing the average company employee’s salary through indie development.”

Three years of profit and loss

The starting point was a cram school on hiatus

KOSAKAI began learning programming in April 2020. During the period when the cram school he ran was closed due to the COVID-19 pandemic, he studied HTML, CSS, JavaScript, Ruby, and Ruby on Rails on Progate, and says he had working software a month later. The origin was a homemade tool to ease the school’s own workload, first released while still in what he calls a “held-together-with-tape state” for his own use. He then broadened the target from teachers to students, running the web and smartphone apps solo (the service name is undisclosed). The design philosophy, in his own words: build things that people “can’t manage without.”

FY2025’s income of ¥4.61 million works out to roughly ¥380,000 a month on average, surpassing Japan’s average annual salary. On the strength of that, in March 2025 he transferred the cram-school business and went independent on indie development alone. Right afterward he traveled to Europe and visited Expo 2025 in Osaka — the disclosure extends even to how a stock-like income changed his way of working.

Revenue structure — two pillars: ads and paid membership

Revenue rests on two pillars: Google AdSense advertising and paid membership (subscriptions). Expenses barely moved (¥542,000 → ¥672,000 → ¥704,000) even as income grew sevenfold. As a result, the profit margin in FY2025 reached 85%. A web service with no inventory and no procurement, expenses don’t scale with income. The inherent strength of indie development’s P&L structure shows up directly in the numbers. Compressed into a single line, these three years are less a story of revenue growing sevenfold than of expenses that never moved. With spending parked around ¥700,000 a year, nearly every additional yen of revenue stacks up as additional profit.

Note his choice of payments infrastructure. Rather than wiring in a full-featured payments SaaS, he introduced the paid membership plan via PayPal and started small. The billing mechanism can be improved endlessly later. Getting the first paying member comes first.

He is candid, though, that one of the two pillars is unstable: ad revenue “fluctuates inexplicably month to month,” and no hack exists on the internet to stabilize it. Raising the subscription share works as the practical hedge, absorbing the monthly swings in ad revenue.

The three structures that produced the growth

Starting from “his own problem” meant demand validation wasn’t needed. The origin was a tool to streamline cram-school work. Because he used it every day himself, the feature priorities could hardly miss. As he puts it: “solving my own problem turned out to solve other people’s too.”

The target pivot was necessity, not whim. The expansion from teachers to students happened because, as the user base widened, a smartphone app became essential. The audience expansion from teachers (hundreds of thousands) to students (millions) coincided with adapting to students’ device reality (smartphone-first). Adopting Flutter lets him run web, iOS, and Android alone.

The arrival of LLMs raised the ceiling on “one person’s development output.” The rapid-growth period of FY2024 coincides with LLM-driven development efficiency. He says: “Tasks that used to take a week sometimes finish in minutes thanks to AI.” Implementation capacity, the binding constraint of indie development, loosened, bringing moves that “used to require a team,” like multi-platform rollout, within an individual’s reach.

Not all smooth sailing — the struggles he discloses

Taking on Flutter was “genuinely hard,” he says. Web apps and mobile apps are “different worlds, different languages, and completely different marketing science.” It was a domain where his web-side wins didn’t transfer, cleared at last with the tailwind of LLMs.

His work rhythm isn’t uniform either. Task management runs on self-direction (“if I thought of tasks the day before, I do them; if not, it’s a day off”) working seven days a week when motivated, taking seven-day weekends when not. He also doesn’t hide that “there’s about a month a year when I tend to get depressed.” On the uncertainty of indie life, his framing is: “isn’t that actually true for employees too?” That behind the up-and-to-the-right revenue graph there are constant waves of output and mood is information anyone considering this way of life should price in.

Lessons and analysis

The exponential ramp of ¥120,000 → ¥1.47M → ¥3.91M is the classic indie-development pattern. Looking at the first year’s ¥122,000 profit alone, quitting would seem correct. Yet two years later it was 30x. It is the same curve as Tsuzuki Blog’s ¥42,000/month → ¥1 million/month in a year and Bannerbear’s 8 months at $0 → $10K MRR in 2 years: for stock-type businesses, first-year revenue is nearly useless for forecasting. What to judge by is not the revenue figure but the slope of leading indicators like user count and retention.

“Go independent once you beat the average salary” is a reproducible threshold. He modestly calls it “improvised,” but in substance he crossed an objective line (¥4.61 million, average-salary level) before going independent, and did so by transferring the cram-school business: prudent risk management. Alongside Tsuzuki’s “quit after saving two years of living costs”, it stands as a practical reference for timing the leap from side project to full-time.

Education × accumulated data is a high-defense position for indie development in the AI era. His read: “maybe only database-centered services will survive,” and “holding a reasonable amount of user data matters.” The premise is a recognition of the moment, frontend chatbots are trying to swallow every service, and simple problems get solved entirely inside ChatGPT. Features can be copied by LLMs. Accumulated learning histories and content data cannot. A practitioner’s first-hand insight for choosing where to enter as an indie developer, and it carries weight. The editorial desk shares his read, with one amendment: the data moat came second. What worked first in this case was problem selection, the ability to pull a “can’t manage without” problem straight out of his own daily cram-school work.

What can be imitated, and what can’t

• Easy to reproduce: the launch pattern of “solve your own work’s inconvenience → open it to others with the same inconvenience” is available to anyone with a job. Choosing problems at the “can’t manage without” level, starting billing early with a simple processor like PayPal, and a low-cost structure of under ¥700,000 a year in expenses are all standard and attainable
• Limits: education is a blessed genre where seasonality and grade-progression cycles stabilize demand. With the service name undisclosed, the acquisition channel (search? word of mouth?) cannot be read, so that part of the playbook is unreproducible. And being a cram-school operator was itself the “closeness to teachers’ and students’ problems” — matching that precision of problem selection in a domain without such knowledge is hard

Related cases

• The “what works, what doesn’t” visible at ¥50,000 a month — the life-size speed of the same indie app development
• Tsuzuki Blog: 100 posts at ¥42,000 → ¥1 million in a year — the blog version of the exponential ramp
• PhotoAI’s “no free users” philosophy — a polar-opposite reference for billing design