10 hours at the day job plus 3 hours commuting, and still ¥94,000/month from a Python automation side business — why he walked away after ¥150,000 total

What’s rare about this record is that the breakdown of a good month and the reason for quitting appear in the same piece. “Ratto,” a systems engineer at a Tokyo IT company, published a record showing he pulled in over ¥90,000/month across two crowdsourcing sites roughly six months into his job — but toward the end, he reveals: “I earned ¥150,000 total and quit.”

The granularity of the numbers is also distinctive: gross payment, fees, and net take-home are broken out separately by platform. This record explains, using the actual numbers in his own account, that side-hustle income isn’t what it looks like on paper.

The October 2022 breakdown

He describes the crowdsourcing platforms’ fees as “roughly 15–20%”, which works out to about 16.7% on the Lancers side and 20% on the CrowdWorks side. Note that simply adding up the per-platform net figures gives ¥77,220, which differs from the article’s stated total net of ¥75,504 by about ¥1,700. Taking the ¥94,380 gross and subtracting a flat 20% gives ¥75,504, so the stated total appears to be a rough approximation. We’ve left the original article’s figures as-is here.

Combined with his day job, he writes, “I think that month’s gross total came to around ¥400,000.” His day job was five days a week, and his stated living conditions were “10 hours of work plus 3 hours of round-trip commuting.” That this side hustle was running on top of that time allocation is one of the notable points of this record.

What he was actually selling

The work was contract development of Python programs that automate browser operations. The typical example given in the article is “a program that periodically checks whether an item is in stock and sends a notification”, a script that runs on a schedule, launches a browser, checks a target site’s stock, and sends the result via something like LINE. At the time of writing (July 2024), he estimates the going rate at “around ¥20,000–30,000.” The rate hasn’t changed much from when he was taking these orders, he adds, but the level of the projects seems to have gone up.

The order in which he learned things is also spelled out.

His overall estimate is “at least about 3 months.” And he notes he’s “given up on programming and started over more than three times,” so these three months weren’t a straight, unbroken run.

The two decisions that mattered most

There’s no dramatic turning point in this case. No spike in revenue, no viral moment, no successful rate negotiation. What mattered were two decisions made at the start of his learning process.

One was narrowing the scope of what to learn to “web browser automation.” Trying to learn programming broadly means the scope balloons, from IT fundamentals to every library under the sun. He acknowledges that, then focuses only on the area that can turn into paid work as a side hustle. The allocation of one month for Python basics and 2–3 months for HTML/Selenium is clearly skewed as general engineering education, but for the single goal of “deliver an inventory-monitoring tool,” it’s neither excessive nor lacking.

The other was registering on crowdsourcing platforms and browsing listings (step 2) before finishing his learning. He gives two reasons: “to confirm what I’m aiming to build” and, more importantly, “motivation (realizing I could earn a specific yen amount by building this.” Because the hardest step, step 3, drags on for 2–3 months, he argues it’s hard to push through while unsure whether “I can actually earn money from this”) without that motivation anchor.

On the receiving end, the one thing he explicitly credits is response speed: “For job applications and any replies after applying, answer within a day.” “There are plenty of listings that draw dozens of applicants.” “Just responding quickly puts you ahead.” The pitch template he shares publicly also leads with the same PR angle, “fast response and delivering a prototype app early.” That he places response speed, not technical skill, as the differentiator says a lot about the character of this case.

An estimation mistake, and the reason he walked away

His failures are also stated concretely. He writes, “I mistakenly took on a job involving over 100 million records for ¥15,000, and regretted it.” Among the things he lists to confirm before accepting a job: whether the client is on Windows or Mac, whether the target site has anti-bot measures, the steps needed to set up Python on the client’s PC, every requested feature in full, and (this last one, drawn directly from that failure) whether the volume of data to be scraped is reasonable.

He gives two reasons for quitting: “the work isn’t stable” and “my day job got busier.” After earning nearly ¥100,000 in a month, he had thought, “maybe I could live anywhere in the world doing this”, but in reality, competition for listings is fierce, and “you basically have to watch listings all day or you’ll miss them, which gets pretty exhausting.” On top of that, he felt that projects accessible to amateurs tend, by nature, to be subcontractor-style work with an inherent ceiling on earnings. His conclusion: “I don’t think this is something you could build a livelihood or stable ongoing income on”, though he adds the caveat that he can’t say for certain.

This case shows that even after a single month of ¥94,000, the cost of sustaining that level, constantly monitoring listings, was judged incompatible with his day job. The ¥150,000 lifetime total reflects how quickly that judgment was made.

What’s replicable, and what isn’t

What’s replicable is the scope-narrowing and the response speed. Neither requires capital or an audience, and both are laid out explicitly as a learning sequence in the article. The pre-acceptance checklist is also directly reusable as-is.

That said, the underlying conditions carry real weight. He’s a working systems engineer, and the experience listed in his pitch (“built internal business automation tools,” “built automated web-testing tools”) comes from his day job. Someone with no experience can’t submit the same pitch. The three-month learning estimate should also be discounted somewhat, given it’s three months for someone already touching code professionally.

The other thing that’s hard to replicate is time endurance. He kept applying, developing, and delivering on top of 10-hour workdays and 3 hours of round-trip commuting, and the result (a single month of ¥94,000) didn’t sustain. The value of this record lies less in the fact that he earned money, and more in the second half: that the level he earned didn’t coexist with his day job.

Related reading

• Until MENTA was sold to Lancers — a case examining the economics of the marketplace where individual skills get traded.
• 24 Kindle books — a record tracing the relationship between time invested and revenue in a side hustle an individual can run in spare time.