TF's apps: ¥103,437 in Year Eight — A Shiga Salaryman Who Kept Publishing His Ad Revenue Every Month, and Hit ¥20,000/Month in Year Nine

When people talk about indie app development as a side project, the cases that surface are the ones that reached several hundred thousand yen a month. In reality, far more common are businesses that earn a few tens of thousands of yen a year and keep going for years. The blog “週末のアプリ作成” (Weekend App Building), run by a 45-year-old company employee in Shiga Prefecture, is a rare record of exactly that band of numbers, published month by month for nine years.

Here are the figures. In 2023, the seventh year, annual revenue was ¥52,859. In 2024, year eight, it was ¥103,437, close to double. That is ¥100,000 a year. Convert it to an hourly rate and the picture is bleak, but that is not where the value of this record lies. From 2025, the ninth year, monthly figures climbed noticeably: ¥20,400 in August 2025 (321% year over year), ¥12,300 in January 2026 (216%), and ¥20,400 again in June 2026 (160%). Clearing ¥10,000 a month has become the normal state.

The revenue source is Google advertising inside free apps. Not paid downloads, not client work, just ad rates multiplied by impressions. Which means download swings show up directly in the money.

The published trajectory

What moved the numbers

December 2024’s ¥21,700 stands out plainly against the months either side of it. By the author’s account, one of his apps landed in Google Play’s Best 15 that month, and downloads jumped by 27,000% versus the previous day. When a store’s editorial slot picks you up, an indie developer’s monthly revenue moves by that magnitude. The flip side: January 2025 fell back to ¥5,672. A single-month spike does not become an asset.

What does become an asset is the old app. The flagship is a bear-bell app, a product the author describes as “about ten years old.” Downloads in October 2022 were around 300. In October 2023 they exceeded 1,500. Roughly five times. In the season when bear sightings make the news, people search, download, and see ads. It is the kind of app where demand moves in a place unrelated to the developer’s own effort. In January 2026 he rebuilt it substantially, he described its accumulated features as being like a house extended over and over, adding a screen lock, new sounds including wolf howls and plastic-bottle noises, a flashing light, and randomized playback intervals. It is reasonable to read that decision to go back into a decade-old app as one of the things propping up year nine.

The other lever was television. His price-comparison app “どちらがお得” (Which Is the Better Deal) shot to 6,000 downloads after being featured on TV. In this case, the only things that can be called acquisition channels are in-store search and recommendations, plus exposure of that kind. No SEO articles, no ad spend, no social media operation appear anywhere in the revenue explanation.

A comparable case is the developer who shipped eight iOS and two Android apps over eight years to reach ¥30,000 a month. That one peaked in 2021 and has been declining since, while TF’s apps turned upward in year nine. The difference is whether new development stopped or continued. A portfolio of ad-supported apps starts decaying the moment the hands come off.

What did not work

Because he keeps going, the misses are also on the record. Of “勇者の宝探し” (Hero’s Treasure Hunt), released in December 2023, he writes plainly that it “has not been downloaded at all.” His first AI-powered app, an “AI Smile Diagnosis” released at the end of February 2025, had cumulative downloads of 266 as of September. Seven months of actual data showing that slapping “AI” on something does not, by itself, make it grow. The two Shiga-themed tourism apps he launched in the first half of 2026, concepts he says have no comparable alternatives, likewise see limited usage so far.

The monetization experiments tell the same story. He implemented subscriptions on some apps in November 2024, yet every monthly report since is written consistently as “Google ad revenue from free apps.” No subscription figure ever appears. Adding a billing model is not hard. Whether it turns into revenue is a separate question, and that gap is visible here.

On the AI Smile Diagnosis app, he also notes being contacted directly by a global company based in Singapore. But it stayed an inquiry and produced no revenue. This record does not confuse “we got a reaction” with “we booked a sale.”

Why nine years were possible

For the first half of 2026: six new apps across Android and iOS, and roughly fifteen apps touched with updates. He writes that AI assistance sped up his build pace considerably. A company employee is turning that volume on weekends alone. The goal he states publicly is one million cumulative downloads, not a revenue figure.

That the target is not money seems important. Had he set out to reach ¥20,000 a month, he would likely have quit three years ago at ¥50,000 a year. Because his goal sits on downloads, a metric that moves before revenue does, progress stays visible even in flat years. And the act of publishing the actual monthly figure on a blog functions as a mechanism for continuing.

A similar structure appears in the developer who earned almost nothing for eight years before reaching ¥200,000 a month in year nine, years of ¥30, ¥1,000, and ¥40,000 preceded the jump. If indie revenue curves are not linear, you need machinery to survive the non-linear stretch. Publishing your numbers can be part of that machinery.

What generalizes and what does not

The time horizon is the part worth carrying into other projects. Reaching ¥20,000 a month in nine years is not, as side income, a large outcome. But given that the 2023 monthly average was about ¥4,400, it is more than a fourfold increase in three years. The fact that a ten-year-old bear-bell app still sits at the center of the revenue shows that this kind of business functions as a stock. As long as what you built does not disappear, next year does not start from zero.

The spike is another matter. A Best 15 slot or a TV feature cannot be arranged on purpose. Try to extract a “do this and you’ll make ¥20,000 a month” from this record and you will inevitably lean on those parts. All that can be extracted is how to design the interval before something connects. Set beside the developer who shipped 68 apps over ten years, the common thread is a stance of never letting the at-bats stop.

We do not regard this case as too small. For someone standing at the entrance to indie development, this band of continuous data is more usable than any summit.