kan.kikuchi: ¥5,913,887 a Year from Solo-Built Joke VR Games — With the ¥420K Build Cost and 813 Hours All on the Record

“A VR game where you lick a recorder without the girl noticing.” The idea that got the biggest laugh in a room full of friends became a product name, and then a business grossing ¥5.91M a year on Steam. The figures Japanese solo developer kan.kikuchi published are, in contrast to the premise, meticulous.

On his blog, he reports that all the VR games he released as an individual grossed ¥5,913,887 between 14 August 2020 and 13 August 2021, netting ¥4,490,828 after Steam’s cut and refunds. More than 80% of that came from one title: 縦笛なめなめVR, released internationally as Chupa Chupa VR. Dividing the year by twelve gives roughly ¥490,000 a month, but the actual monthly curve spikes hard around major sale events — it is not a smooth line.

He goes further in a separate post about the flagship title alone, publishing a receipt-level breakdown of development costs, the hours logged, the effective hourly rate, and the refund rate. Disclosure at this level is rare in indie game development, so it is worth taking apart.

The numbers on the flagship

Chupa Chupa VR launched on Steam (Windows VR) on 14 August 2020. The developer has kept appending sales figures alongside the days elapsed since release.

(Profit and hourly rate are his own calculations: revenue less Steam fees, refunds and development costs, divided by measured development hours.)

The unit breakdown: 3,311 copies of the base game at a $10.99 list price, 1,543 costume DLC at $0.99, and 468 male-character DLC at $0.99. Close to half of base-game buyers also bought DLC.

On the cost side, development came to ¥420,544: ¥65,258 for Unity Asset Store assets, ¥28,563 for 3D models from BOOTH, ¥32,978 for music and sound effects from AudioStock, ¥91,625 for a logo and voice work commissioned via Lancers, ¥173,740 for English, Chinese and Korean translation through GENGO, and ¥28,380 for the website on Wix. The largest single line item being translation matters for the market-selection story below.

Development time, measured with Toggl, came to about 813 hours including updates and DLC. At day 151 the implied hourly rate was ¥2,616. By day 535 it had risen to ¥5,830, because the numerator kept growing while the denominator stopped.

Why a “slightly lewd joke game,” and why VR

The most instructive part of this case is that the market logic was articulated in advance. He separates “how much you could earn” from “how likely you are to earn it.” Gacha-driven mobile games score high on the first and terribly on the second: giant competitors, and an ocean of apps.

He chose PC (Steam) VR instead. There were enough users to make the revenue worthwhile, but not enough return for companies to pile in, and individual developers were deterred by expensive hardware and thin documentation. The result was a shelf without price competition and without a runaway quality bar. Steam’s “New & Trending” and top-seller lists are per-category, so on a shelf with a small denominator you can rank high without enormous unit volume. At launch the game hit #1 in the VR category and #56 across all Steam games.

The genre choice follows the same logic. Mildly risqué games sell, yet companies find them hard to publish and individuals avoid them out of embarrassment. He also argues that a product’s rating is not absolute quality but the gap between expectation at purchase and satisfaction afterwards, and joke games start with low expectations on systems. In practice the game holds 43 reviews at 93% positive, and he notes with some pride that reviewers praise the game design, not just the gag.

Idea selection involved actual test marketing. He floated concepts as small talk among a group of friends and judged them on three criteria: the idea lands in a short explanation, the friends start riffing on it themselves, and it gets a laugh the moment it is said. The winner became the game. The title came from something someone said in that same conversation.

What ¥1,111 was for

Pricing was reverse-engineered from the shelf too. The Japanese price is ¥1,111, partly a pun, but mainly because a 10% discount takes it just under ¥1,000, qualifying it for Steam’s “under ¥1,000” feature.

The discounting is unusual as well. Where many titles cut deep, this one stayed at 10% (later 20%). His own observation: a solo 10% discount during a quiet period pushed the game back onto page one of the sales ranking, and running a 10% discount during a big seasonal sale produced roughly the same unit volume. On the VR shelf, you can buy visibility without widening the discount.

Promotion ran on zero ad spend. He made the trailer himself, choosing the music first and cutting the footage to it. The Japanese and English versions together passed 50,000 views. Through Steam’s Curator Connect he sent out ten keys and got four positive write-ups. A streamer covered it repeatedly, and he bundled it with his previous title. That earlier game, 札束風呂VR, took ¥370,136 in the flagship’s first 151 days, clearly better than when he had discounted it 58% before the new release. When a hit lands on the shelf, the back catalogue moves with it.

What did not work

The same developer has also published a miss made with the same process. SLASH OF BULLET, built to sell worldwide with the United States as the primary market, “flopped” in his own words.

His post-mortem is specific. VR games are few enough that users find them without heavy marketing, so this was not a promotion problem, and the reviews from people who played it were not bad. What remained was that the game was “easy to compare.” A near-future cityscape, robots and a sword leave plenty of comparison points even inside VR, and he had assumed that “throwing slashes” was differentiation enough. On top of that, he had run informal test marketing with friends before the flagship but skipped it entirely this time, citing the pandemic.

The flagship has its own weak spot. Its refund rate is 14.1%, well above Steam’s roughly 10% average, because average playtime is 23 minutes and Steam grants unconditional refunds under two hours. And of 3,311 copies sold, only 2,526 people actually played, 785 buyers never launched it. The strength (people buy it as a joke) and the weakness (they return it) come from the same property.

He is critical of his own international push as well. He spent ¥173,740 on English, Chinese and Korean localisation and picked an English title chosen for sound rather than meaning, yet 58% of buyers were in Japan, a sharp contrast with his previous title’s 21% Japanese share. He writes that this side does not feel like it worked.

Conditions and limits for repeating this

What generalises is the shelf selection: rather than winning on quality, find a shelf competitors are structurally unlikely to enter, then ride its exposure machinery, rankings, features, curators. But VR being under-supplied while companies stayed lukewarm was a window that existed around 2020, not a permanent condition. A ¥470 walking simulator that sold 30,000 copies on day one and passed 1.4 million in a little over a year was likewise a short-form design fitted to the shelf of its moment.

The limits are equally clear. More than 80% of that ¥5.91M sits in one title, and he calls the concentration excessive himself. There is no guarantee the revenue repeats next year, and given that it is not a salary, he does not consider it a high figure. In practice, solo VR development is not his whole income: he co-develops another game, helps a mobile game company, still earns from older apps, and runs a long-standing technical blog. As with a creator who reached ¥10M a year selling 3D assets for VRChat, the picture is a bundle of income sources rather than one product.

The accumulated base matters too. Before going independent in January 2017, this developer had shipped more than 30 mobile apps (roughly one million downloads combined) including during his student years, and had run a Unity technical blog for years. His build speed, his choice of outsourcing vendors, and the care he took to email Steam and confirm that publishing sales data was permitted all rest on that base. Read alongside a developer whose 13 games earned ¥700 in total, it becomes clearer where the outcomes diverge.

Finally, his corrective measure is thoroughly practical. Steam lets you publish a store page before the game exists, and wishlist counts before launch give a rough sales forecast. He has run two titles through that test: one came back as “probably will not sell if released,” the other as “not bad, but not enough to greenlight full development.” Instead of discovering failure after building, confirm before building. In a business whose cost of goods is 813 hours, reordering those two steps is the single largest saving available.