Jammsworks: Two Childhood Friends, 49 Escape Games and ¥600M in Cumulative Revenue — It Started by Reading 100 Rivals' 1-Star Reviews

Escape-room games are one of the most crowded categories on mobile. Within it, Jammsworks (a two-person studio formed by childhood friends from Oizumi, Gunma Prefecture) shipped 49 titles between October 2016 and August 2024 and passed 30 million cumulative downloads. Cumulative revenue crossed ¥600 million as of July 2024. All of these figures come from programmer Asahi Hirata speaking on the record in an interview with Levtech LAB.

Revenue comes from exactly two places: in-game banner ads, and a video ad that plays before a stuck player can see a hint. Look at the studio’s App Store developer page and every published title is listed at ¥0. There are no paid unlocks and no item sales. Free games with ads attached, 49 of them over eight years, add up to ¥600 million.

Spread across the 94 months from the first release to the interview, that ¥600 million averages ¥6.38 million per month. But this is not “current monthly revenue.” The years with the most releases were the first one or two after founding, and the release pace today is slower, not faster. The gap between the smoothed figure and the actual curve is itself a good description of how this business works.

Eight years in numbers

The studio’s official Instagram bio has since been updated to “distributed in 21 countries worldwide” and “over 35 million cumulative downloads”, the founders themselves disclosing that the numbers have kept moving since the interview.

Using bad reviews as a blueprint

Before building their first title, the pair did research rather than ideation. They downloaded roughly the top 100 escape games on the App Store, played all of them over two or three weeks with almost no sleep, and read every word of the review sections.

Two patterns dominated the negative reviews: “the atmosphere is creepy or scary,” and “the difficulty is too high.” In Saito’s words: “So if we just make the opposite, we get a game nobody gives a bad review to.” That produced the concept, a game with a light, non-claustrophobic atmosphere and easy puzzles.

What matters here is that the research looked at low ratings, not high ones. Positive reviews reflect a maker’s personality and luck. Negative ones read more like a log of where players dropped off. Jammsworks converted a competitor-generated record of frustration directly into its own spec. The crowdedness of the category worked in their favour, because it meant there were 100 titles’ worth of material to read.

The habit persists. New reviews are piped automatically into Slack and checked every morning at 8. When a puzzle asked players to cut a string using a stag beetle and two 1-star reviews called it too far-fetched, a version replacing the beetle with a box cutter shipped the next day. Changing a spec over two 1-star reviews is a speed that only really works at a headcount of two.

The second game, where trying to raise revenue halved it

They stumbled on title two. To increase views of the video ad shown before hints, they deliberately raised the difficulty. The intent was to make players get stuck more often. What actually happened was mid-play abandonment, and revenue for the second title came in at half of the first.

The failure left traces on the platform. On the App Store the average rating for the second title, Red Room, currently sits at 3.5, while the third, Christmas Eve, is in the 4.5 range, closely matching Saito’s recollection in the interview of “3.5 stars on the previous one, 4.6 on the third.” Even the debut 1K sits in the 4.0 range, below the 4.7-ish median of everything that followed. The numbers show a rating trough confined to the first two releases.

For an ad-funded app, changes that increase ad exposure look rational in the short run. But in an escape game, the time a player spends stuck is also the time they spend deciding to leave. They tried to grow ad inventory and shrank the audience for it instead. That is a different risk from the one facing a solo developer whose ten apps peaked at around ¥50,000 a month in 2021 and have declined since, where the pressure came from outside.

When it stops, it stops on the platform’s terms

Hirata says that in app businesses, “depending on what the distribution and ad platforms decide, revenue could go to zero tomorrow.” Something close to that has already happened. Apple told them not to release so many similar games in such a short window, and for three months every new title was rejected, apparently treated as churning out duplicate content. No particular countermeasure worked. After three months, submissions suddenly started passing again. The founders still do not know why.

For a business whose weapon is volume, volume itself becomes a review risk. Set beside a developer who shipped 90+ personality-quiz apps for 8.2 million downloads and ¥20 million in revenue, or a decade of solo development that reached two million downloads across 68 apps without a single hit, Jammsworks occupies an unusual position: a volume player whose per-title ratings are high. The 56 escape games on the App Store carry about 1.07 million ratings in total, with a median average rating in the 4.7 range. Refusing to trade satisfaction for volume is part of why the business could come back after the rejection window.

The pace has, in fact, been deliberately slowed. The first year or two ran at one release a month. Today it is about four a year. Counting App Store release dates, twelve titles in 2017 gives way to three or four a year from 2021 onward. The shift toward more content per title shows up directly in the numbers.

What transfers, and what doesn’t

The transferable part is treating reviews as spec input rather than sentiment. In any crowded category with a deep pile of reviews, the same move is available regardless of industry, and it works best where the complaints share a common shape.

Plenty here does not transfer. The business rests on a property specific to escape games: Japanese appears only in UI strings and hint text. As Saito puts it, localization costs almost nothing, and few products can be shipped to 21 countries without translation. Even so, roughly 80% of users are in Japan, and a genuine overseas hit is something the founders still describe as ahead of them.

The two-person division of labour also fit from the start. Saito came from architectural design and could produce the backgrounds. Hirata could write the code. A static-image escape game is one of the few genres that closes with exactly those two roles. Choosing a technically undemanding genre was itself a product of constraint: Saito’s parents had told him to go back to salaried work if he could not show results within two months.

The largest limit is that ¥600 million is an eight-year cumulative figure, not a guarantee of stable monthly income. The three-month rejection showed that revenue can stop for reasons unrelated to anything the operators do. The founders’ repeated insistence that they feel anything but secure is, for reading this case, roughly as important as the numbers.

Note: the revenue and download figures here come from the founders’ own statements in the Levtech LAB interview. Title counts, release dates and ratings were corroborated against public App Store data, but there is no external way to verify the revenue itself.