A crypto side-hustle app went from ¥300,000/month to a few thousand yen. 20,000 installs, ¥18,000-per-referral, and how it vanished

Side-hustle app revenue write-ups usually stop the pen right at the peak. This case is the opposite: both the peak and the year after are told by the same person. “BitVirtual,” an app built on the side by manufacturing-company employee Shota Nakamura, reached ¥300,000/month during the 2017 crypto boom, and had fallen to a few thousand yen a month by the time of the December 2018 interview, a roughly 99% decline. All figures below reflect that December 2018 point in time.

Peak versus present

Dividing ¥300,000/month by the ¥18,000 referral rate works out to roughly 17 account openings a month. In practice, some referrals surely paid less, so the real count was likely higher, but either way, a 20,000-install app was generating something like a dozen to several dozen account openings a month. That’s an order of magnitude beyond what a typical side-hustle app of that scale produces.

The “few thousand yen” figure after the collapse gives a useful yardstick. Other solo-developer apps mentioned in the same interview: a 43,000-download drone flight-map app earning ¥15,000/month in ad revenue, and a 50,000-download meditation timer earning ¥20,000/month. The going rate for a 20,000-install app run on pure ad monetization really is a few thousand to low tens of thousands of yen a month. BitVirtual’s ¥300,000/month, then, was generated by the unit economics of the market it plugged into, not by the app’s scale.

The demo-trade mechanism

The app itself is a crypto demo-trading simulator. Its origin traces to the developer’s own experience: he bought Bitcoin at ¥300,000 and “panicked and sold it immediately”, which made him wonder if there’d be value in a way to simulate trades first.

This product design turned out to be a powerful lead-qualification device. Anyone who bothers to install a demo-trading app has interest in crypto trading but hasn’t yet committed to depositing real money, exactly the population an exchange wants far more than the general public with zero interest. And the app could see exactly who was performing well and by how much.

The actual tactic pointed straight at that:

Taking advantage of the demo-trading feature, we’d show a “want to open an account?” dialog to people who were posting good results — that kind of thing.

Right after a simulated trade shows a profit is exactly when a user’s sense of self-efficacy is highest. That “I can do this” feeling lowers the psychological barrier to opening a real account. Rather than prompting every user uniformly, the offer only appeared right after a win. Same funnel, different conversion, by design.

Both the rise and fall came from outside

The turning point in this case was the market itself, not anything the developer did. He offers data backing that up directly:

When I overlaid the app’s unique-user trend with the Bitcoin price chart, they moved in almost exactly the same shape.

If traffic tracked Bitcoin’s price, demand was fully exogenous. The November–December 2017 price spike drove installs up to 20,000. And once the price fell, the same curve retraced in reverse.

Worse still, traffic wasn’t the only thing that dropped. “The affiliate terms got worse too,” he says. Affiliate revenue is a product of traffic × conversion rate × payout, and when two of those three factors drop simultaneously, the result falls by an order of magnitude. The roughly 98% decline from ¥300,000/month to a few thousand yen is exactly what that multiplication produces.

To make matters worse, the payout rate was a variable entirely outside the developer’s control. Exchange ad budgets swelled during the boom and got cut once the market cooled. ¥18,000 per referral was the abnormal rate exchanges were paying for new account acquisition at the time, not a price the app’s value set.

What he says in hindsight

Two points of reflection: one, half-joking, is “mostly, I think, I should have just bought and held Bitcoin.” The other bears on the business decision itself.

I also think I should have done more promotion while there was momentum.

This is an important point. The window during which an ¥18,000 referral payout made economic sense for paid acquisition was one of the few moments when ad spend would have penciled out. Boom periods also drive up the cost of paid installs, but with a ¥18,000 payback per referral, aggressive paid acquisition could plausibly have multiplied the scale several times over. Instead, the earned cash was never reinvested into further installs, and he waited for the exogenous demand to recede on its own. When the peak came from outside, scale can only be built by your own hand during that peak, and nowhere else.

That said, there’s a silver lining here. Production costs were extremely low. The app’s promo video used an overseas template purchased for about ¥2,000, with only asset swaps outsourced via crowdsourcing, for a total of about ¥10,000. As he put it, “with this method, even a solo developer can produce a video fairly easily on a budget of about ¥10,000.” With such low fixed costs, falling to a few thousand yen a month didn’t force him to shut down. Combined with keeping his day job while developing this as a hobby, the structure meant there was little to lose in the downturn.

What can and can’t be reproduced

What’s reproducible is the idea of designing a product as a lead-qualification device. Gather the “interested but not yet committed” segment, use behavioral data to identify the most motivated individuals within it, and present an offer at the moment they’re most receptive. This same shape can be built anywhere high-payout affiliate programs exist, certification exams, investing, side hustles, career changes, not just crypto demo-trading. The lightweight operation (a ¥10,000 promo video) is directly copyable too.

What isn’t reproducible is the payout rate itself. ¥18,000 per referral was the product of a specific moment: Japan’s crypto exchanges in 2017 competing fiercely for new customers. Build the same app a year later and you won’t see the same revenue. This case’s peak was 80% timing, 20% execution, and he acknowledges as much himself.

The lesson to take from it is simple: revenue built on a single affiliate program, a single theme, and exogenous demand rises and falls outside your own hands, in both directions. Whether that revenue can be converted, while it’s flowing, into something you actually control (a user base, a product in a different category, know-how in paid acquisition) determines what’s left once the boom passes.

Related reading

• Hitode’s blog operation — a case of diversifying revenue sources while exposed to an external factor, the search algorithm.
• levelsio, running Photo AI solo — another solo build that rode a “wave,” offering a contrasting record of building scale during the wave itself.