Okashi Kaigyo Lab: a 13-tsubo cake shop booked ¥27,073,172 in year three and published every monthly sales figure to the yen

A small Japanese cake shop has published 24 months of sales, down to the yen. The blog Okashi Kaigyo Lab shows ¥21,730,380 in annual revenue for the second year of trading and ¥27,073,172 for the third. Averaged monthly, that is ¥2,256,098, seven times the median of the eight brick-and-mortar cases on this site that disclose monthly revenue (¥320,000) and nine times the median of the 88 Japanese cases (¥250,000). It also sits almost exactly level with the ¥2.25M median across all 251 revenue-disclosing cases.

The author goes by Marron. He finished a confectionery school, trained at four shops, and opened a 13-tsubo store at 32. He and his wife split production and the counter between them, the business is takeout-led, and this is year nine.

The instinct is to credit disciplined cost control. Yet the cost ratio he publishes elsewhere runs to 45% in practice: 35% for ingredients plus 10% for packaging, fifteen points above the “30% for a cake shop” figure the trade repeats. Rearrange the numbers and the profit here is being decided in the lease, not in the kitchen.

Twenty-four months, recorded to the yen

Setting his two published years side by side against the same months gives this.

Only August fell below the prior year, and in that one month customer count dropped 13.6%, from 955 to 825. Average spend rose from ¥1,754 to ¥1,926, which held the shortfall to ¥86,498. A shop where fresh cake stops moving in summer, yet average spend jumps to the second-highest level of the year, is telling you that the reason people come has shifted toward gifts.

The growth breaks down too. Revenue rose 24.6%, customer count rose 17.7% from 12,358 to 14,540, and the annual average spend rose only 5.9%, from ¥1,758 to ¥1,862. Marron lists “average spend rises year by year” as the third of his five rules, but most of the year-three increase came from the number of people who walked in. His sense that regulars accumulated shows up in the count, not the ticket.

“December decides the year” turns out to be 1.67x

Marron calls December the peak of the year and describes it as close to double a normal month. Our reading differs. December of year three was ¥3,568,570 against an average of ¥2,136,782 for January through November, a ratio of 1.67. Year two lands at 1.65, the same ratio twice over. December accounts for 13.2% of annual revenue, so ordinary months are carrying the other 87%.

What matters more than the multiple is his own note that December is not the profit peak. Christmas loads up on cream, strawberries, decorations and dedicated boxes, then adds short-term staff wages and waste risk. March, by contrast, sells baked goods for graduations and farewells, which keep longer, waste less and cost less to make. Top line peaks in December and profit lands in March, he writes. The tallest month on the sales table drops to second place on the income statement.

Four hundred cakes made, two hundred left over

The number worth remembering here is not the annual revenue but the count of cakes thrown away. In his first Christmas he prepared roughly 400 Christmas cakes and sold about 200. He made that many because he was afraid of turning customers away. The reason they did not sell is plain: a shop that has just opened is not known in its area, so nobody wanders in on the day.

Year two he made about 250, cutting 150 units, and changed exactly one thing about the method: build to confirmed reservations, then add a little. Reservations open at the end of October precisely so that this firm number can anchor the production plan. From December 23 onward, when finishing begins, he slept one hour a night in years one and two.

As a fix it looks unremarkable, but what happened was a design change that moved inventory risk from forecasting to booking. He paid for the premise that walk-in sales do not work until recognition exists, and the price was that first year of waste.

What survives sits in the lease, not the kitchen

Marron gives two reasons a 45% cost ratio still leaves roughly 35% margin: rent runs about 2.5% of sales, and advertising spend is zero.

The property is a 13-tsubo shell unit near a rural station, a former office, at ¥60,000 a month with ¥360,000 in acquisition costs. Against average monthly sales of ¥2,256,098 the rent ratio is 2.66%, under a third of the 10 to 15% the trade treats as the benchmark. His opening business plan assumed ¥30,000 a day across 25 days, or ¥750,000 a month, and even against that plan the ratio was 8%. He looked at a vacated convenience store with excellent frontage and walked away because ¥200,000 rent and over ¥1,000,000 in acquisition costs would have put the ratio at 26% of projected sales.

The lease, not the cost ratio, is what decided the economics of this shop. He can afford not to skimp on ingredients, and he could absorb inflation pushing his cost ratio from 31% to 35%, because the fixed costs were crushed first. One year he raised prices across the existing range by 30%: customer count fell that year while profit rose, and the following year customers came back. The tolerance to raise prices at all comes out of that low fixed base.

The ¥10.16M he spent opening the shop fits the same design. Keeping property acquisition to ¥360,000 left the rest for equipment and working capital.

Zero ad spend, one story a day

Customer acquisition is Instagram and nothing else, and the account passed 10,000 followers. The operating pattern runs against the advice to post more. The only daily act is one Instagram story showing what is in the display case that day. Feed posts go up when a new item launches, roughly every two weeks, plus a monthly notice of closing days. He does not post repeatedly in a day because, in his words, it is annoying.

Stories work here because the product changes daily. Showing every morning what is available today is closer to publishing inventory than to advertising. The 24-hour lifespan of the format matches the shelf life of fresh cake that has to sell that same day.

The area is crowded: two other independent cake shops, a Japanese confectionery, and a large chain. The reasons he gives for lasting nine years are that he ran the social account from the build-out period (with under 100 followers at the time), that he swapped the lineup based on what customers asked for, and that he handed the counter to his wife and stayed in the kitchen.

How far this carries

Among Japanese brick-and-mortar cases we have published a food truck that stopped in month four at ¥380,000 a month, a crepe stand in five tsubo clearing ¥1M a month whose owner still called it a failure, and a yakitori takeout shop opened for about ¥500,000. At ¥2,256,098 a month, this shop sits at the top of that ladder.

• A 45% cost ratio leaves profit only when rent has been pushed down to around 3% of sales and the reason to visit can be created without paying for it
• Building to reservations plus a margin works only when most of the product moves through bookings on a fixed calendar date, so the count holds even if walk-ins are written off
• One story a day becomes acquisition only when what you sell changes daily and the people who see it live close enough to come the same day

The limits belong on the record. That ¥60,000 rent rests on a rural station-front location and on individual terms from a landlord who waived both key money and a deposit. Reproducing the ratio in a city is a different problem. The 4am starts, 14-hour days and four hours of sleep sit directly behind the 35% margin. What is disclosed stops at sales, customer counts, average spend and approximate cost and profit ratios; actual labour and utility costs, and first-year revenue, are not published. The figure of a personal income in the ¥10M range is self-reported, with no third-party verification.