Raised ¥200M, earned a 4.8 rating, closed anyway. What stopped dark-store startup Mesh was the basket size one neighborhood over

“The story of a dark-store startup that raised ¥200 million from VCs and burned through all of it.” That is the title Shun Sato, CEO of Mesh, Inc., gave on October 21, 2026, to his retrospective on the business he closed.

Despite the title, the numbers in the text are all good ones. An App Store rating of 4.8. A 99.1% satisfaction rate in in-app reviews. Sales growing about 140% month over month. Orders kept coming from the first day of service, and by the next day the team was so short-handed that engineers went into the warehouse.

Normally numbers like these are read as reasons to keep going. Mesh closed the business with these numbers in hand. The deciding factor Sato wrote about was a single point: Hiroo had many customers placing ¥5,000 orders several times a week, but Ebisu, in the same central Tokyo, did not.

Mesh took venture capital, which puts it outside this site’s criteria for cases, which record businesses that started small. Its sales are also given only as a growth rate, with no amounts. Even so, the way the numbers are laid out in this retrospective is worth reading, so it is handled here as a column rather than a case.

From the raise to the exit and the pivot to development

The dates of the closure and the pivot, the basket size in Ebisu, cumulative order counts, and the amount “burned through” in the title are not written in the retrospective.

The year it raised four times the norm

Mesh raised about ¥200 million in its seed round in March 2022. Sato writes that seed rounds at the time were typically ¥30–50 million, so this was more than four times the norm.

On its own the amount looks aggressive, but set beside the industry that month the impression changes. According to BRIDGE, the following April another dark-store operator, OniGO, raised ¥720 million in its seed round, bringing its total since founding to about ¥1.18 billion. Sato himself opens his retrospective by citing how the valuation of America’s Gopuff rose 3.8 times in eight months. In a sector awash with money, Mesh was on the modest side.

The plan and the reality also diverged. BRIDGE’s April 2022 article said Mesh planned to launch in Tokyo’s wards as early as that summer. In Sato’s retrospective, the business started about a year after the raise, once a property was secured, and the beta app came out in January 2023. A section heading about not even being able to decide the width of the warehouse aisles tells what that half-year delay was made of.

Nobody had run a supermarket

Midway through the retrospective comes this scene. How do we manage inventory? Where do we source products? How do we find warehouse space? Every answer from the team was “No idea!” Nobody had ever run a supermarket business.

When raising money, an investor also told them, “I can’t picture Japanese consumers, who are fussy about quality, buying fresh food with natural variation by delivery.” As it turned out, that worry was wrong. Orders came from day one and the app earned a 4.8 rating. Customers who would buy fresh food online did exist.

What the desk weighs most in this retrospective is that the business stopped after the worry had been disproved. The question people doubted, whether customers would buy, got its answer. The question left unanswered was where they would buy, and for how much.

Satisfaction is a number about people who bought. What decides a business is the number for the neighborhoods that didn’t

The 4.8 rating, the 99.1% satisfaction and the roughly 140% monthly growth all came from people who had already ordered. Those who ordered were satisfied, and their numbers were growing. That much is established.

What Sato named as decisive lay outside them. Mesh started in central Tokyo on the hypothesis that relatively affluent families with young children would pay a delivery fee. In Hiroo, many customers put ¥5,000 in the cart several times a week. But the same buying was confined to a very narrow part of the city center, and in Ebisu it did not occur. Sato writes that they could only conclude the business would work in parts of central Tokyo but not elsewhere.

A dark store can deliver only around the place where its warehouse sits. So how much one warehouse earns depends on how many Hiroos lie within its reach. The quality of the app does not add to that number. What follows is inference: however good the rating and growth rate, they were no reason to add warehouses. If the next neighborhood was an Ebisu, the same warehouse would not earn the same.

What buyers wanted was not the business but what was built for it

On the day it announced the closure, Mesh received acquisition approaches from three companies: a national retailer, a large e-commerce platform and a leading startup. Sato gives four reasons they valued it: the UX reflected in the 4.8 rating, the fact that all systems were built in-house, zero major incidents, and growing sales and KPIs. The systems were built by three full-time and two part-time engineers.

Mesh turned down every offer. The only reason Sato gives is that the team could not give up wanting to build businesses themselves. Afterwards, requests for planning and development arrived from companies including the convenience-store chain Lawson, and Mesh became a development company.

Lined up, the reasons for the approaches read as if buyers were looking less at the dark-store business itself than at the systems built to run it and the people who could build them. The value left after the business closed lay not inside the business but in how it had been built.

The pattern stands out beside the column on the end of the food delivery service menu. menu’s shutdown announcement contained not a single number, and its acquisition served to wind the business down. Mesh’s retrospective gives the rating, satisfaction, basket size, even the number of engineers, and buyers tried to put a price not on the business but on the ability to build. Both were delivery businesses, yet the records of how they closed could hardly be more different.

The numbers that are not written

The limits need stating. The amount “burned through” in the title appears nowhere in the text. How much of the ¥200 million was spent and how much remained at the pivot is unknown. Neither Ebisu’s basket size, cumulative orders nor the closing date is given. “Sales were growing about 140% monthly” does not say over which period, or whether it is month over month or some other basis. The raise date and shareholders were checked against the third-party database INITIAL, but sales and basket size can be checked only against Sato’s own retrospective.

For a count of reasons businesses close, there is the column classifying the causes of death of 11 closed cases, where the “profitable but not sustainable” type outnumbered running out of money. And the record of an SEO tool that drew 1,000 people on its beta’s first day and only fell from there reads as another example where first-day numbers were no reason to continue.

Records of closing with good numbers are rare

Many people write retrospectives after closing a business, but most explain why the numbers were bad. Mesh’s does the opposite, writing why it closed while the rating, satisfaction and growth were all good.

The deciding factor was a difference on the map: Hiroo’s ¥5,000 did not repeat in Ebisu. The numbers from people who used the app were all good, and the numbers from the neighborhoods that did not use it decided the business. The title says it burned through everything, but what the record keeps is not the amount burned. It is the numbers accumulated before that difference was found.