# Annexes — The Six Cases

*Companion to "Eighteen Months to Find Out. Or Twelve Weeks."*

*Annexes to Perkd's whitepaper on smart unmanned retail. Main paper: https://www.perkd.me/whitepapers/smart-unmanned-retail.md · PDF: https://www.perkd.me/whitepapers/smart-unmanned-retail.pdf · 繁體中文: https://www.perkd.me/zh-tw/whitepapers/smart-unmanned-retail-annexes.md*

---

## How to read these

Each annex covers one documented operator, in a common structure: a fact strip, how it works, the pivotal fact, what it does and does not evidence, and a source-confidence table. Where a figure is contested or unfiled, the annex says so rather than picking the flattering version.

They are ordered by **weight of evidence**, not by size or fame. Daalchini and the Japanese frozen-vending operators come first because their numbers are the most verifiable; Genki Forest comes last because its figures are brand-supplied with no published methodology. A case's position says nothing about how interesting it is — only how hard you can lean on it.

**A caution that applies to all six.** None of these operators is an international franchisor entering a new market with unattended units. They are workplace catering, a domestic vending platform, restaurants selling beside their own shops, a venture-funded fresh brand, a low-capital franchise category, and a beverage brand's own cabinets. Together they evidence two things well: that the format works at scale for food businesses, and that capturing identity at the point of sale changes the customer economics. The entry sequence argued in the paper is reasoning built on top of them, not a finding drawn from them.

---

## The six at a glance

| | Operator | Scale | Funding | Access | Product | The trade they made | Ch |
|---|---|---|---|---|---|---|---|
| **A** | Daalchini (India) | ~5,500 machines | Franchise partners (>95%) | App required | Ambient + fresh, multi-brand | Identity on every sale; loses the casual buyer | 2, 4, 5 |
| **B** | Restaurant-owned frozen machines (Japan) | 10,000+ units | The restaurant itself | None | Own-brand frozen | After-hours revenue; keeps nothing about the buyer | 3, 5 |
| **C** | Felfel (Switzerland) | ~800 fridges | Felfel; employer pays a fee | Employer-enrolled | Chilled fresh, curated | Frictionless enrolment; no trade outside contracted sites | 2, 4 |
| **D** | Farmer's Fridge (US) | 2,000+ locations | The company | Open, loyalty rewarded | Chilled fresh, own kitchens | Every passing sale; ~9% of spend to get identity later | 2, 5 |
| **E** | Unmanned frozen shops (Japan) | 131 → 1,282 stores | Individual franchisees | None | Own-brand frozen, single SKU | The cheapest franchise rung; no customer knowledge, no moat | 1, 5 |
| **F** | Genki Forest (China) | Tens of thousands | The brand | Join after buying | Beverage | An uncontested channel; capture happens after the sale, if at all | 4 |

Each annex sets out both halves of that trade in full, alongside product, sites and what the case does and does not evidence.

**Read the column as choices, not as scores.** Every operator here traded something for something. Farmer's Fridge did not fail to capture identity — it declined to, correctly, because an airport traveller passes once. Genki judged that a soft drink would not carry a sign-up. The only trade in the set that looks one-sided is the Japanese machines', and even that was inherited from the equipment available rather than chosen.

**What does not vary.** Every funding model in this table works somewhere, and so does every access model. What holds across all six is the relationship between route density and profitability, and between captured identity and repeat economics.

---

## Contents

- **Annex A** — Daalchini, India
- **Annex B** — Japan's restaurant-owned frozen vending machines
- **Annex C** — Felfel, Switzerland
- **Annex D** — Farmer's Fridge, United States
- **Annex E** — Japan's unmanned frozen shops
- **Annex F** — Genki Forest smart cabinets, China

---

## Annex A — Daalchini (India)

**Why it is here.** It is the only operator in this set that requires membership to buy, publishes repeat-rate and profitability figures, and runs a franchise-funded fleet at scale. It is the closest thing available to a test of whether a mandatory identity gate suppresses a food business.

---

| | |
|---|---|
| **Scale** | ~5,500 machines at FY26 close, up 62% year on year; 80+ cities |
| **Who funds the units** | Franchise partners — over 95% of the fleet |
| **Access** | App mandatory. No app, no purchase |
| **Who holds the customer** | The platform, from the first purchase |
| **Product** | Ambient and fresh, multi-brand; 190+ D2C and FMCG suppliers |
| **Sites** | Offices, hospitals, campuses, metro stations, malls, residential |
| **Access Gate score** | 3/3 — friction absorbed by an everyday payment app, captive sites, a catalogue to sell into |
| **Evidences** | Chapter 2 (the gate), chapter 4 (what identity unlocks), chapter 5 (density) |

---

### How it works

Daalchini sits between three parties rather than operating alone. Brands get distribution into six-square-foot retail points. Franchise partners buy machines and handle day-to-day operation. Corporate sites host them. Daalchini supplies the technology, the supply chain, the brand relationships and the demand analytics.

Two models run in parallel: a rental model where the company owns the machine and runs everything, and a franchise model where the partner owns and operates it. The franchise model won decisively — over 95% of the fleet is partner-owned. Revenue is deliberately spread across product sales, machine sales, subscription fees on installed units, and on-machine advertising, so no single line carries the business.

**Supply and route.** The supply chain is local and multi-brand — packaged goods from established Indian FMCG and D2C suppliers, plus fresh items from cloud kitchens, assembled into an assortment designed around six eating occasions a day. Replenishment is the partner's job, with a published expectation of one to two hours daily for up to five machines. This is the structural reason the model scales without the company building routes: the route is somebody else's spare capacity.

### The trade they made

**Won.** Identity on every single transaction, from the first purchase, with no separate enrolment step. A 68% repeat rate and customer value it can measure rather than infer. Scale without owning the machines. And operating leverage: fleet up 62% in FY26 while EBITDA rose about fivefold.

**Gave up.** Every casual buyer who will not install an app — which in a transient site is most of them, and is why the estate is overwhelmingly captive. Also the footfall a single brand cannot generate alone: the assortment spans 190+ brands because one would not fill the machine.

### The pivotal fact

Identity is not optional and the business did not suffer for it. The company reports 68% repeat customers and customer lifetime value to acquisition cost at roughly three times its e-commerce benchmark.

The financial trajectory matters as much. Daalchini turned profitable in FY25 on a net profit of INR 2.5 crore against a loss of INR 8.58 crore the year before. In FY26 it reported revenue doubling to INR 140 crore, with EBITDA rising about 5.7 times to roughly INR 21 crore, while the fleet grew 62%. **Fleet up 62%, EBITDA up more than fivefold** — that gap is operating leverage arriving, and it is the clearest published evidence that density, not machine count, is what makes this category work.

### What it does and does not evidence

**Does:** that a mandatory app gate is survivable, and profitable, in captive sites where the app is also the payment rail; that franchisee-funded fleets scale faster than company-owned ones; that revenue diversification beyond product margin is how these operators reach profit; that density produces operating leverage.

**Does not:** that a gate works in transient or street-facing sites — Daalchini's estate is overwhelmingly captive. Nor that a single-brand fridge behaves like this: Daalchini aggregates 190+ brands, which concentrates footfall in a way one brand's fridge cannot. Nor anything about cross-border entry; this is a domestic operator in its home market.

**Transfer caution.** Indian labour and rent costs make partner-operated replenishment cheaper than it will be in most developed markets. The payback guidance of 10–18 months should be read against that.

---

### Source confidence

| Claim | Weight |
|---|---|
| Net profit INR 2.5 crore FY25; prior-year loss INR 8.58 crore | **Filed accounts** (consolidated) |
| FY26 revenue INR 140 crore; EBITDA ~INR 21 crore; ~5,500 machines | **Company-disclosed** — CEO statement to a trade outlet, not yet filed |
| >95% partner-owned; two business models | **High** — consistent across independent coverage and company material |
| 68% repeat; CLV/CAC ~3× | **Low-medium** — CEO statement, no published methodology |
| 10–18 month payback; 1–2 hours daily per five machines | **Low** — franchise recruitment material |
| Brand and client rosters | **High** — repeatedly named in independent coverage |

**Two reporting bases.** Daalchini's filed accounts and its CEO-stated figures are not the same series. Filed FY24 accounts show standalone operating revenue of INR 18.3 crore and consolidated INR 37.31 crore; the INR 70 crore FY25 and INR 140 crore FY26 topline figures come from CEO statements. The profit figures reconcile to the consolidated filings. Use the profit trajectory with confidence and the topline with the qualifier attached.

### Sources

1. Inc42 — "Daalchini's FY26 Revenue Doubles YoY To ₹140 Cr" (April 2026) — FY26 revenue, EBITDA, fleet, FY25 profit
2. Inc42 — "Exclusive: Daalchini Turns Profitable, Posts INR 2.5 Cr PAT In FY25" (April 2025)
3. Inc42 — "Daalchini's Standalone Revenue Jumps 56% To INR 18.3 Cr in FY24" — filed standalone and consolidated accounts
4. YourStory — company profile and CEO interview (November 2024) — fleet history, revenue series, data practices
5. SME Futures — CEO interview — repeat rate, CLV/CAC, monthly active users
6. Company franchise material — the two models, payback guidance, operating time per machine

*Currency approximate at INR 85 ≈ US$1. All sources in English.*

---

## Annex B — Japan's Restaurant-Owned Frozen Vending Machines (冷凍自販機)

**Why it is here.** Ten thousand established food businesses bought an unattended box to sell their own product. It is the largest documented population of food brands choosing this format, it has the clearest published operator economics in the set — and it is the clearest demonstration of what happens when the customer is not captured.

---

| | |
|---|---|
| **Scale** | Over 10,000 units shipped (Sanden「ど冷えもん」, since early 2021) |
| **Who funds the units** | The restaurant or food manufacturer itself; ~90% of buyers |
| **Access** | None — cash and card. No identity captured |
| **Who holds the customer** | Nobody. The sale is anonymous and stays that way |
| **Product** | The operator's own frozen product; typical ticket ¥800–1,000 |
| **Sites** | Beside the operator's own shop, car parks, roadsides |
| **Access Gate score** | 2/3 — product pull is exceptional and the destination sits ten metres away; only the friction absorber is missing |
| **Evidences** | Chapter 3 (the cost of anonymity), chapter 5 (frozen economics) |

---

### How it works

An existing restaurant or food maker installs a frozen machine, usually on its own frontage, and sells packaged versions of what it already cooks. The machine runs when the shop is closed. Nothing else about the business changes.

Adopters include major chains — 餃子の王将, リンガーハット, 松屋フーズ with frozen beef-bowl and curry lines, 丸山製麺's famous-ramen network — alongside a long tail of independents.

**Supply and route.** This is the case's most transferable feature and the reason its economics work. There is no route: the kitchen is metres away, and restocking is a staff task rather than a logistics operation. Production is batched into slow periods and frozen, converting idle kitchen capacity into inventory with one to three months of shelf life. Machine cost runs around ¥2m, electricity ¥7,000–9,000 a month, depreciated over five years.

### The trade they made

**Won.** Revenue in the hours the shop is shut, on a kitchen, a brand and a frontage already paid for. Near-zero waste on one to three months of shelf life. A bounded downside of roughly ¥40,000 a month, and a machine that can be moved.

**Gave up.** Every customer. Ten thousand operators, no records. They cannot tell whether a machine buyer ever ate in the restaurant ten metres away, cannot bring one back, and cannot prove the halo they believe in. This is the only case in the set where the trade looks one-sided — and the operators did not choose it so much as inherit it from the equipment available.

### The pivotal fact

Published trade models put monthly operating profit at roughly ¥50,000 at 300 meals and ¥230,000 at 900, against fixed costs of about ¥40,000 a month — so almost all incremental revenue falls through. One roadside operator running two machines reported around ¥550,000 a month at over 30% margin with roughly eight-month payback. Operators have reported that about 20% of sales arrive after normal closing hours.

And then the loss. Almost all of these machines take cash. The restaurant is ten metres away — the destination condition is satisfied completely — and nothing is captured, because a vending stack records a transaction, not a person. Ten thousand careful operators reached the same outcome because the equipment available to them made no other outcome possible.

**This is why the case scores 2/3 rather than 0.** These operators are not missing the conditions that make identity worth having. The product pull is exceptional and the destination is in sight of the machine. The only thing absent is anyone to absorb the friction of joining — and because nothing asked, nothing was recorded. A case that fails the test tells you little. A case that passes two of three conditions and still ends the sale anonymously tells you the gap is in the equipment, not the opportunity.

### What it does and does not evidence

**Does:** that established food businesses adopt unattended units at scale to extend their own brand; that frozen product from an existing kitchen removes both the waste problem and the route problem; that after-hours trade is genuinely incremental; that fixed costs are small enough to make the downside bounded.

**Does not:** anything about entering a market where the operator has no kitchen. Every one of these units sits beside an existing restaurant — which is precisely why its supply model is cheap, and precisely why it cannot be read as a market-entry precedent. Nor does it evidence anything about identity, except by absence.

**Transfer caution.** Japanese consumer acceptance of restaurant-quality frozen food is unusually high, and the trade profit models are illustrative rather than audited. The spread between the ¥50,000 and ¥550,000 outcomes is site quality, not machine specification.

---

### Source confidence

| Claim | Weight |
|---|---|
| 10,000+ units; ~90% bought by restaurants and food makers | **Medium-high** — industry body and manufacturer statements |
| Machine cost, electricity, depreciation | **Medium** — dealer and vendor published figures |
| Profit at 300 and 900 meals per month | **Low-medium** — illustrative trade-press models, not accounts |
| ¥550,000/month, ~8-month payback | **Low** — single operator account |
| ~20% of sales after closing hours | **Low-medium** — trade journal report |
| Named adopters | **High** — publicly announced deployments |
| Machines take cash and capture no identity | **High** — product specification |

### Sources

1. Japan Chamber of Commerce / Sanden Retail Systems — cumulative unit shipments and buyer profile
2. Sanden Retail Systems product material — machine specification, running cost, customisable exterior
3. Japanese vending trade press and dealer material — operator profit models, electricity, depreciation
4. 「激流」trade journal (2022) — after-hours share of sales
5. Operator interviews in Japanese business media — the roadside two-machine account
6. Company announcements — 餃子の王将, リンガーハット, 松屋フーズ, 丸山製麺 deployments

*Currency approximate at ¥155 ≈ US$1. Sources 2–5 are Japanese-language.*

---

## Annex C — Felfel (Switzerland)

**Why it is here.** It is the only operator in this set where a third party absorbs the cost of enrolment. The employer signs, and the employee arrives already a member — which makes it the reference design for any brand that wants identity without friction at the door.

---

| | |
|---|---|
| **Scale** | ~800 fridges, ~600 corporate clients, ~150,000 people served daily |
| **Who funds the units** | Felfel owns them; the employer pays a monthly service fee |
| **Access** | App, with enrolment effectively handled at the employer level |
| **Who holds the customer** | Felfel, via the employer's contract |
| **Product** | Chilled fresh, ~25 SKUs rotated weekly; Felfel cooks nothing itself |
| **Sites** | Offices, production sites, hospitals, schools — 25 to 250 people per fridge |
| **Access Gate score** | 3/3 — the employer absorbs friction, the alternative is a long walk, and the app is the destination |
| **Evidences** | Chapter 2 (the gate), chapter 4 (owned relationship) |

---

### How it works

Felfel is a food brand that does not cook. It co-creates exclusive dishes with Swiss family-owned producers, tastes every dish at its own premises, and controls the assortment, the fridge, the logistics and the app. The supplier gets volume and a shelf; Felfel gets culinary identity without a factory.

The customer is the employer; the consumer is the employee. The employer pays a monthly service fee covering the machine, logistics and daily restocking, and chooses the subsidy level — full, partial or none. Employees pay the balance in-app. Reported pricing is around US$1,000 per week per client, or from roughly US$1.50 per employee per day. The alternative being displaced, a conventional staff canteen, is described as requiring investment on the order of CHF 1 million.

**Supply and route.** Production is distributed across partner producers rather than a central kitchen, and Felfel runs daily restocking into a dense urban estate. Waste on a daily-fresh assortment is attacked three ways: data-driven daily replenishment, a late-afternoon half-price window, and donation of the remainder to a national food bank, with the tail converted to biogas.

### The trade they made

**Won.** Enrolment without friction, because the employer completes it. Recurring service-fee revenue that does not depend on how many people buy on a given day — the protection that matters when a site empties. And daily fresh at scale, with a waste routine designed before launch rather than after.

**Gave up.** Everyone outside a contracted workplace. There is no walk-up trade, no street presence and no impulse purchase, and growth moves at the speed of a B2B sales cycle rather than a consumer one. The model also forgoes owning production, which caps margin in exchange for capping risk.

### The pivotal fact

Felfel did not persuade 150,000 individuals to register. **It persuaded 600 companies, and the individuals arrived enrolled.**

That single design choice removes the conversion risk that every other gated operator carries. It also changes what the gate is for: where the employer subsidises a meal, the app is the only route to the subsidy, so membership is the benefit rather than the toll.

### What it does and does not evidence

**Does:** that the friction of identity can be moved off the shopper entirely by contracting at the B2B layer; that a service-fee model decouples base revenue from daily footfall, which is what protects an operator when a site empties; that chilled fresh is viable at scale with a designed waste routine; that curation can substitute for owning production.

**Does not:** that this works outside a high-price, high-density, high-trust office market, or where employers do not already treat food as a funded benefit. Nor does it evidence anything about franchising — Felfel is vertically integrated and does not franchise. Nor about single-brand estates: the assortment rotates weekly precisely because a captive audience tires of a fixed menu.

**Transfer caution.** Swiss willingness to pay for quality fresh food is among the highest in the world. The model's transferability rests on the employer-subsidy culture more than on the technology.

---

### Source confidence

| Claim | Weight |
|---|---|
| ~800 fridges, ~600 clients, ~150,000 daily | **Medium** — co-CEO statement in trade interview; company site frames it differently elsewhere |
| Service-fee model, employer-set subsidy, 25 SKUs weekly | **High** — company-published and independently described |
| ~US$1,000/week; from ~US$1.50/employee/day | **Medium** — one press report plus company material |
| CHF 1 million canteen comparison | **Medium** — trade interview |
| Waste routine: Happy Hour, food-bank donation, biogas | **High** — company-published |
| Claimed employee-engagement uplift | **Low** — vendor marketing; not used in the paper |
| Share of the Swiss office-lunch market | **Low** — press characterisation, no methodology; deliberately not used in the paper |

### Sources

1. Food Inspiration — co-CEO interview: fleet, clients, daily reach, model
2. Felfel company site — service model, subsidy mechanics, supplier curation, waste process
3. WorkLife — pricing characterisation and market commentary
4. Swiss and international trade press — canteen investment comparison, US market entry
5. Schweizer Tafel — food-bank donation partnership

*Currency approximate at CHF 0.80 ≈ US$1. Sources in English and German.*

---

## Annex D — Farmer's Fridge (United States)

**Why it is here.** It is the counter-case to Daalchini: the same category, the opposite access decision, and a business that works anyway. It also prices what identity costs when you buy it after the sale instead of capturing it during one.

---

| | |
|---|---|
| **Scale** | 2,000+ locations across roughly 20 cities; ~500 staff |
| **Who funds the units** | The company — over US$100m raised. It does not franchise |
| **Access** | Open. Anyone buys at the screen; loyalty is rewarded, not required |
| **Who holds the customer** | The company, for the share who join the loyalty programme |
| **Product** | Chilled fresh, ~50 items at US$5–14, made daily in own commissaries |
| **Sites** | Airports, hospitals, offices, universities, retail |
| **Access Gate score** | 1.5/3 — the brand part-absorbs the cost of joining (~9% of spend) and the destination exists; the transient population is what fails |
| **Evidences** | Chapter 2 (when not to gate), chapter 5 (venue economics and concentration risk) |

---

### How it works

A vertically integrated fresh-meal brand where the fridge is the storefront. Commissaries produce daily, an in-house fleet restocks, machines display a "made on" date, and unsold stock is removed on a fixed cycle. Beyond its own estate it sells wholesale to retail partners including Target, and runs direct-to-consumer delivery.

**Supply and route.** The most capital-intensive supply model in this set, and the reason the comparison is instructive: own kitchens, own logistics, own machines. That is what daily-fresh at national scale requires, and it is precisely the commitment the paper argues a brand entering a new market should not make on day one.

**Venue economics — the instructive part.** Farmer's Fridge does not apply one commercial structure everywhere. Hospitals often take units as a subsidised amenity, because the institution wants 24-hour food for staff and families. Airports and retail treat them as a traffic driver, where the host benefits from the offer existing. Offices and campuses share a percentage of sales with the site. The same machine is sold to three different buyers with three different arguments.

### The trade they made

**Won.** The sale from anyone walking past, in airports and lobbies where a sign-up demand would have lost it. Venue terms flexible enough to place units in institutions that would never pay a revenue share. Reach across roughly 20 cities.

**Gave up.** Roughly 9% of spend, permanently and on every transaction, to attach identity after the fact — plus every customer who buys and never joins. This is the only place in the set where the cost of the alternative is actually priced, which is why it is worth stating plainly rather than treating as a failure. For its estate, it is the right trade.

### The pivotal fact

The company pays roughly **9% of spend** — via a loyalty programme worth about US$9 per US$100 — to attach identity *after* the sale. That is the price of the alternative to capturing it at the door: permanent, on every transaction, versus a one-time sign-up.

It is also the right decision for its estate. An airport traveller passes once; demanding registration destroys the sale and gains a customer who will never return to that machine.

**Which condition actually fails is worth being precise about**, because it is not the obvious one. The brand is willing to absorb the cost of joining — that is what the 9% buys — and it has somewhere to send a member once it has one, through the app, the delivery business and the wholesale range. What fails is product pull *measured against a sign-up*: a fresh salad is wanted enough to buy and not wanted enough to register for, when you are walking to a gate. The population in front of the machine is the binding constraint, and it is the only one of the three that the operator cannot change.

### What it does and does not evidence

**Does:** that open access is correct where populations are transient; that identity has a market price a brand can compare against the cost of gating; that venue-differentiated commercial terms materially widen the site pipeline; that channel concentration is the format's real fragility — COVID removed roughly 85% of sales when offices and airports emptied, and survival came from a rapid pivot to delivery.

**Does not:** that a franchisee-funded model works — the company deliberately refuses to franchise daily-fresh. Nor that this is achievable without heavy capital: over US$100m raised to reach 2,000 machines is a venture build, not a self-financing programme.

**Data caution.** Per-fridge revenue and basket figures circulating online appear to be templated or AI-generated with no primary source, and are not used here or in the paper.

---

### Source confidence

| Claim | Weight |
|---|---|
| 2,000+ locations, ~20 cities, ~500 staff | **Medium-high** — company statements in national business press |
| Revenue tripled in three years; 10–15 fridges placed weekly | **Medium** — CEO statement to Crain's Chicago |
| ~85% COVID sales decline; DTC pivot; US$40m round | **Medium-high** — contemporaneous business press |
| Wholesale partners including Target | **High** — company-published |
| Greens loyalty mechanics (~9% effective rebate) | **Medium** — derived from published programme terms |
| Does not franchise | **High** — company position |
| Venue-differentiated commercial structures | **Medium** — company and press descriptions |
| Per-fridge revenue or basket figures | **Not used** — unverified, likely templated content |

### Sources

1. Inc. — company profile (January 2026): footprint, staffing, history
2. Crain's Chicago Business — CEO interview (July 2025): revenue growth, placement rate
3. Bloomberg — COVID sales collapse and delivery pivot
4. Farmer's Fridge company site — menu, pricing, "made on" dating, Greens programme terms
5. Trade coverage — venue structures and wholesale partnerships

*All figures in USD. All sources in English.*

---

## Annex E — Japan's Unmanned Frozen Shops (餃子の雪松 and the category)

**Why it is here.** It is the only case in this set that is a franchise system, and it shows how cheap a franchise rung can be made. It is also the strongest argument *against* gating: the category grew tenfold while asking nothing of the customer.

---

| | |
|---|---|
| **Scale** | Category: 131 stores (2020) → 1,282 (2022). 餃子の雪松: ~19 → 400+ |
| **Who funds the units** | Individual franchisees; many were restaurateurs whose trade collapsed in 2020 |
| **Access** | None — honour box, cash |
| **Who holds the customer** | Nobody. A cash box records nothing |
| **Product** | Own-brand frozen, single SKU — 36 gyoza at ¥1,000 |
| **Sites** | Small vacant storefronts, 2–5 tsubo |
| **Access Gate score** | 1/3 — exceptional product pull, no friction absorber, no destination |
| **Evidences** | Chapter 1 (the low-capital rung), chapter 5 (frozen and licensing) |

---

### How it works

A franchise format stripped to its minimum: a freezer, a cash box and a security camera in a vacant shop. No kitchen, no counter, no staff, no point-of-sale system. Start-up runs around ¥1–2m for a small site — roughly one-third to one-tenth the cost of a conventional restaurant — and some programmes market a two-week path from signing to opening, with no food-service operating permit required for pre-packaged frozen goods.

**Supply and route.** The franchisor manufactures and wholesales; the franchisee restocks a freezer perhaps weekly. 餃子の雪松 charges **no royalty at all**, taking its margin in the wholesale price of the product — a structure far easier to sell to a cautious investor than an ongoing percentage deduction. 大阪王将's competing programme charges on sales volume instead.

Product design carries much of the load. One SKU at a round ¥1,000 eliminates change-making, choice paralysis, planogram management and most reasons to hesitate.

### The trade they made

**Won.** A franchise rung cheap enough to open the investor pool to almost anyone — a tenth of a restaurant, two weeks to trade, no operating permit for packaged frozen. Tenfold category growth in two years. Near-zero waste and no staff.

**Gave up.** All customer knowledge, and with it any ability to remarket, measure repeat or lift the basket. One SKU at ¥1,000 is a ceiling with no upsell mechanism. The format also gave up defensibility: it was copied at scale within two years, which is what a tenfold rise in a category with no moat looks like.

### The pivotal fact

The category grew roughly tenfold in two years, with about 60% of stores opening in a single year, and it did so on an honour box. **Membership is not what made unattended food work.** Where product pull is exceptional and there is nowhere to send a customer afterwards, the frictionless version wins — and any brand proposing a gate should be able to say what the customer gets for it.

COVID supplied both sides of the boom: restaurateurs with collapsed trade and business-restructuring subsidies, meeting consumers who wanted contactless food at any hour.

### What it does and does not evidence

**Does:** that a low-capital unattended format widens the franchise investor pool dramatically; that frozen single-SKU retail can be run by non-specialists; that margin-in-wholesale is a viable alternative to royalties; that packaged frozen carries a materially lighter licensing burden than fresh.

**Does not:** that the format travels to markets with lower social trust — an honour box is a Japanese answer. Nor that it builds anything durable: operators know nothing about their buyers, cannot remarket, and cannot measure repeat. Nor that the growth is stable; a tenfold rise with 60% of stores opened in one year is the signature of a land rush, and some observers already call the category oversupplied.

**Transfer caution.** Single-SKU simplicity caps the basket at ¥1,000 with no upsell mechanism, and shrinkage under an honour system is not publicly quantified anywhere.

---

### Source confidence

| Claim | Weight |
|---|---|
| 131 → 1,282 stores; ~60% opened in FY2022 | **Medium-high** — Teikoku Databank survey, reported in Forbes Japan |
| 餃子の雪松 store counts | **Medium-high** — multiple independent Japanese outlets |
| ¥1–2m start-up; freezer/cash box/camera format | **High** — consistent across franchise and trade sources |
| No royalty at 雪松; volume royalty at 大阪王将 | **Medium** — franchise comparison sources |
| Two-week opening; no operating permit required | **Low-medium** — franchise recruitment material; varies by prefecture and format |
| Shrinkage under honour system | **Not published** — unknown |

### Sources

1. Teikoku Databank survey, reported in Forbes Japan — category store counts
2. TOUCH TO GO and Japanese business media — 餃子の雪松 history and store count
3. Japanese franchise comparison sites — start-up cost, royalty structures, opening timelines
4. 餃子の雪松 and 大阪王将 franchise material — format, product, terms
5. Japanese trade coverage of the 2020–22 unmanned boom — COVID subsidies and operator origins

*Currency approximate at ¥155 ≈ US$1. Most sources are Japanese-language.*

---

## Annex F — Genki Forest 元气森林 Smart Cabinets (China)

**Why it is here.** It is the clearest example of a brand building its own unattended estate specifically to own the customer relationship rather than to add a sales channel. It is placed last because its performance figures are brand-supplied with no published methodology — the reasoning is instructive, the numbers are not evidence.

---

| | |
|---|---|
| **Scale** | Fleet in the tens of thousands; announced plans of 80,000 rising to 100,000 coolers and cabinets |
| **Who funds the units** | The brand, which also bought into cabinet manufacturing |
| **Access** | Scan to open; the shopper is invited to join a brand channel afterwards |
| **Who holds the customer** | The brand, for those who join after buying — an unknown share |
| **Product** | Beverage — the only non-food case in the set |
| **Sites** | Offices and apartment buildings, deliberately away from convenience stores |
| **Access Gate score** | 2/3 — friction near zero since scanning is the payment motion; weak product pull; strong destination |
| **Evidences** | Chapter 4 (what identity unlocks; the network as a channel) |

---

### How it works

Genki Forest could not win cooler placement in China's convenience chains, where incumbents had spent decades and billions — Coca-Cola alone holds hundreds of thousands of coolers and vending machines there, and Nongfu Spring covers millions of retail terminals. So it opened a different front: its field teams shifted from placing ordinary coolers into shops to placing smart cabinets into offices and apartment buildings, where no incumbent had yet fought.

The purpose was explicit. A shopper scans to open, is invited to add the brand's enterprise messaging account, and lands in the brand's private traffic pool — reachable thereafter at near-zero marginal cost. That relationship is unavailable through a convenience-store sale, where the retailer owns the customer.

**Supply and route.** Beverage is the easiest possible product for this: shelf-stable, no cold-chain fragility, no expiry management, high frequency, low ticket. The brand went vertical to support the estate, adding vending leasing to its business scope, acquiring a cabinet factory and investing in a refrigerator manufacturer.

### The trade they made

**Won.** An uncontested channel in offices and apartments, where incumbents' hundreds of thousands of coolers counted for nothing. A messaging audience the brand can reach at near-zero marginal cost — unavailable through any convenience-store sale. And eventually a network other brands pay to launch into.

**Gave up.** Capture at the moment of sale. Joining happens after the purchase and by invitation, so an unknown share of buyers never joins and is never counted. For a low-ticket drink that is a rational trade; for a branded meal it would leave more on the table.

### The pivotal fact

Once the network existed, **it became a channel the brand could sell to others.** Starbucks' ready-to-drink line used Genki's cabinets to launch a new product, with stated intent to use the network for further launches and to follow it into inland markets.

The second lesson is about restraint. The cabinet business is reported to have reached full-year profitability in part *because* it did not expand aggressively — with site selection, assortment, warehousing and fine-grained operations named as the continuing challenges rather than scale.

### What it does and does not evidence

**Does:** that a consumer brand will spend real capital to own a customer relationship a retailer would otherwise keep; that an unattended estate becomes a second-order asset other brands will pay to access; that disciplined footprint growth, not maximum placement, is what reaches profitability.

**Does not:** anything about food. No spoilage, no cold-chain complexity, no food-safety licensing, low ticket, high frequency — a ready-to-eat meal estate inverts nearly all of it. Nor does it evidence a technology moat: as Chinese analysts noted at the time, if Genki could deploy smart cabinets so could Nongfu and Coca-Cola, and they did.

**Numbers caution.** The reported sales lifts — around 118% for outlets with a cooler versus without, and around 74% for smart cabinets over conventional coolers — originate with the brand and have no published methodology or control group. Use the strategic reasoning; do not lean on the percentages.

---

### Source confidence

| Claim | Weight |
|---|---|
| Strategic shift to smart cabinets; office and apartment targeting; messaging-channel routing | **Medium-high** — consistent across Chinese business press citing internal staff |
| Announced fleet plans (80,000 → 100,000) | **Medium** — founder statements at distributor events |
| 118% and 74% sales-lift figures | **Low** — brand-supplied, no methodology |
| Hardware premium of roughly RMB 300–400 per unit | **Low-medium** — press citing an unnamed informed source |
| Full-year profitability of the cabinet business | **Medium** — Chinese financial press |
| Starbucks RTD launch through Genki cabinets | **Medium-high** — reported with a named executive |
| Competitor cooler counts | **Medium** — widely cited industry and prospectus figures, varying by year |

### Sources

1. Chinese business press (Jiemian, Sina Finance and others) — channel strategy, cabinet deployment, profitability
2. Founder statements at distributor conferences — fleet targets
3. Chinese trade coverage — hardware premium, vertical integration into manufacturing
4. Interview coverage with Starbucks RTD channel leadership — the cabinet launch
5. Industry and prospectus data — competitor cooler and terminal counts

*Currency approximate at RMB 7.1 ≈ US$1. Most sources are Chinese-language.*