The warehouse builds the credit.
The bank moves the money.
This is not a plan to become a coffee bank. It is a plan to turn real inventory into credit collateral that banks accept, markets can transfer, and risk systems can control. The warehouse holds title and sets the standard. The bank lends against receipts it can verify. The core enterprise and the industry fund stand ready to buy in a downturn. Three parties, one job: turn dead stock into working cash flow.
Who contributes what, and who earns what
The trust structure decides the business structure. Each party contributes something the others cannot replace; remove any one and the loop fails.
Credit infrastructure
Real title control, grading, receipt registration, IoT monitoring, daily mark-to-market, disposal execution. We earn the fee for making a receipt trustworthy.
Money and accounts
Lends against verified receipts, runs the closed repayment account, settles supply-chain payments. Earns the spread; never touches the goods, the grading or the storage duty.
Buyer of last resort
Buy-back commitments, disposal priority, counter-cyclical stockpiling. Earns the spread between distress prices and next season's cost — the obligation is also an option.
Independent trust
Independent QC, goods and price insurance, credit insurance, on-site supervision. Paid in service fees — their independence is itself part of the collateral.
Eight steps of the client journey, eight billing points
Every step is both a service and a fee. From intake to settlement, the client pays for trust at every stage.
S1 Intake QC — the production cost of trust
Independent inspection (weight, moisture, defects, SCA cupping), lot grading, batch passport, insurance binding, sealed samples kept 18 months. The QC firm has no relationship with the borrower; the sample is the only physical evidence in a dispute.
S2 Tiered storage — maturity layers, not a funding pool
Active vault (traders, fast turns), term vault (roasters stocking up, lower rate), pledged vault (title locked, bank-supervised), delivery vault (settlement-ready). The layers are physical and service layers — never a maturity mismatch on money.
S3 E-receipt — the IOU everyone recognises
Unique number + central registry entry + batch passport binding + state-machine lock. A bank accepts the receipt because it can be looked up, audited, and released only by instruction — those three facts are what we charge for.
S4 Financing — the gateway, never the money
Receipt-pledge loans, purchase-order finance, storage-rent instalments, roaster equipment-plus-inventory packages, cross-border LCs with warehouse supervision. We feed the bank QC data and a mark-to-market API; the bank prices and lends.
S5 Risk custody — the second door before the stampede
Daily mark-to-market, dynamic LTV corridor, concentration caps, hedge onboarding (futures / options / price insurance), margin management. When prices fall, top-ups trigger automatically — before everyone heads for the same exit. Hedge onboarding is billed as an origination service through licensed futures partners.
S6 Transfer — receipts move, beans stay
Receipt transfer, agency purchase and sales, basis trading, scheduled delivery. Title moves, coffee does not — the direct translation of "paper money saved metal coins from travelling", and the source of the platform's network effect.
S7 Disposal — the counter-cyclical line
A three-tier ladder: core-enterprise buy-back (92% of market) → platform auction (7 days) → industry-fund stockpiling (index −5%). When prices fall, other lines shrink and disposal revenue rises — a built-in portfolio hedge. Executing government and industry-fund stockpile mandates adds a second counter-cyclical earner.
S8 Settlement — organising the cash flow
Escrow accounts for storage fees, margin accounts, closed management of loan repayment, agency payments, farmer payroll, multi-party splitting. All money moves inside bank accounts; the platform issues instructions and splits only.
Eight revenue lines, eight different natures
Design logic: a stable base (storage), a deep moat (registry and standards), elasticity (origination and transfer), a counter-cyclical hedge (disposal), and a long-duration compounder (data).
| Revenue line | Pricing | Payer | Nature | Y3 share |
|---|---|---|---|---|
| Storage | ¥0.14–0.32 /kg/mo | Owner | Base · cyclical-stable | 46% |
| QC & certification | ¥3,000/lot + ¥800/lot/yr | Owner | Event-driven · cost of trust | 6% |
| Registry & supervision | ¥600/receipt + max(¥0.04/kg/mo, 0.3–0.5%/yr of balance) | Owner + bank | Recurring · compliance moat | 8% |
| Financing origination | 2.5%/yr of loan + hedge origination | Borrower | Recurring · cycle-sensitive | 12% |
| Transfer & delivery | 0.8% of value | Buyer + seller | Event-driven · network effect | 9% |
| Cross-border services | 0.3–0.5% of deal + per-document | Importer / exporter | Event-driven · bonded vault, LC, negotiation | 4% |
| Disposal & stockpile | 1.5–3% disposal · 0.5–1% stockpile execution | Bank / fund / government | Event-driven · counter-cyclical | 4% |
| Data & index subscription | ¥60,000/yr/inst + SaaS by capacity + API | Bank / trader / research | Recurring · high margin | 10% |
Annual contribution of three client types
One infrastructure, three contribution profiles. Traders contribute turnover, roasters contribute leasing and financing, co-operatives contribute seasonal intake and data.
| Client type | Profile | Storage | QC/reg/sup | Origination | Transfer/other | Year total | % of value |
|---|---|---|---|---|---|---|---|
| A · Trader | 200t active vault, always in, pledged for fast turns | ¥480k | ¥113k | ¥300k | ¥240k | ¥1,133k | 10.5% |
| B · Roaster | 80t term vault + bean leasing + order finance | ¥134k | ¥10k | ¥539k | ¥0 | ¥683k | 15.8% |
| C · Co-op | 120t intake (8 mo) + agency sales + payroll | ¥134k | ¥20k | ¥0 | ¥224k | ¥378k | 5.8% |
An illustrative model, not a forecast
Assumption-driven, for calibration of direction only. Every parameter must be re-based against actual rents, power prices and labour costs.
| Key assumptions | Y1 | Y2 | Y3 |
|---|---|---|---|
| Climate-controlled capacity (t) | 600 | 2,000 | 5,000 |
| Average utilisation | 65% | 70% | 75% |
| Inventory value (¥10k) | 1,187 | 4,794 | 14,290 |
| Pledged loan balance (¥10k) | 620 | 2,490 | 7,420 |
| Platform transfer volume (¥10k) | 2,500 | 8,000 | 20,000 |
| QC lots / subscribing institutions | 50 / 3 | 150 / 12 | 350 / 30 |
| Total revenue (¥10k) | 175 | 654 | 1,746 |
| Total cost (¥10k) | 156 | 465 | 1,132 |
| EBITDA (¥10k / %) | 19 / 11% | 189 / 29% | 614 / 35% |
Revenue mix, three-year evolution (¥10k)
Y3 revenue composition
Build the second door before everyone rushes the narrow one
When coffee falls, every pledgor sells at once to repay → price falls further → liquidity dies. The answer is not predicting the crisis; it is pre-building the分流 into the contract.
| Trigger | Automatic action |
|---|---|
| C-price −8% | System alert; hedging recommended |
| −12% / coverage > 55% | Partial top-up or partial repayment (negotiable) |
| −15% / coverage > 60% | Margin call T+1, no verbal waivers |
| −25% / coverage > 72% | Disposal ladder starts, no further negotiation |
| Relief mechanism | Design |
|---|---|
| Hedge onboarding | Pledged portfolios matched to futures / options / price insurance; a hedge policy earns a higher LTV |
| Concentration caps | Single client ≤25% of loans, single origin ≤35%, single variety ≤40% |
| Three-tier disposal | Core buy-back (92%) → platform auction (7 days) → fund stockpiling (index −5%) |
| Counter-cyclical credit | Pre-signed rollover and extension frameworks with the bank: refinance, not recall — a client that survives keeps the vault full |
Clear enough that the bank shares the spread
Legal duties must be explicit down to who owes what to whom. Only then will a bank give up part of its spread to play in this structure.
Storage supervision agreement
Daily mark-to-market API · goods released only on the bank's instruction · pledge registration precedes disbursement · annual three-way stocktake. The bank gets collateral it can see, control and sell.
Buy-back and disposal agreement
Buy-back at 92% of market · disposal priority · counter-cyclical stockpiling triggers (index −X%) · fund participation. The core enterprise trades an obligation for an option: beans bought at 92% in a crash are next year's cheap input.
Closed repayment
Sales proceeds enter the bank's escrow → repay the loan first → residual released → platform splits per agreement (storage / supervision / origination). The platform never handles the goods money — structurally eliminating any funding-pool suspicion.
Twelve months, four milestones
Foundation: one bank, one vault, three lots
Sign the supervision agreement and data API with one bank; complete the 600t climate-controlled vault; run three lots end-to-end (QC → passport → receipt → registry).
First loan: run the loop once
First receipt-pledged disbursement, then the full cycle — lend, mark, repay, release. Price index v1 goes live for internal mark-to-market.
Transfer: receipts start moving instead of beans
Receipt transfer and delivery go live; the core-enterprise buy-back and the fund stockpiling framework are signed; bean leasing launches.
Replication: from facility to standard
Data subscription sold externally (target: 3 institutions); second site scouted; standard + index + supervision agreement packaged as a replicable franchise template; a cross-border bonded-vault pilot starts with the partner bank.
Four lines we do not cross
1 · No empty receipts
No intake, no receipt. A hollow receipt is a bounce cheque from the exchange house — every yuan lost at Shanghai steel and Qingdao port traces back to this.
2 · No double pledging
One receipt pledged twice is a run in slow motion. Unique numbering + dual registration + mutually exclusive states — the system refuses, not a clerk.
3 · No funding pool
The warehouse is a venue, not a bank: no deposits, no guaranteed returns, no touching margin. All money moves through bank accounts; the platform only instructs.
4 · No naked price risk
Coffee runs 30–40% annualised volatility. LTV, margin, hedging and a disposal plan must all be live before any disbursement — missing one means no loan.