Risk changes shape at every handoff.
FromOne business risk→ToRisk that moves with the transaction
A lender usually starts with one question: will this business succeed? On an open network a transaction moves through states, and the risk changes with each one. This shift asks whether finance could read those states and price the risk that exists now.
One batch, five questions
Meera’s first batch is 2,000 packs illustrative. Before production, the question is whether the product works. Once Arjun agrees to carry 800 of them, it becomes whether they will sell in his district. When retailers order, it is whether the orders will be fulfilled. Once the packs are delivered, it is whether the retailers will pay. When payment settles, most of the uncertainty is gone.
One batch, five questions, put to different participants at different moments.
Risk with a timestamp
“Will this business succeed?”
“Will this verified order settle?”
| Today | On an open network |
|---|---|
| One business risk, assessed upfront | A distinct risk at each stage |
| Risk fixed at the point of lending | Risk that changes as evidence arrives |
On an open network, risk hands off between participants, just as the goods do.
How it would work
Shared transaction states. Every transaction on the network carries states that the network can read: committed, ordered, fulfilled, delivered, invoiced and settled. Any entitled provider can see which risk an exposure carries now.
Six kinds of risk are worth telling apart: creation; market and inventory; assets and their use; the transaction itself; the counterparty; and, once exposures are pooled, the portfolio. They overlap, and one participant may carry several at once.
Go deeper
What already exists
Finance already treats a transaction differently at different stages. On TReDS, once a buyer accepts an invoice, the Reserve Bank’s Directions give the buyer an unconditional obligation to pay on the due date, and discounting is without recourse to the seller[1]. That is a different risk from an open order. Open networks also already pass status between participants: ONDC’s draft protocol note has the seller app send pickup status and the logistics provider send delivery status[2]. This shift lets finance read those states.
Scope. The ONDC note is a draft. We haven’t checked which states the live protocol exposes to a third party, or whether they match what happened closely enough to price on. We found no data on losses by state.
Rules and access
A transaction state is business data. Who may see it is a matter of consent and contract (Shift 7). Reading a state doesn’t make a participant a lender: whoever finances an exposure does so under their own regulator’s rules. Discounting an accepted invoice on a TReDS platform follows the Reserve Bank’s Directions[1], and financing outside such a platform follows rules we haven’t reviewed.
Getting it right
Keep the states few and plain. Six is enough to start.
Make them hard to fake. A state counts only when the party who would know has confirmed it.
Show the risk, not a score. “Ordered, not yet delivered” tells a provider more than a number.
Who it helps. Providers can fund the stage they understand instead of lending against a whole business. Meera is no longer judged on a risk that has already passed.
What could go wrong
- States that don’t match what happened.A wrong “delivered” would be priced as safe, so a state counts only when the party who would know has confirmed it.
- Too many states.Complexity shuts out small participants, so there are six.
- A false sense of precision.States say where a transaction is, not whether it will succeed.
- States seen by the wrong people.They are business data, so access follows consent (Shift 7).
How we’d test it
Take a sample of past transactions in one category, with their recorded states.
- The test
- Losses by state
- What we would measure
- Check whether losses and delays differ by state, and whether the recorded states match what actually happened. Price one stage on its state alone and compare it with how it would have been priced on the firm.
- What would end the idea
- Losses don’t differ by state, or the recorded states don’t match what happened closely enough to price on.
Sources
- [1]Reserve Bank of India, Reserve Bank of India (Trade Receivables Discounting System) Directions, 2026 (RBI/DPSS/2026-27/406, 23 June 2026), paragraphs 15 and 20Checked; the Reserve Bank’s text
- [2]ONDC, Harmonizing statuses in cascaded retail & logistics transaction (draft protocol note)Checked; a draft note in ONDC’s public protocol repository
Spotted a source we missed or a rule we misread? Tell us what we got wrong.
If risk changes shape at every handoff, who should carry each shape?