Judge the decision, not only the outcome.
FromOutcomes→ToDecisions in context
A good decision can have a bad outcome, and a repayment record can’t tell the difference. This shift asks whether transaction records could carry the context of each decision, so that finance can judge the decision as well as the result.
A miss that was the right call
Meera’s first full cycle illustrative. Her projections supported 5,000 packs. She orders 2,000, because the monsoon roads into two of her districts are unreliable. 1,700 sell and 300 are left over. As an outcome, that is a shortfall.
In context, a landslide closed one route for eleven days, and the stock she chose not to order would have left a far larger unsold pile. It looks like a miss. The decision was sound.
Decision, context, outcome
“What happened?”
“What was decided, in what conditions, and what happened?”
| Today | On an open network |
|---|---|
| Outcomes | Decision, context, capital and outcome |
| A single snapshot | Repeated cycles |
| Repayment | Judgement, execution and response |
One cycle tells you little. Four cycles tell you how someone decides.
How it would work
A contextual evidence schema. Each transaction record carries its context, with a status for how well each element is evidenced: role, category, geography, counterparties, conditions and the capital committed.
- Decision
- context
- capital
- order
- supplier
- fulfilment
- outcome
- settlement
- reputation
One cycle tells little. Repeated cycles build verified performance, and a pattern of sound decisions in hard conditions is worth more to a provider than one good result.
Go deeper
What already exists
Open networks already pass order and fulfilment details between participants, from collection preferences and package dimensions to pickup and delivery confirmation codes[1]. Lenders already read behaviour over time rather than a snapshot. Revenue-based and cash-flow-based lending, offered by fintechs, lets repayments move with a brand’s revenue[2], and Account Aggregator data enabled an estimated ₹3.82 lakh crore of loans in FY26, about 8.4 per cent of retail and MSME lending by value[3]. This shift adds the context behind the numbers.
Scope. We found no evidence on whether decision-and-context records predict next-cycle performance better than outcome-only records. The pilot builds both and compares them.
Whose record it is
A record of someone’s decisions is personal data when it concerns an individual, and commercially sensitive when it concerns a business. The participant owns it and decides who sees it (Shift 7). Consent, correction and erasure apply[4].
Getting it right
Keep evidence honest. It has to be accurate, relevant, hard to manipulate, predictive and worth what it costs to capture.
Capture context where it is already recorded: route closures, weather, delivery logs. Don’t ask people to write explanations.
Don’t let context become an excuse. A claim of context counts only when a record backs it.
Who it helps. Meera is judged on how she decides, not only on what the monsoon did. Providers get a clearer read of who adapts well, which is what a repeat borrower really needs.
What could go wrong
- Context used as an excuse.Anyone can explain a bad result, so context counts only when a source other than the participant has recorded it.
- Too heavy to capture.If it takes effort, small participants won’t do it, so context comes from data the network already holds.
- A good story mistaken for evidence.Models use recorded context only.
- Gaming.Once context matters, it can be staged, so the pilot tests how hard that is.
How we’d test it
Using past transactions from one corridor, build two models of next-cycle performance: one on outcomes only, and one on decisions with their context.
- The test
- Context against outcomes
- What we would measure
- See which model predicts the next cycle better, and by how much. Keep the context fields that add predictive power and drop those that don’t.
- What would end the idea
- Models using decisions and context predict next-cycle performance no better than outcome-only models.
Sources
- [1]ONDC, Harmonizing statuses in cascaded retail & logistics transaction (draft protocol note)Checked; a draft note in ONDC’s public protocol repository
- [2]CredAble, How are fintechs powering India’s D2C expansion? (11 September 2024)Checked; a lender’s own blog
- [3]IANS, India’s Account Aggregator ecosystem drives Rs 3.82 lakh crore loans in FY26 (27 August 2026), reporting a Sahamati reportChecked; Sahamati’s own estimates, as reported by a news agency
- [4]PRS Legislative Research, The Digital Personal Data Protection Bill, 2023Checked; a summary of the Bill as passed by Parliament in August 2023
Spotted a source we missed or a rule we misread? Tell us what we got wrong.
Meera’s evidence is her own. What does it mean when someone else stakes their resources on her product?