Measure
An applicant completes a short gamified simulation inside the partner’s existing application flow. Every decision is a revealed choice with a consequence inside the simulation — there are no questionnaires, and nothing is self-reported.
How it Works
Patent PendingThe BYLAS™ Engine
Credit has two axes: capacity and willingness. Data covers the first. The BYLAS Engine was built to measure the second — directly, quantitatively, from behavior.
Why willingness has never been measured
Capacity — can they repay? — is increasingly commoditized. Bank transaction data, open banking, and platform records already measure cash flows well. Willingness — will they repay? — has only ever been inferred from history. A credit bureau reads a file and assumes the past predicts the person.
An applicant without a file has nothing to infer from. That is how the IFC’s estimated US$5.7 trillion financing gap for micro, small and medium enterprises in emerging markets is built: not from people who will not repay, but from people no one can read. The BYLAS Engine measures willingness directly, from revealed choices in a gamified simulation, so a thin file is no longer a blank one.
What we measure
Behavioral dimensions from prospect theory
Prospect theory describes how people actually decide under uncertainty — not how a textbook says they should. The BYLAS Engine estimates each applicant’s behavioural profile across these dimensions.
- A
Risk sensitivity
How the applicant’s valuation of outcomes curves as stakes grow — the shape of their appetite for gains and their tolerance of downside.
- B
Loss aversion
How much more heavily a loss weighs than an equivalent gain when the applicant decides. Our field evidence shows this is a state as much as a trait: it shifts under stress, which is why the engine measures it in context rather than assuming it is fixed.
- C
Probability weighting
How the applicant distorts likelihoods — overweighting rare events or discounting probable ones — when choosing under uncertainty.
How we measure it
From revealed choices to a credit decision
01 02 Estimate
A Bayesian engine takes the full record of choices and extracts the applicant’s behavioral decision signals — risk sensitivity, loss aversion, and probability weighting. Together they form a quantitative signature of how this person makes financial decisions.
03 Underwrite
These signals power second-look underwriting on the partner’s declined applicants — pricing the willingness axis that transaction data cannot see, and returning a decision on an anonymized key via API.
Working With Lenders
We ask partners to believe nothing. We take the risk first.
Stage 1
Shadow validation
Nothing changes for the partner or the applicant. The simulation is embedded in the live origination flow, the partner decides exactly as before, and we observe. Our predictions are then matched against real repayment outcomes — not proxies — in a genuinely high-stakes context.
Stage 2
Second look
Once validated, we underwrite the partner’s rejected applicants on our own balance sheet, holding the full loan economics. Servicing and collections ride the partner’s existing processes, so there is nothing new to build or staff. For the partner, that means incremental revenue from applicants they already decline — with zero change to their approved book.
We never ask a partner or an investor to believe what we have not risked our own capital on. Our own balance sheet is the argument.
Our commitments
What we never do
No questionnaires or self-reported data
The engine reads revealed choices in a gamified simulation. We never ask applicants to describe themselves.
No personal data taken
Scoring runs on anonymized keys via API. Names, identifiers, and documents stay with the partner.
Partners own their raw data
Origination and repayment data remain the partner’s property, inside the partner’s systems.
Ready to see what your decline flow is worth?
We take online meetings with lenders and investors.