Credit where no bank will look.
600,000 retailers without a credit score any lender trusts. Serai builds the behavioral data layer. The wholesale marketplace is how we capture it.
01The problem
The business is real.
Stock turns, staff get paid, the shutter goes up every morning. Boutique retail moves billions of rupees of inventory a year.
The paperwork is not.
Turnover on the books is a fraction of turnover through the till. That is how this market has always run.
So on paper they look like a bad risk.
The best-funded attempt at this market wrote off its way back out of it. Every large NBFC has a version of the same story.
Not one bank will underwrite them.
More than 100,000 boutique retailers run real businesses and are uncreditworthy on paper. Paper is the only thing a lender has ever been able to pull.
A lender reads what the borrower wrote. What predicts repayment happens inside a transaction the lender is not part of.
Sell to them yourself and every line above is observed. The three that matter most arrive with the first order.
of what predicts repayment is visible from outside the transaction.
- GST turnover, FY filed
- ₹18,40,000
- Bank credits, 12 mo
- ₹22,10,000
- ITR declared income
- ₹9,60,000
- Collateral offered
- None
- Bureau history
- Thin file
- How often they reorder
- How fast they pay
- What they buy
What predicts repayment is how a shop buys, not what it files.
None of it appears in a filing. All of it shows up the moment a shop starts ordering from you.
How often they reorder
Back in 15 days, or 45?
This moves first — weeks before a filing would show anything, months before a default.
How fast they pay
Paid on day 28, or day 58?
Not whether they pay in the end, but when. A shop drifting later every month is in trouble long before it misses.
What they buy
A wider basket, or one line?
A shop adding categories is growing. A shop down to one is buying only what it knows will sell.
Nobody has this data, because nobody sells to these shops and waits to be paid.
Brands will not ship without payment. Shops will not order without terms. Someone has to put their own cash between the two from the first order. Nobody would. We do.
We pay the brand on day 7 and collect from the shop on day 30.
The commerce is real and it funds the company. We chose terms that put our own cash in the gap: brands paid in 7 days, retailers paying us in 30, and twenty-three days carried on our own balance sheet.
Order lifecycle
30 minutes to 30 days- 01First 30 minutes
A brand lists and starts trading
It uploads a catalogue, accepts the terms and takes its first order. No integration, no account manager, no negotiation.
- 02Day 00
A retailer orders on Net-30
The shop pays nothing today. No collateral, no personal guarantee, no paperwork beyond the order itself.
- 03Day 07
The brand is paid in full
Out of our own balance sheet, against a confirmed order, twenty-three days before the retailer owes us anything.
- 04Day 30
Repayment falls due
Paid on time, paid late, paid in part, or not paid at all. Whichever it is, we are the ones it happens to.
We run all four of these steps ourselves, so we see all four. Thirty days after an order we know whether the shop paid, how many days late it was, and how much came back.
One order, on its own
- Out to the brand · day 07
- −₹1,00,000
- The shop has put up nothing, and has paid us nothing yet.
- Back from the shop · day 30
- +₹1,00,000
- 23 days later, and the repayment is a record only we hold.
Illustrative. Assumes a steady ₹1 Cr of orders a month, brands paid on day 07, shops paying on day 30. Every order is out for 23 days, so what we have out settles at one day of orders — ₹3,33,333 — times those 23 days, or ₹76,66,667. That is what the money buys: at the end of every one of those 23 days we know whether the shop paid, and no one else does.
The trade
Brands
They hand over the billing relationship. The brand invoices Serai and not the shop, so it is paid on a fixed date and never has to find out whether the shop paid at all.
- Paid in
- 7 days
- Live in
- 30 minutes
- Credit risk carried
- None
- Shops in reach
- 100,000+
Serai
We put up the cash. ₹1,00,000 an order, out of our own balance sheet, twenty-three days before any of it comes back, and all of the risk that it does not.
- Cash out per order
- ₹1,00,000
- Days carried
- 23
- Credit risk carried
- All of it
- Repayments recorded
- 1 per order
Retailers
They agree to be seen. Every order, part-payment and late day is recorded against the shop, and that record is what takes its limit from ₹10,000 to ₹1,00,000.
- Pays in
- 30 days
- Upfront
- ₹0
- Collateral, guarantee
- None
- Limit
- ₹10,000 → ₹1,00,000
Every order we clear is a row nobody else has.
Paid on time, paid late, part-paid — each one attached to a shop we have watched since its first order. Carrying the cash is what buys the data.
What we see
How often they reorder
0.41days between orders
How fast they pay
0.69which day inside Net-30 they pay
What they buy
0.38how many categories they order
Score and limit
Score
388/1000
Limits by score band
Credit limit today
₹10,000
Score 388 falls in band D, so the limit follows from the table. Nobody at Serai approves it, and nobody at the shop is asked for collateral.
Illustrative record. Limits are set by a published table today and move only on repayment we have observed. Learned scoring begins at 18 months of transaction data.
Cumulative labelled repayment events
2,40,000
by month 18 — enough to train on
Illustrative projection
1L = 1,00,000 events
Illustrative. The shape is the argument: labelled events accumulate only for whoever is carrying the float while the money is owed.
Months from public launch · 01 Oct 2026
Rules
₹10,000 → ₹1,00,000
A published table. Repayment history moves a retailer up a band automatically — no discretion, no relationship manager, no file to argue with.
Labels
2,40,000 labelled events
Order, reorder, part-payment, delay, resolution. Each one a row with an outcome attached, against a shop no bureau holds a file on.
Learned scoring
18–24% p.a.
Enough labelled events to train on rather than legislate for. The three observed signals stop being heuristics and become features in a model — one that prices embedded retailer credit, co-lent with an RBI-registered NBFC.
Demand showed up before the product did.
Nothing here is scale. It is people committing money before there was anything to buy.
50+
letters of intent
Signed before launch, including brands that came off Shark Tank India.
80
onboarding before launch
Boutique retailers joining before the marketplace opens. Demand before supply.
We built the marketplace first. The credit business is what it turns into.
We are raising to sign more brands before the October launch, and to reach eighteen months of repayment data as fast as we can.