NOVA
SYNTHETIC DATA · FINANCIAL INCLUSION · WEST AFRICA

Banks have the data.
Builders have none.

Sensitive customer data can’t leave the building, so the models that could widen financial inclusion in West Africa never get trained. NOVA breaks the deadlock: generate privacy-safe, realistic financial data on demand, from your own rules or from a file you already have.

THE DATA DEADLOCK

80%
of AI projects stall on data problems
Gartner
60%
of AI budgets spent preparing data
McKinsey
70%
of banks cite privacy as their #1 AI barrier
Deloitte
$20B
projected synthetic-data market by 2030
MarketsandMarkets

TWO WAYS TO MAKE DATA

Create from nothing

Tell NOVA what you want: the columns, and the rules behind them, like “a brand-new account making a large international transfer is likely fraud”. It builds realistic records from nothing. No dataset needed. Seven ready-made domains, or define your own.

Copy from real data

Already have a real dataset? NOVA studies its patterns (the ranges, the relationships, the unusual cases) and creates a brand-new set that behaves the same way, but belongs to no real person.

WHAT YOU CAN GENERATE

Loans
Credit scoring
Transactions
Fraud detection
Insurance
Actuarial modelling
Remittances
Economic analysis
Macro
Economic indicators
Investment
Portfolio risk
Corporate
Credit analysis
Your own domain →
Define columns + rules

HOW IT WORKS

01

Describe what you need

Pick a domain (loans, transactions, insurance, or define your own) and tell NOVA the columns you want and the rules behind them, like “rural schools have lower pass rates”.

02

Generate new data

NOVA creates brand-new records that follow your rules. No real data required, just domain knowledge. Or upload a file you already have and get a safe synthetic twin.

03

Check it’s good

NOVA checks four things: does it look realistic, are the relationships right, can you train a model on it, and is it truly private? All four pass.

04

Download and use

Get a CSV you can use straight away, for training, testing, or sharing, without exposing a single real customer.

EVERY BATCH IS CHECKED FOUR WAYS

Does it look real?

The shape and spread of every field matches reality.

Are the relationships right?

Income still drives loan size; risk still drives default.

Can you actually use it?

A model trained on it performs almost as well as on real data.

Is it truly private?

No record is a copy of, or traceable to, a real person.

Try it yourself.

Pick a domain, or upload your own CSV. No sign-up.

Open the studio →