Avra @ Bloomberg Línea Summit · Nov 23See the agenda

Cases · Price

How a leading payments company sets the right merchant fee from day one

Avra predicts how much each new merchant overstates its expected volume at signup, placing 80% of leads in the right or adjacent band and reaching up to 92% accuracy where pricing errors cost the most.

Client
A leading payments company
Stage
Price
Measured on
Out-of-time test on recent leads
Status
Under evaluation for onboarding

The challenge

A leading Brazilian payments company sets each new merchant's fee based on the monthly volume the merchant declares at signup. Merchants who overstate that volume get a fee that is too low.

Around the third month, the company reprices them to match their real volume. Many leave right after, so the company pays to acquire customers it ends up losing.

What Avra built

Avra built a model that predicts, for every lead, how far the declared volume is from the volume the merchant will actually process. The answer comes as one of eight bands, from accurate to heavily overstated, available at signup.

The model combines the client's onboarding data with signals from Avra's foundation model on each CNPJ, such as registration data, legal records and proprietary sources. It was tested out of time on thousands of recent leads.

Results

With the overstatement band known at signup, the first offer can start at the right price and avoid the repricing that drives early churn.

80%
of leads placed in the right or adjacent band
92%
peak accuracy, in the bands where overstatement is highest
0.90
ROC AUC across the eight bands

Next

The client is now evaluating the model to set fees at onboarding.

Price every new customer right.

Avra turns the CNPJ of every lead into a prediction your pricing and onboarding teams can use on day one. Talk to our team to see where your current pricing leaks revenue.

Talk to sales