Avra @ Bloomberg Línea Summit · Nov 23See the agenda

Cases · Retain

How a leading utility found that 78% of its commercial churn was still recoverable

By classifying every lost business customer, Avra showed that 78% of commercial disconnections came from companies still operating, and flagged business closures with 98% precision.

Client
A leading utility
Stage
Retain
Measured on
Two and a half years of disconnections
Status
Blind test next

The challenge

A leading Brazilian utility loses commercial customers every month. When a business disconnects, the reason is self-reported, so the company cannot tell a business that closed its doors from one that is still operating and may have switched supplier.

Without that distinction, the retention team treated every disconnection the same way and spent effort on businesses that no longer existed.

What Avra built

Avra built a model that classifies each disconnected business as recoverable (still operating) or closed. It uses the company's CNPJ and signals from Avra's foundation model, such as registration status, lawsuits and protests, regional economic activity and the behavior of similar companies.

The model was applied to two and a half years of commercial disconnections. Self-reported reasons turned out to be a weak guide: the share of closed businesses was about the same for every reason given, between 20% and 26%.

Results

For the first time, the retention team can see which lost customers are worth going after.

78%
of commercial churn came from businesses still operating, a base retention can act on
98%
precision when flagging a business closure
89%
accuracy when flagging a business as recoverable

Next

The next step is a blind test, followed by a model that predicts disconnections before they happen.

Win back the customers you can still keep.

Avra models churn on your own customer base, enriched with what happens to each company outside your walls. Talk to our team to find out which of your lost customers are still within reach.

Talk to sales