Moving to payroll loans under the rules of several banks
ConCrédito was growing with advances on the FGTS birthday withdrawal, a low-risk, fast-approval product. In October, a change to the FGTS rules hit the company's main product. “Basically they said: we're ending the product in 30 days,” summed up Willian Conzatti, ConCrédito's founder. Private payroll lending (consignado CLT), which had been in testing for a few months, became the operation's priority.
The new product called for a different model. Private payroll lending involves credit analysis, audience targeting and more qualified conversion: to close 3,000 to 4,000 contracts, the company needs to reach about 100,000 people a day. Leadership had defined the risk appetite, but it still had to become a documented, configurable policy, which is natural for a product line that was just being born.
From the start, the operation worked with at least two partner banks, with a third in negotiation. Each had its own criteria for minimum employment tenure, payroll margin, restrictions by employer sector and score cut-offs. Applying these rules by hand, lead by lead and bank by bank, would not keep up with the planned volume.
Then there was the cost of data. The analysis requires checking employment, indebtedness, the employer's financial health and score. In a simultaneous-query model, with a 15% approval rate, 85% of the query cost falls on applicants who would not be approved: R$ 85,000 in a batch of 10,000 analyses. With plans to double volume the following half-year, the projection exceeded R$ 1.5 million over 12 months spent on queries for applicants who would not be approved.
A formal policy, engines per bank and cascading analysis
VAAS worked with ConCrédito's leadership and with the credit specialist at the fund manager to formalize the credit policy in parallel with the technical rollout of the engine. The result was a decision engine in three integrated layers.
A formal credit policy. Age range, minimum employment tenure, maximum margin commitment, number of active contracts, score bands and employer criteria were consolidated in a living document, already configured in the engine and adjustable as the risk appetite evolves.
Independent engines for each partner bank. Each bank has its own engine on the same platform. The same lead can be eligible for one partner and not another, and the system handles this automatically.
Cascading analysis. Queries run in increasing order of cost and the flow stops at the first rejection. It starts with eligibility using government employment data, then moves to registration and compliance checks, borrower indebtedness (SCR for individuals) and the employer's financial health (SCR for companies). The score, the most expensive query, only runs for applicants who passed every earlier step. The engine then calculates limits per bank and reaches a decision.
The rules belong to ConCrédito and its partners; VAAS orchestrates the queries and applies the defined criteria. Profiles within the policy are approved automatically, borderline cases go to the credit team's review desk, and rejections are recorded with a full log. Rule changes follow three access levels: view, draft editing and publishing. Only the publishing profile can put a change into production.
Cost per approved customer fell from R$ 67 to R$ 24
Results were measured after the first full batch in the new flow. A batch of 10,000 analyses now costs R$ 36,250, compared with R$ 100,000 in the simultaneous-query model. Savings were R$ 63,750 per batch, and the cost spent on rejected applicants fell from R$ 85,000 to about R$ 21,250.
Beyond cost, the operation gained the structure to scale:
- Fast rollout. The first partner bank's workflow was ready with 90 to 95% of the rules in less than one business day after technical alignment.
- Policy delivered with the engine. The credit policy was built from scratch in parallel with the technical rollout, already integrated into the engine.
- Autonomy without IT. The Head of Credit adjusts risk parameters in minutes, without opening a ticket with development.
- Borrower and employer analysis. Using SCR data for individuals and companies filters out employers with social-security debts, short-term commitments above revenue or in judicial recovery before the score query.
- Auditability. Every decision is recorded with the steps taken, the data queried and the rules triggered.
