Prioritizing collections with data: who to contact, on which channel and when

A collections portfolio has debtors in very different situations: someone who forgot a bill, someone months behind, someone with a high amount and someone with a low one. Treating them all the same spends messages where they pay off least and arrives late where it matters. Prioritizing with data means deciding, every cycle, who to contact first, on which channel and when.

Start with concentration

Before defining the sequence, look at the portfolio. A Pareto chart shows where most of the volume is. By days overdue, it shows whether the problem is in recent or old delinquencies. By value, it shows whether a few contracts concentrate the balance. By region, it shows where the operation needs more attention.

In DeepMiner, the imported portfolio becomes an index and the Pareto comes from a question to the AI assistant: “make a Pareto of outstanding balance by days overdue in the Collections Portfolio”. Count and sum tell different stories, so it is worth asking for both.

One sequence step per range

Each relevant range becomes a step in the sequence, with its own audience, channel, message and timing. A common structure:

• Recent delinquency: a short SMS reminder, right at the start of the delay.
• Intermediate delinquency: a message with negotiation options and action buttons, by RCS where the customer’s device supports it.
• Long delinquency: a contact with a specific proposal and referral to a service desk.

In DeepMiner, each step is a strategy in the index’s flow: filters by range, the channel’s provider and daily, weekday or cron recurrence, with a time zone. The strategy shows the expected records before saving, and the per-run limit controls the volume of each send.

Without repeating the same contact

Contacting the same debtor twice on the same day through the same channel irritates and does not raise the chance of payment. The option to not resend to anyone who already received on that channel prevents that within the strategy.

When the portfolio is updated, the key field makes the difference. Importing with the contract as the key, each contract’s record is updated instead of duplicated, and the contact history stays on the same record.

The limits of collections

Prioritizing does not authorize any approach. Brazil’s Consumer Protection Code forbids exposing a delinquent consumer to ridicule and subjecting them to embarrassment or threat during collection (art. 42). Messages must be clear and respectful, and the chosen channel cannot expose the debt to third parties.

On the data side, the LGPD provides credit protection as a legal basis for processing (art. 7, X), but the principles of purpose and necessity still apply. In DeepMiner, fields such as CPF and phone are recognized as sensitive, and their values do not enter the AI prompt.

Measure by the result on the record

Each send creates a run with total, sent messages, failures and the amount debited. Afterward, the provider’s webhook writes the status of each contact on the debtor’s record. The status shows up in the index’s record list and in each record’s event history.

Cross-reference that result with the next portfolio update. The contracts that left the portfolio after the contact and the ones that remain in it show where to adjust filters, channels and timing in the next cycle.

A dashboard for the operation

A portfolio dashboard with a KPI for total contracts, a Pareto by range and the evolution of imports, with automatic refresh, lets the operation see the portfolio without building a report. Global filters let you look at one range or one region at a time.

Want to build your sequence on a sample portfolio? Book a demo.