Some debtors pay simply because they saw that their bank was trying to reach them. No conversation, no promise, no minutes on anyone’s invoice, and yet the instalment arrives. We see this every month in early collections, and it is the clearest reason why counting minutes, or even counting promises, misrepresents what a collections contact is worth.
Our view is simple. In early delinquency, the act of contact carries value of its own, and the only honest way to measure it is at the level of the account, in the bank’s systems, against a comparator. Anything else rewards the wrong behaviour.
What actually happens after the phone rings#
In our early-collections projects, most customers who are a few days late are not refusing to pay. They forgot, their salary came in late, they moved money between accounts, or the reminder letter went unread. For them, a call from the bank is the reminder. Some answer, hear who is calling and say they know, they will pay today, and they have no time to talk. The conversation ends before a promise to pay can be recorded. Some do not answer at all, see the missed call, recognise what it is about and pay.
The reverse also exists. A customer makes a clear promise in a well-run conversation and then does not pay. It is less frequent than people fear, but it is real, and it is why we never treat a promise as a payment.
| What happened | What a minute-based report shows | What account-level measurement shows |
|---|---|---|
| No answer; customer pays the same day | Nothing: zero minutes | A cured account after contact |
| Short call, “I know, no time”; customer pays | A few seconds, no promise | A cured account after contact |
| Full conversation, promise to pay, payment arrives | A long, ‘productive’ call | A cured account after contact |
| Full conversation, promise to pay, no payment | A long, ‘productive’ call | An account still in delinquency |
Illustrative categories drawn from early-collections operations; not a measured distribution.
Read the table from the right and the conclusion is uncomfortable for anyone who prices or reports collections by the minute. Two of the three paying outcomes are invisible or nearly invisible in minutes, and the one non-paying outcome looks like good work.
Why minutes are blind to this#
Minute-based reporting grew up with human call centres, where the cost of a seat is time and the natural unit of output is talk time. It answers the question ‘how busy were we’. It does not answer ‘did the account cure’. In early collections, where a large share of the work is reaching people, scheduling retries, rotating numbers and sending texts, a huge amount of effort produces very few minutes, and some of the best results produce none at all.
Billing by the minute makes the blindness worse, because it rewards long conversations. A supplier paid per minute has no economic reason to design a short call, to stop when the customer has already said they will pay, or to value the attempt that never connected. We said years ago that this puts the supplier and the client on different sides of the same question, and we still think so. New agreements at Vee are no longer billed by the minute.
If we are paid by the minute, we are not playing towards the same goal as our client. – Dawid Wójcicki, Founder & CEO of Vee
Contact is not the same as conversation#
This is not an argument for dialling without purpose. The value we describe comes from contact that the customer can recognise and connect to their bank. A call placed as part of the bank’s own operation, at a sensible hour, from a number that is managed rather than burned, with a text before or after where the bank has agreed it, is a reminder the customer understands. A storm of unexplained calls is not, and it damages reach for every later attempt.
When the call is answered, the customer hears at once that they are speaking with an AI on behalf of their bank. We have always disclosed this, and Article 50 of the AI Act has required it since 2 August 2026. Identity is verified before any detail of the debt is mentioned, which is enforced by the platform. The customer who says ‘I know, I’ll pay today’ and wants to go is allowed to go. That is a good outcome, not a failed call.
Measuring results where nobody answered#
If contact has value without conversation, then the result must be measured where the result lives: in the bank’s ledger. That means account-level data coming back from the client, matched to the cases we handled, including cases where nobody answered and cases where the call lasted seconds.
In practice, the outcome data is the hardest information to obtain, because it does not sit in our systems. We work with three modes, depending on what the bank can provide.
- Periodic outcome files with the contract number, so that each payment can be matched to the cases and attempts that preceded it.
- Aggregated monthly effects per portfolio and bucket, when contract-level data cannot leave the bank.
- Automatic return after each closed allocation, the target state, where outcomes flow back as routinely as the allocation files flow in.
Input data matters as much as output. Whether a customer is new or long-standing, how they paid in the past, which hours and language they prefer: the more we know at the start, the better the correlations and the hypotheses we can test. All of this stays within the scope of processing the bank has agreed, and none of it touches credit decisions. Our job is reachability and the quality of contact, not scoring.
Why this changes how a supplier designs the work#
Once you accept that contact itself carries value, several design decisions follow, and most of them run against what a minute-paid operation would do. The first is that reach becomes the main lever. Getting through to the right person, at a time they answer, from a number they do not ignore, is worth more effort than polishing a long script. That is why so much of our work in collections sits in the contact plan, which we tune per portfolio with the bank.
The second is that the conversation should end when its job is done. If a verified customer says they have already paid, or will pay today, the call has achieved what it needed to. Keeping them on the line to recite terms they did not ask for adds seconds, irritates the customer and adds nothing to the probability of payment. A short call is often the respectful call.
The third is that the whole assignment matters more than the best-looking part of it. If an unanswered attempt can still produce a payment, then leaving a quarter of the allocation without any attempt for several days is a measurable loss, not a scheduling detail. Each day without contact moves some of those customers into the next bucket, where the bank has to start thinking about reserves.
The fourth is the least comfortable. A supplier who designs for the cure must be willing to be judged on it, with data from the client’s side, including the cases where the supplier cannot see what happened. That requires trust in both directions and a clear agreement on what counts, over what window, and how disputes are settled.
What the bank’s own teams gain#
There is a practical benefit inside the bank as well. When results are measured per account and per bucket, the collections team can see which segments respond to an early, light contact and which need a specialist conversation. Those findings feed the bank’s allocation strategy, which stays with the bank’s analysts. We do not take over that strategy; we make sure each case it sends us is reached, treated correctly and reported back with a status the bank can act on the same day.
The honest version of the claim#
There is a trap on the other side, and we want to name it. If every payment after any contact is credited to the contact, the numbers look spectacular and mean nothing. Some customers would have paid without a call. The only defensible figure is the incremental effect against a comparator: the same buckets, the same portfolios, the same period, handled by a different channel or by the existing process.
That is how we report early-collections performance, and it is why the figure above is a difference against a comparator rather than a gross recovery rate. It includes the accounts that cured after a missed call and the ones that cured after a five-second ‘I know’. It excludes the promises that were never kept. We think that is the only version of the number a risk committee should accept.
What to change in your reporting#
- Report collections contact at account level, with outcomes taken from your own systems, not from the supplier’s call logs.
- Include every handled case in the denominator, including unanswered and very short contacts.
- Compare against a comparator on the same buckets and portfolios, over the same window.
- Track kept promises separately from promises made.
- Treat minutes as a cost diagnostic, never as an output metric.
- Ask any supplier how they would be paid if half their best results had zero talk time.
That last question tends to clarify a supplier relationship quickly. If the honest answer is that the supplier would earn less for a better outcome, the commercial model is working against you. We moved away from it because we saw the problem from inside, and we are glad to show any bank how account-level measurement works on a part of its own portfolio.
Follow The First Second
Occasional issues on the mechanics and economics of AI-powered customer contact. No noise, no vendor news.
Follow via RSS