Why collections breaks at scale for digital lenders — and what's actually working.

Most digital lenders hit the same wall at roughly the same point in their growth. The loan book is performing. The product is working. The team is executing. And then delinquency starts climbing, the collections team cannot keep up, and the NPL ratio begins moving in the wrong direction.

The instinct is to hire. Add agents. Build out the function. For a while it works. Then the book grows faster than the team can, and the problem returns — larger this time, and more expensive.

This is not a staffing problem. It is a structural one. And until you see it clearly, you will keep solving the wrong thing.

The scale wall

A collections agent working a dialler can make 60 to 80 contact attempts per day. Of those, a third to a half will result in actual conversations. Of those conversations, a fraction will produce a payment arrangement. Follow-up, broken arrangements, dispute handling — this takes time. An experienced agent, working well, can actively manage 200 to 300 accounts at any given point.

At 5,000 delinquent accounts, this is manageable. A team of 20 can touch the book. At 50,000 accounts — a number that arrives faster than most lenders expect — the same team is looking at 2,500 accounts each. That is not collections. That is triage.

The math does not lie. At 80,000 accounts, a team of 25 agents working full capacity can make first contact with roughly 3,000 accounts per month. The other 77,000 accounts age. The early-stage debt becomes late-stage debt. The recoverable becomes unrecoverable.

The unit economics

The second problem lives in the P&L, not the headcount plan.

A single contact attempt — agent time, dialler cost, telephony, infrastructure overhead — costs somewhere between £6 and £14 depending on your market and setup. Reaching a delinquent borrower typically requires eight to twelve attempts. You are now looking at £80 to £150 before a single conversation has happened.

For a £400 debt with a 30% recovery probability: expected recovery value is £120. Cost to recover: £100 or more. The margin barely exists, and that is before you account for the accounts that go to zero.

For a £100 debt — common in salary advance, BNPL, and micro-loan portfolios — the unit economics are simply negative. Chasing it costs more than recovering it. These accounts are written off not because they are unrecoverable, but because the cost of recovery makes it irrational to try.

The lenders who will win are not the ones with bigger collections teams. They are the ones who figured out how to make the unit economics work at any scale.

The tiering trap

Most lenders respond to the scale wall by tiering the book. Prioritise high-balance accounts. Focus agent time where the expected recovery value justifies the cost. Let the small balances sit or move them to a third-party agency.

This is rational. It is also a trap.

In digital lending — BNPL, embedded finance, salary advance, micro-loans — the vast majority of the book is under £500. Tiering away the small balances means ignoring most of what you are owed. The loss rate on your small-balance book becomes structural. It is priced in. Accepted.

The better lenders know this is not a solution. It is a managed retreat.

What is actually working

The lenders seeing genuine improvement in recovery rates share a few things in common.

First contact speed matters enormously. The probability of recovery drops significantly with each day that passes after first missed payment. The lenders recovering at the highest rates are making contact within 24 to 48 hours — not 7 to 14 days. They are treating the first contact window as critical infrastructure, not a process to be managed.

Negotiation beats notification. An automated SMS saying “you have missed a payment” produces a fraction of the outcomes of a conversation that understands the borrower's situation and offers a realistic path forward. The lenders recovering well are the ones whose collections function can actually negotiate — propose a partial payment, restructure an arrangement, acknowledge difficulty without losing the thread of repayment.

Coverage at scale is the multiplier. A lender who can contact 100% of their delinquent book within 48 hours — every account, not just the high-balance ones — sees materially better outcomes than one who contacts 50% of accounts within 14 days. The improvement compounds across the entire portfolio.

The implication

The structural insight is this: collections is a conversation problem, not a headcount problem. You do not need more agents. You need more conversations that can actually negotiate — at any scale, on any account, from the first day of delinquency.

The technology to do this exists. The lenders who are deploying it are building a significant structural advantage. The ones who are not are managing the same problem, at growing cost, with diminishing returns.

The wall is real. The way through it is not more people.