The payment is due next week. The opportunity is now.

Consider an illustrative water account. The next bill is unexpectedly high. The customer has already reported a possible leak and asked for an explanation. Payment is due in seven days. The service record and billing record each hold part of the story. Read together, they suggest a useful next step: investigate the bill and contact the customer about the unresolved issue.

A reminder alone leaves the question unanswered. Earlier support creates an opportunity to resolve the cause, understand any payment difficulty and agree an appropriate next action. The account remains in the paying population. That is precisely why it deserves attention.

The most likely customer to miss a payment is not necessarily the person whose outcome an intervention can change. Some customers will pay without contact. Others need support that a reminder cannot provide. A useful decision considers the likely cause, available options, customer circumstances and the cost of acting.

This gives prevention a more demanding question than ranking accounts: which permitted action is worth taking for this customer, now? The answer might be a billing correction, a conversation about support, a clearer explanation or a review by a specialist. It can also be to wait for information rather than send another message.

Predictive intelligence draws on connected payment, account, service and conversation knowledge to identify risk. Judgement then weighs current circumstances, permitted actions and their likely usefulness. It determines what to do, when to act and which channel fits. The customer's response can change that judgement as the interaction develops.

ellio's prevention service connects that intelligence and judgement to digital-first delivery. Digital agents carry out approved actions across the appropriate channels, including digital voice. The choice reflects the task, contact preferences and customer circumstances. Human expertise supports the interaction as questions arise.

In the illustrative water case, a digital agent can query a billing expert about the unresolved explanation while retaining the customer conversation. Clarifying the bill does not itself authorise an account adjustment. Where the decision or circumstances require a specialist to take over, the agent transfers the context and continues supplying the human with relevant knowledge and guidance.

The engagement defines which records can be connected, who can amend the account, which contact preferences apply and who owns an unresolved action. Expert consultation and transfer are distinct steps, with responsibilities for each. The customer needs an accurate account of what has happened and what remains open.

The response, expert input, completed action and subsequent payment outcome enrich the shared customer knowledge. That evidence informs the next decision and the review of guidance and predictive models. A favourable outcome becomes useful learning when the service understands the circumstances behind it.

If the customer pays, the observed outcome is encouraging. It does not establish that the intervention prevented arrears. The customer might have paid anyway. A credible evaluation needs a suitable comparison and enough time to see whether payments remain sustainable.

Where appropriate, compare an added preventive treatment with normal service among similar eligible accounts. Essential support remains available to everyone. Where a controlled comparison is impractical, use the strongest available baseline and make the uncertainty clear. Changes in balances, seasonality and customer mix can affect the result.

Measure entry into arrears alongside customer outcomes and delivery cost. Avoided arrears, lower expected losses and recovered cash describe different benefits. Counting the same improvement under several headings makes the business case look better than the economics.

Start with customers whose accounts are current and who share a specific, potentially resolvable issue. A payment date mismatch and an unresolved billing dispute may both precede arrears, but they call for different information, authority and support. Combining them into one treatment makes the result harder to explain.

Write the eligibility and exclusion rules before choosing cases. Record why each account qualifies, when the signal became available and whether the intended response can still arrive in time. A useful prediction made after the payment was missed cannot establish an early-intervention capability.

The review should include people who look at risk but remain current, and people who fall behind without being identified. Review the cost and customer experience of contact among those who remain current, using comparison evidence to assess whether the intervention added value. Examine missed opportunities and limits in the evidence among those who fall behind. Both groups belong in the evaluation.

The observed difference is 40 accounts among the 2,000 offered the intervention. It is an estimate to investigate, with uncertainty. It is not evidence that every contacted customer avoided arrears because of the contact. The comparison also needs checks for allocation, starting balances, missing outcomes and other changes that could explain the difference.

Count the cost of identifying, contacting and supporting the whole treatment group. Include unsuccessful contacts and people who would have paid anyway. Establish whether the improvement lasts beyond the first payment date and whether any difficult cases were transferred elsewhere.

The next decision depends on more than the entry-rate calculation. Did the intervention resolve the underlying issue? Did it create an unnecessary burden for customers? Can the service perform the same work reliably for the next eligible group? Those answers determine whether to expand, refine or stop the treatment.

Successful prevention can change the composition of the accounts that still reach collections. If easier-to-resolve cases remain current, the remaining arrears population may be more difficult. A lower recovery percentage within that smaller population can therefore coexist with a better whole-portfolio result.

Follow the original eligible population through payment, entry into arrears, recovery and repeat difficulty. Keep the prevention and recovery measures visible, but assess their combined economic effect. This makes it possible to distinguish better customer outcomes from a change in which cases each team receives.

Choose one recurring issue that appears before payment difficulty. Reconstruct recent cases across billing, service and payment records. Ask what information arrived first, what action was possible and why it did not happen. Then define a treatment, its owner and the evidence needed to judge it.

Some debt cannot be prevented by a better service. Some apparently useful signals will prove unreliable. The discipline is to find the situations where earlier action improves a customer's position and creates measurable value, then expand from the evidence.