Before you scale the service, agree what success means.

Imagine a collections operation reports more cash recovered in June than in May. The headline looks encouraging. June also contained more accounts, larger balances and a different mix of debt. Without understanding those changes, it is difficult to know how much credit belongs to the service.

The same problem appears elsewhere. Sales can rise after a seasonal campaign. Contact volumes can fall after a channel becomes harder to use. Complaint closures can increase while customers continue to challenge the decisions. Each measure needs context before it becomes evidence of value.

Define the customer group, the service being changed, the observation period and the result that would justify expansion. Record the starting performance and the costs that belong to it. Include quality and customer safeguards alongside the financial measures.

Where practical, use a suitable comparison group to distinguish the service's effect from wider changes. Other approaches can include a staged introduction or a carefully adjusted baseline. Their strength depends on the available data and the differences between the groups. The method should match the decision being made.

Keep the original baseline visible as performance improves. A rolling view helps manage today's operation; the starting position helps explain the value of the change over time. Both are useful when their purposes are clear.

For a growth service using propensity to buy, keep likelihood and additional benefit distinct. Selecting customers already more likely to purchase can raise conversion without establishing that the intervention helped. Define a comparable eligible population before evaluating the chosen action, timing, channel and conversation. Retain the outcome of each group on the same basis.

The commercial case includes connected knowledge, predictive intelligence, integration, digital delivery, quality and ongoing service fees. Include digital voice where used, expert consultation during a digital interaction and human work after transfer. Measure the full journey. A reduction in handling time can create capacity while leaving payroll unchanged; keep that benefit separate from expenditure the operating plan can actually remove.

Avoid overlap between benefits. An account that stays current can contribute to lower arrears and lower expected losses, but those measures cannot simply be added as though they represent unrelated gains. Earlier cash receipt also differs from additional cash that would otherwise never have been collected.

Finance and operations should agree these definitions together. A business case that they interpret differently will become a commercial disagreement when results arrive.

A service fee can pay for agreed delivery, while an outcome element can recognise verified improvement where attribution is strong enough. The contract needs to specify the measure, baseline, exclusions, review period and treatment of factors outside the service's control.

The incentive also needs to support the customer outcome. Paying for more conversations should not reward avoidable repeat contact. Paying for collections alone can miss the value of keeping customers out of arrears. The right structure depends on the scope and the evidence available, which is why ellio's commercial approach starts with the value case.

An outcome fee is only as credible as its definition. If both parties cannot explain how a result becomes a payment, the arrangement needs more work before it scales.

Consider a deliberately illustrative sales test with two comparable groups of 5,000 eligible enquiries. The normal-service group produces 400 orders, of which 20 cancel. In the revised journey, connected knowledge and propensity inform the approved action, timing and channel, with digital agents supported by human expertise. This group produces 460 orders, of which 35 cancel. The headline difference is 60 orders; after cancellations, it is 45 retained orders.

If each additional retained order contributes £40 before the incremental service cost, the observed difference represents £1,800 of contribution. If the test also incurs £1,100 of incremental delivery and allocated implementation cost, the provisional net value is £700. Every number in this example is invented.

The calculation remains conditional on a credible comparison. It does not establish that the intervention caused all 45 additional retained orders, that the difference exceeds normal variation, or that future groups will behave the same way. Discounts, returns and fulfilment costs must be treated consistently in the contribution figure.

Record which conditions are fixed and which can legitimately change. A new product, a tariff change, a major service failure or a different customer mix can affect the result independently of the intervention. Agree how such events will be recorded and reviewed before they become reasons to dispute the fee.

Where a rolling baseline is used, retain the version, date and population behind each calculation. The purpose is to understand current performance without losing the original business case. A better baseline should not quietly erase earlier progress or be used to attribute external changes to the service.

Also specify the treatment of missing records, cancelled work and corrections after reporting. A result should remain traceable to its original evidence and any authorised restatement. Both parties need the same calculation, rather than separate spreadsheets that agree only when performance is favourable.

A positive calculation is one input to expansion. The operating review also needs evidence that the service can deliver reliably, that customer safeguards held and that the next population resembles the one tested. An attractive result in one straightforward journey does not establish the case for every service.

A focused evaluation should lead to one of three practical choices: expand, adjust and test again, or stop. Agree those conditions at the outset, including what would count as an unacceptable customer or operational result.

Set a review that includes the client outcome owner, operations and Finance, with technology and assurance involvement where needed. Present the estimated value, uncertainty, operational exceptions and the cost of expansion together. The decision can then be proportionate to the strength of the evidence.

If the test is inconclusive, identify the specific uncertainty that further work would resolve. More analysis is useful only when it could change the decision. That discipline keeps proof of value connected to the commercial choice it was created to support.