Measure the value that survives the sale.
Propensity to buy estimates who is likely to purchase. The commercial test asks whether the recommendation and intervention made a worthwhile difference. We follow the result through fulfilment, cancellation and retention.
Define the outcome before choosing the measure.
A model can identify likely buyers without creating an additional sale. We assess prediction separately from incremental contribution. Conversion, onboarding and retention each need a defined population, comparison and review period.
Connected customer history, interactions and business knowledge inform the recommendation, timing and channel. We examine the action taken and what followed. Suitability and permission remain separate requirements.
We agree the smallest worthwhile improvement and its cost. All realised sales remain in the comparison; advice failures are reported separately as a condition for continuing the service.
Follow the result beyond the initial response.
| Measure | What it establishes |
|---|---|
| Eligible-customer conversion | Completed purchases divided by eligible customers over the agreed period. Keep the audience and qualification rules stable. |
| Retention | The share of a defined customer group still active at the agreed review point, alongside changes in price, usage and contribution. |
| Retained contribution | Revenue less the relevant product, fulfilment, discount, refund and cancellation costs. Match the costs to the same population and period. |
| Incremental contribution | The additional contribution attributable to the intervention relative to a credible comparison, after the additional delivery costs. |
| Advice and fulfilment quality | Whether the recommendation was appropriate, expert advice was used correctly and the promise was delivered. Include failed onboarding and avoidable post-sale contacts. |
| Customer response | Complaints, opt-outs and other evidence of unwanted or unsuitable contact, with the affected population and channel shown. |
Distinguish a sale associated with contact from value created by it.
Keep a suitable comparison.
Where feasible, eligible customers from the same population are assigned to the intervention or usual service. Random assignment can reduce bias. We retain unsuccessful contacts and refusals in their groups and record other campaigns, so selecting likely buyers is not mistaken for an intervention effect.
Check whether the test can answer the question.
Available volume, baseline behaviour and the size of the worthwhile change determine how informative the comparison can be. Small samples can produce large-looking differences by chance. We state uncertainty and keep inconclusive findings distinct from evidence of an effect.
Allow time for the outcome to mature.
We review conversion first, then contribution after relevant cancellations, refunds and fulfilment. For retention, we select a period that reflects the buying or renewal cycle. Early indicators remain distinct from the final commercial assessment.
More orders can leave less value behind.
Suppose a test group places 80 orders and 20 later cancel. An equally sized comparison group places 70 orders and five cancel. The first group has more initial orders, but only 60 remain against 65 in the comparison group.
That arithmetic is not an ellio result or proof of a treatment effect. It shows why the review must continue after conversion. Order value, margin, service costs and uncertainty still need to be assessed before judging the intervention.
Include the expertise behind the conversation.
Suitable work runs digital first through the selected channel, including digital voice. A digital agent may complete the conversation alone, query an expert while retaining it, or transfer with full context and continuing intelligence assistance. We account for the contribution and cost of each route.
The assessment includes digital delivery, expert input, specialist time, implementation and fees. We separate one-off investment from ongoing cost and remove overlap between activities. Outcomes feed reviewed improvements to knowledge, models and recommendations.
Our commercial structure combines defined service consumption with outcome fees where additional value can be verified. We agree measurement and customer safeguards for each engagement. A before-and-after observation informs the next decision without making the whole change attributable to ellio.
Scale the evidence, then the service.
| Decision | Evidence needed |
|---|---|
| Extend | A worthwhile contribution effect, acceptable uncertainty, sound advice and reliable fulfilment across the tested journey. |
| Refine | A promising customer response with a specific weakness in timing, offer, delivery or measurement that can be addressed. |
| Stop | Customer treatment fails the agreed standard, the economics do not justify the work or the available evidence cannot support the proposed scale. |
Which growth result would change a decision?
Bring a conversion, onboarding or retention opportunity. We agree the service scope, economic baseline and measures for additional value.