Illustrative example
A diagnostic example, with a company that does not exist.
Illustrative example; it does not describe a real case or present results obtained. It shows how a diagnostic leads to prioritizing one action and ruling out another.
Example conclusion
A welcome contact is prioritized and a blanket discount is ruled out.
All figures in this section are hypothetical and belong to a company that does not exist. They describe a purchasing pattern, not the effect of any action.
Prioritized action: test a welcome contact
Hypothetical evidence: of 1,000 new customers in the previous semester, 200 (20%) made a second purchase within 90 days. Decision: this pattern justifies testing a welcome contact with eligible new buyers, in a defined window after their first purchase. Eligibility and the window are set before starting, and a comparable group without contact is kept. The customers of that semester, including the 800 who did not repeat, are historical evidence, not the audience of the test.
Ruled-out action: blanket discount for high-value repeat customers
Hypothetical evidence: 450 of 500 repeat buyers (90%) keep their pace, and 1,200 of their 2,000 orders (60%) already carried a discount. Decision: their value is high, but nothing indicates that another discount would change their behavior. The 50 whose pace dropped are reviewed separately.
Not observed: how many customers would buy thanks to the contact, and whether that effect would leave margin after its cost. A high value does not anticipate a high response either. That is why one action is tested, and the other is ruled out only for this cycle.
Example
Four audiences of a hypothetical portfolio.
Illustrative example · hypothetical company · not a real case and no results obtained
High current-value repeat buyers
- Evidence
- They buy often and account for much of the realized value. (Observed)
- Historical value (realized)
- High · On revenue alone they look like the most valuable audience. (illustrative level)
- Reading
- High value does not mean intervening changes anything: many would buy anyway. A blanket discount could just transfer margin.
- Relevance for the customer
- A discount may benefit the buyer without generating additional purchases; before offering one, it is worth understanding what they need and what would change with the action.
- Possible decision
- Ruled out this cycle: a blanket incentive. Attend only to those whose purchase pace drops relative to their own pattern. Valid option: do not intervene with the rest.
- Measurement
- Compare those who receive attention with a comparable group that does not, over a fixed window. Economic result: incremental contribution margin, with the discount already reflected, minus additional execution costs. If costs are missing, incremental revenue is reported, labeled as such, and profitability cannot be concluded. Counting purchases is not enough to judge an incentive that reduces margin.
- What we do not know
- We do not know whether the drop in pace is a normal pause or a sign of loss.
Customers with a single recent purchase
- Evidence
- They bought once in recent months. (Observed)
- Historical value (realized)
- Low · Little realized value so far. (illustrative level)
- Reading
- Their expected value depends on whether a second purchase happens. It is an uncertain estimate, not a fact.
- Relevance for the customer
- The contact must be useful to the customer and not just push for another purchase.
- Possible decision
- Prioritized: test a welcome contact to facilitate the second purchase, only with eligible new buyers and in a defined window after their first purchase.
- Measurement
- Group with contact versus a comparable group without contact, with random assignment when feasible. Response: second purchase within a window set in advance. Economic result: incremental margin minus the cost of the contact. If costs are missing, incremental revenue is reported, labeled as such, and profitability cannot be concluded.
- What we do not know
- We do not know what share of second purchases would happen without the contact.
Low current value, with inferred affinity signals
- Evidence
- They bought little, but their purchases resemble those of customers who later bought higher-margin products. (Inferred)
- Historical value (realized)
- Low · They seem of little value. (illustrative level)
- Reading
- Low current value does not equal low potential, but here the potential is a model hypothesis and must be validated.
- Relevance for the customer
- A suggestion is relevant only if it answers a real need; that is why it is validated before being expanded.
- Possible decision
- Small, limited test. Check against later behavior or a direct question before expanding.
- Measurement
- Test with comparison. Response: purchase of the categories of interest within a fixed window. Economic result: incremental margin minus the cost of the action. If costs are missing, incremental revenue is reported, labeled as such, and profitability cannot be concluded.
- What we do not know
- The inferred affinity may not reflect a real need. Its confidence is limited.
Long-inactive customers
- Evidence
- No purchases over a long period. (Observed)
- Historical value (realized)
- Medium · They bought quite a lot in the past. (illustrative level)
- Reading
- Historical value is real and already happened; it does not say they will return. Recovery deserves a test if there is a hypothesis that the additional value generated by the contact may exceed its cost.
- Relevance for the customer
- Insisting on someone who has already left may annoy them; the contact must offer something useful to them.
- Possible decision
- Selective test, only with those who meet agreed criteria of relevance, eligibility and cost, and with a comparison group. The rest, no intervention.
- Measurement
- Contact versus no contact in a subgroup. Economic result: incremental margin minus the cost of the action, not just the number of reactivations. If costs are missing, incremental revenue is reported, labeled as such, and profitability cannot be concluded.
- What we do not know
- We do not know whether they are still potential customers, or whether the contact would change anything.
A real diagnostic depends on the data, the business question and the agreed scope. The High/Medium/Low levels in this example do not come from real data.
Want to see how it would look with your portfolio?
A real diagnostic depends on your data, your question and the agreed scope.