TITAN-CORE 2.4

Customer record

Customer 48213 · consumer

9 orders · €281.00 · €31.22 average basket · 47-day cadence · last order 67 days ago

Scores served on 18/06/2026 — unfold a score for its reason

Recommended action

Demonstration: these settings are not saved and train no model. They do recompose the segmentation plan at the bottom of this section, though — you read there the exact split your setup would produce across your 45,800 customers.

Over what span the model sums a customer’s expected value. The horizon touches neither churn risk, nor repurchase dates, nor the 90-day potential — each has its own deadline. It changes the scale of the value, and the uncertainty that comes with it.

What puts one customer in the same group as another. A cohort always reads over time: it is what lets you see whether customers recruited in May are worth more than January’s, and to know it before the six months are in.

The four splits shipped with the model. Each is kept as is or removed — a removed axis is no longer pushed to your CRM or your ad platforms, and the model keeps producing its five scores without it.

The split you write yourself, by crossing the model’s outputs. Name it, pick its criteria, the headcount computes as you go. That is the segment that will land in your CRM under the name you gave it.

The headcounts of the four native axes are your base’s, as of 18/06/2026. Your own segments’ headcounts are estimated by crossing status and potential — an inactive customer with very high potential does exist, there are 30 of them — then rounded to the nearest ten; in production they are counted customer by customer. Moving from one horizon to the next holds in one rule: value(24 months) = value(12 months) × (1 + r) and value(36 months) = value(12 months) × (1 + r + r²), where r is the tier’s observed retention — 0.74 · 0.58 · 0.39 · 0.17 from the top tier down.