Chocolate Factory in the Rhythm of the Season | Module 9 of 9 · beyond the usual

The Direct Customer:
Whom the message moves, not who is at risk

Direct sales are small, but they are the only channel in which the manufacturer sees the buyer. The common rule "whoever threatens to leave gets an offer" is often ineffective. An effect model targets the customers for whom the message changes something.

Model calculation - fictitious sample company

This module is part of a nine-part model calculation on a fictitious mid-sized chocolate manufacturer (€95 million net revenue, 140 items, 4 lines). All euro amounts are derived from the documented assumptions of the sample company. They are not results of real customers and not a performance promise. Which values are achievable on your data is shown only by a pilot.

01 The problem - risk is not effect

Whoever leaves anyway leaves with a voucher too; whoever stays anyway does not need one

The sample company earns €2,850,000 in direct sales, 3 % of net revenue: online shop, corporate gifts, resellers without a head office. Small in revenue, large in data: order rhythm, basket, seasonal reference per customer.

The usual rule sends offers to the customers with the highest churn risk. Research shows: that is often the wrong target group. What matters is for whom the message changes behaviour, and that can only be measured with a randomly held-out comparison group.

Targeting by effect, measured against a comparison group

Without a held-out comparison group there is no effect, only a guess.

Order history
Campaigns with comparison group
Effect model
Target group by effect
Additional revenue

02 The model - two models, one difference

Purchase probability with message minus purchase probability without message, per customer; the target group is the top deciles of this difference

▸ Example values - illustration of the method, not an analysis carried out
Mean effect top 30 %: 5.8 percentage points
Mean effect random selection: 2.5 percentage points (example values)
Effect of a message per decile of customers - sorted by risk against sorted by effect (example values)
↳ The last decile is negative

Part of the customer base reacts to a message by turning away: a voucher nobody needs reminds customers that they can leave. The effect model shows this decile and leaves it alone. A risk model would write to it.

03 Business impact - small, clean, measurable

The lever: additional revenue in direct sales at the direct-sales contribution margin

€57,000
Additional contribution margin / year
+5 %
Additional revenue (assumption)
€142,500
Additional revenue / year
Model calculation · How the amount arises - derived in the open

Every assumption comes from the sample company and is stored centrally. With your figures the input changes, not the method.

PositionValue
Base: Revenue in direct sales / year€2,850,000
Lever: Additional revenue through targeting by effect (assumption)+5 % = €142,500
Contribution margin in direct sales40 %
Result: savings / year€57,000

Assumptions of a sample company - in a real project your data replaces these values.

↳ Why the smallest channel stands here

Direct sales are the only place where effect can be measured cleanly because the manufacturer can form the comparison group itself. What is learned here about seasonal reference, basket and message later travels into the conversations with retailers. The amount is small and stays small; the method is the gain.

04 Next steps in your factory

From the campaign to the comparison group

① Introduce the comparison group

From the next campaign a random part of the customer base is held out. Without this step there is no effect model.

② Measure two campaigns

After two campaigns with a comparison group the first effect model emerges. You see the effect per decile against the previous rule.

③ Target group per season

Before every seasonal campaign a list of the customers with the highest expected effect - and a list of those to be left alone.

All 9 modules: AI in the chocolate factory