From a class to an incentive policy
The class syncs into Braze as an attribute and picks the policy. Three of the four classes get no incentive at all: the discipline of this model is where it refuses to spend, and the persuadables get everything that saves.
Full price, full presence
They were going to purchase anyway. Keep the relationship warm at full price and save the incentive budget for a customer it can actually move.
A light, time-boxed incentive
The only customer an offer genuinely moves. Keep it small, time-boxed and specific: enough to tip an open decision, not enough to reset their price expectations.
Leave well alone
A healthy habit that does not need a reason. Recognition, not discounts: an offer here can only give margin away or disturb a routine that was working on its own.
No incentive, efficient always-on
An offer will not bring this customer back at a price worth paying. Keep the lowest-cost always-on presence running and spend nothing extra until a real signal shows.
The one offer that earns its cost
The persuadable play is light and time-boxed: enough to tip an open decision, small enough to pay for itself in the incremental margin it creates, and short enough that it never becomes the new normal price.
Where intent is lower, the model escalates to a stronger incentive, and nowhere else. Every other class gets presence, recognition or silence, at full price.
This week only: a little thank-you
A light, time-boxed reason to purchase, sent only to the customers an offer genuinely moves.
The do-not-discount discipline
The hardest part of this model is not the offer, it is the restraint. Two classes are explicitly protected from the discount budget, whatever the redemption rate says.
Sure things
They redeem enthusiastically, which is exactly the problem: every voucher replaces a full-price purchase that was already coming. Spend they were giving you anyway is not a campaign result, so suppress them from every incentive send.
Sleeping dogs
A champion habit or a subscription that renews without a thought. A discount invites them to rethink a decision they had stopped making. Recognition, not price: the cheapest retention programme is the one that leaves well alone.
Prove the offer did anything
Redemption is not lift. Hold a slice of every incentive campaign back, count only the gap above the control, and feed the result straight back into the quadrant: this model is trained by the experiments you run there.
Raw before-and-after would claim the whole targeted bar. True attribution counts only the gap above the control, the part that would not have happened without the model. Illustrative.