Stop paying people to do what they were already doing
Discounts feel like performance because they get redeemed. The quadrant asks a harder question: would this customer have purchased anyway? Four classes, and only one of them deserves an incentive.
Where it sits: The targeting half of proof. It decides who gets an incentive; the holdouts in /prove show what the incentive actually bought, and train the next version.
Most discounts buy customers you already had
The quadrant
Two questions split your base four ways: would they purchase anyway, and does contact actually move them? Who buys anyway, who is genuinely persuadable, who reacts badly to contact, who is beyond an offer.
Margin saved, not response bought
Redemption rate flatters every discount. A sure thing cashing a voucher is margin given away for behaviour you were getting for free; the model counts only the lift an offer creates.
Holdouts are the other half
This model targets; /prove verifies. Every incentive campaign holds a slice back, and those experiments become the training data that turns the heuristic into a true uplift model.
The same customers, through the offer lens
The cast you met in RFM, re-read for discount sensitivity. Tap any customer to see their full journey.
The habit is live: they were coming back with no nudge at all.
The habit is live: they were coming back with no nudge at all.
A genuine maybe: the decision is open, so a nudge can tip it.
A genuine maybe: the decision is open, so a nudge can tip it.
A genuine maybe: the decision is open, so a nudge can tip it.
That is Offer Sensitivity working on sample data. The service page has the rest: what it needs from you, how long it takes to build, what it pairs with, and how the lift gets proven.