MLCompoundingHigh complexityLevel 3 Advanced decisioning
Offer Sensitivity
A machine-learning model that classifies customers by how they respond to incentives: sure things, persuadables and those who react badly to contact.
Stops discount budget going to people who buy anyway, and aims incentives where they change behaviour.
Who it's for
Discount-heavy businesses that want to stop paying people who would have bought anyway and aim incentives where they change behaviour.
The engagement
What you get
- Offer-sensitivity ML model (persuadable / sure-thing / do-not-disturb)
- Offer-sensitivity attribute written to Braze
- Incentive-targeting recipes tied to holdout results
What Fuse does
- Use holdout / experiment results to label incentive response
- Train the offer-sensitivity model (sure-thing / persuadable / do-not-disturb)
- Write the sensitivity attribute into Braze
- Design incentive-targeting recipes tied to the classification
- Refresh as new experiments run
What we need from you
Data & access
- Offer and redemption history
- Holdout / experiment results
Your responsibilities
- Provide offer and redemption history
- Run the holdout experiments (via the Incrementality service) that train the model
- Provide Braze workspace access
Outcomes and proof
- Discount budget redirected away from sure-things
- Incentives concentrated on genuinely persuadable customers
- Margin protected while incremental volume holds
Assumptions
- Holdout experiment data is available to label response
- Offer and redemption events are tracked
Out of scope
- The experiment framework itself, which is the Incrementality service and a prerequisite
- Funding the incentives