Push Notifications: Who to Send To and Who NOT to Bother
A delivery app sends a notification with a €5 coupon every Friday to its 3M users. The CRM team celebrates that 18% of those who receive it order that day. But the CFO notices that half of those users already ordered every Friday without a coupon, and that uninstalls rise after each mass send.
Where the money comes from
What to do now
- 1Always keep 10% of users in a control group.
- 2Uplift model for ordering and for uninstalling.
- 3Send only where the net value is positive.
For your committee
One page with the decision, the impact, the risks, the steps and the KPI to track.
The print dialog opens: choose "Save as PDF".
Watch out: more volume is not always more profit
Every extra order won with a coupon has a low margin, and at peak hours delivery also costs more. Look at the margin of the incremental order net of coupon and logistics.
KPI for the committee
Incremental margin per euro of coupon.
Detail for your data team
Proposed technique: Uplift Modeling
Why: The question is not who will order, but who will order THANKS to the coupon. Uplift Modeling uses the control group to estimate the individual causal effect and avoid giving coupons to people who were going to order anyway or bothering those who will end up uninstalling.
Cost of getting it wrong: Coupons to Sure Things are wasted margin; notifying Sleeping Dogs destroys a free and valuable communication channel.
Data needed: 6-month history with 10% of users in a random control group (no notification).; Behaviour: order frequency, usual day and time, average ticket, coupon use.; Interaction with notifications: opens, opt-outs, subsequent uninstalls.