Skip to content
Sebastian.Gebhardt

Draper

AI Customer Decisioning for Omnichannel Retail

An omnichannel decisioning brain for the holding's customers: it decides who to contact, what to offer, when, and on which channel — consent first, with measured lift.

Live pilot · Hoka
17SIGNALSDRAPERAPPROVALHOLDOUT · MEASURED LIFTCUSTOMER
Decide → approve → dispatch → measure
Banners holding customer data
17
Live pilot (Hoka: 3 stores + cl.hoka.com)
1
Data law, native to the model
21.719

The problem

A holding with 17 retail banners knows the same customer in fragments. Each brand messages on its own, nobody measures the incremental effect of any given message, and Chile's data-protection law (Ley 21.719) demands consent handling that generic tools don't ship with.

How it works

Draper unifies the customer into a golden record, decides the next best action per person — who, which offer, when, which channel — and dispatches with human approval from its own operations console. Every decision lands in a ledger with holdouts, so incremental revenue is proven by measurement, not faith. Consent is native: Ley 21.719 lives in the data model, not as a patch.

What it does

  1. 01

    Golden record

    One unified customer record across the group's banners, instead of 17 partial versions.

  2. 02

    Decisioning and dispatch

    AI proposes the play; a person approves before any message reaches a real customer.

  3. 03

    Measured lift

    A decision log with holdouts and an outcome ledger: incrementality is demonstrated, not declared.

  4. 04

    Consent-native

    The consent and preference spine is designed on Ley 21.719 from day one.

Limits

A live pilot, not a full rollout: the data scope covers 17 banners, but the live operation today is Hoka — 3 stores and cl.hoka.com. Lift is reported against holdouts, not as attributed revenue.

Run a retail operation and recognize the problem? The advisory side of this work lives at Menlo & Oak — and you can always write to me directly.