AI · Lifecycle · ML · Marketing memo
Next-best-action is a marketing problem before it is an ML problem
Machine learning can rank possibilities. Marketing still has to decide what is useful, acceptable and commercially meaningful.
A point of view
The hardest part of next-best-action systems is not prediction. It is translating prediction into a customer experience that improves a decision.
Start with the decision
Ask what decision the customer is making and what signal indicates readiness. A model built without that customer logic can become an impressive scoring system with no useful action attached to it.
Marketing defines the action space
Marketing needs to define offers, messages, suppression rules, timing, priority and customer experience constraints. ML can help rank the options. It should not be responsible for the entire customer strategy.
Incrementality is the test
The right question is not whether recommended customers convert. They may have converted anyway. The right question is whether the recommendation created incremental behaviour compared with a credible control.
Apply the thinking
Bring me the marketing problem, not just the channel.
I work backwards from the commercial constraint, customer behaviour and evidence, then decide what intervention makes sense.
Talk about your problem