Lookalike Modeling: Acquiring Twins of the Best Buyers
A digital acquisition agency has the base of a brand's 5,000 best customers (highest LTV) and needs to find, within an external base of 2 million users, which profiles are most similar so it can target them with advertising.
Where the money comes from
What to do now
- 1Define the 5,000 best customers by 24-month margin, not by revenue.
- 2Build the lookalike and compare it with the current segmentation.
- 3A/B campaign: lookalike audience versus standard audience.
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
Looking for twins of the customers who bill the most rather than those who leave the most margin: you may clone people who buy a lot on promotion and then return. Define "best customer" by margin and LTV.
KPI for the committee
LTV / acquisition cost ratio by audience.
Detail for your data team
Proposed technique: K-Nearest Neighbours (KNN)
Why: The very concept of 'Lookalike' (twin audience) is to calculate direct closeness or similarity in a multidimensional space, exactly what KNN does.
Cost of getting it wrong: Targeting users far from the ideal profile drives up the cost per click and sinks campaign profitability.
Data needed: Advertiser CRM: Interest profile, average ticket and digital habits of their VIP customers.; DMP / External Audiences: Behavioural and interest variables of potential users.