One CPA Target Hides Your Best and Worst Customers

@Homeshield operates in smart home security DTC. High-consideration purchase. Long decision cycle. But if you are using one CPA target for all customer segments, the polynomial is averaging away your most valuable information.

The reflexive invariant: "Our target CPA is $X, and we optimize against it." That closure treats CPA as a single-variable polynomial. But different customer segments have radically different value profiles. A customer who buys a full home security system and subscribes to monitoring is worth 8x a customer who buys one camera and churns in month three. Optimizing to one CPA optimizes to an average that does not exist.

When I built cohort-based pricing for a DTC operation, we discovered that three distinct segments had CPA targets that should differ by 300%. We were overpaying for low-value segments and under-investing in high-value ones.

The Cohort-Based Pricing Framework:

1. Segment customers by predicted LTV at acquisition. Behavioral signals in the first session predict cohort membership with surprising accuracy.
2. Set different CPA targets by segment. Pay 3x more for predicted high-LTV cohorts. The lifetime math justifies it.
3. Different creative for different cohorts. The message that converts a full-system buyer is not the message that converts a single-camera buyer.
4. Track cohort profitability over time, not just at acquisition. Some cohorts improve. Some decay. The polynomial changes over the customer lifecycle.
5. Adjust acquisition spend in real time based on cohort performance data. We reallocate 15% of budget monthly based on cohort-level returns.

The S4 construction: a cohort-level acquisition engine that treats every segment as its own polynomial with its own solution.

I have cut CPA by 66% and lifted ROAS 67%. Cohort-based pricing was the refinement that made both possible.

Do you have one CPA target or segment-level targets? What would change if you segmented?