Churn scores are cheap. Knowing where to spend retention is not.

Any competitive retail energy business can build a churn model. Feed it usage, billing, payment history, and a few interaction signals, and it will hand back a probability for every customer. The trouble is that a churn score, on its own, quietly points retention spend in the wrong direction — because the customers most likely to leave are not the same as the customers most worth keeping.

Churn risk × lifetime value, not churn risk alone

A high-churn, low-margin customer subsidized by everyone else is not a retention win; letting them go might improve the book. A high-value customer with a modest churn signal may deserve more attention than a near-certain leaver worth very little. The decision that matters isn’t “who will churn?” — it’s “where does a dollar of retention spend protect the most recurring revenue?” That requires two models sitting together:

• Churn likelihood, grounded in the same 360° view and segmentation used across the customer base — so the signal is consistent, not a one-off score.
• Lifetime value, tied to profitability by tariff, rate, and usage pattern — so “worth keeping” is measured in margin, not headcount.

Close the loop or the model decays

The part most churn programs skip is the scorecard. If you can’t tell which retention mechanism actually moved which segment — and correlate the intervention back to retained recurring revenue — you’re flying blind on the only question that compounds: is retention spend paying for itself? Modeling churn is table stakes. Scorecarding the retention response is what turns a model into a program.

Plot churn risk against lifetime value and the retention budget nearly re-allocates itself: defend the high-value customers with real churn signals, let the unprofitable near-certain leavers go, and stop spending equally on both. A churn score tells you who might leave. Pairing it with value — and a scorecard that closes the loop on what retention spend actually recovered — tells you which of them is worth a dollar to keep. That is the only question the budget is really asking.