Customer analytics · Live dashboard

Retention dashboard: who should we contact first?

Churn risk for 7,043 customers, from my churn project. Set what the retention offer costs and how often it works, filter the customers, and the dashboard works out who is worth contacting and what the campaign should earn.

Customers–

Expected to leave–

Revenue at risk–per month

Worth an offer–

Expected profit–

Who's worth an offerEach dot is a customer. Above the line, the offer pays for itself.

Worth an offer Not worth it

Why they're at riskMain reason the model flags each customer worth an offer.

Revenue at risk by contractExpected monthly revenue lost to churn (risk × bill).

Contact first

Top 25 by expected profit of the offer.

Customer	Risk	Bill	Expected profit	Main reason	Second reason

Risks are out-of-fold predictions from a logistic regression (5-fold cross-validation), so no customer is scored by a model that saw them. "Expected profit" is the model's own estimate. Scored against what customers actually did, the same targeting at the default settings earned about $92,000. Data: IBM's Telco Customer Churn sample (a fictional company). In production this would score active customers every month. Code ↗