Blog

Thinking about retention, plainly

Written for customer-success leads and subscription operators who want to understand what the data says, and what to do about it.

Customer retentionJuly 15, 2026 · 5 minute read

Churn shows up in the data first

The signals that precede a cancellation are almost always visible weeks before the account leaves. The problem is not detection. It is attention.

Read it

Written by

Maya Kessler

Co-founder and CEO

Machine learningJune 18, 2026

Why our churn model is a logistic regression

We tried gradient boosting and a small neural network on our cohort. Both scored marginally better. We shipped the logistic regression anyway, because the model a customer-success lead cannot inspect is a model they cannot use.

Priya Natarajan · 6 minute read

Customer successMay 9, 2026

The save play: what to send when the model says 30 days

The save play is not a last-ditch email asking for another chance. It is a specific action matched to a specific signal. The match is almost always wrong in customer-success teams that are new to working a ranked list.

Maya Kessler · 6 minute read