Behavioral signals vs. NPS which actually predicts churn in your B2B?

Quick take, then I want yours.                                                                                                                                                                                                                                                 

NPS asks "would you recommend us?" a hypothetical, answered by whoever happens to be in the survey

window, scored against a benchmark nobody actually validated for your segment. It's lagging by       

definition. By the time someone scores you a 6, they've already half-checked-out.                                                                                                                                                                                                                                        Behavioral signals logins, admin-seat activity, feature depth, billing health, ticket patterns are what people actually do (or stop doing). They lead NPS by 30–90 days. A drop in admin-seat activity

combined with a billing dispute and a ticket spike is a far stronger churn signal than any survey answer.                                                                                                                                                                                    

But it gets messy:                                                                                                                                                                                                                                                                                     

- Behavioral alone misses the "we got acquired and the new CTO hates your category" exits — those don't show up in usage data until the contract ends.                                                    

- NPS sometimes catches sentiment shifts in champions before usage drops when champions even answer.

- Most teams I've seen do neither well. They track NPS quarterly as a board metric and call it retention insight.                                                                                                                                                                             

Genuine questions for CS / RevOps folks here:                     

1. What's your actual predictive lead time on NPS vs. behavioral in your B2B?                                                              

2. Has anyone combined them into a single risk score — or do you keep them as separate views?

3. If you had to kill one tomorrow, which one goes?                                                                                                                                                     

I'm building ChurnBase on the behavioral-first thesis with NPS as a confirming signal, not the lead. Tell me where I'm wrong.