Here's a question most CS teams can't answer cleanly.
What behavior, specifically, drives your retention rate?
Not "engagement." Not "health score." Not "relationship quality." A specific, observable action that a specific customer either took or didn't take.
If you can't name it, you can't engineer it. And if you can't engineer it, you're hoping.
Retention, expansion, and advocacy are outcomes. They're what happens downstream when customers do certain things consistently. But the things themselves — the actual behaviors — are almost never named, tracked, or designed for. We measure the outcomes and then wonder why they're hard to move.
A behavior is observable. It either happened or it didn't. "They're engaged" is not a behavior. "They logged in four times this week and responded to our last two emails within 24 hours" is a behavior. "They're a good customer" is not a behavior. "They've referred two colleagues and agreed to a case study" is a behavior. "They're at risk" is not a behavior. "They missed their last check-in, haven't opened a campaign in 30 days, and asked about their contract end date" is a behavior.
The distinction matters because you can't trigger what you can't see.
So what are the behaviors that actually drive your outcomes?
Retention lives on: consistent logins, completing onboarding steps, responding to outreach, showing up to calls, using the product in ways that produce results, asking questions that signal they're building something.
Expansion lives on: asking about features they don't have yet, hitting usage limits, referring a colleague, engaging with content about advanced use cases, saying things like "when we scale this up" instead of "if this works."
Advocacy lives on: sending an unprompted referral, agreeing to a case study, defending you in a conversation you weren't part of, introducing you to someone in their network without being asked.
None of those are feelings. None of them are scores. They're things people do.
The reason this matters operationally is that behaviors are triggerable. Once you've named the behavior you want, you can work backwards to the trigger that makes it more likely to happen. Once you've named the behavior you're worried about, you can design an early warning system that catches it before it becomes a trend.
You cannot design a trigger for "engagement." You can design a trigger for "hasn't logged in in 14 days." You cannot monitor "satisfaction." You can monitor "hasn't responded to the last two outreach attempts." You cannot engineer "loyalty." You can engineer the sequence of actions that makes loyalty the natural outcome.
This is the shift. From measuring outcomes to designing the behaviors that produce them.
Name the behaviors. Then figure out how to make them happen.
