Ask five product managers at five different PLG companies what "activated" means and you'll get five different answers. Completed onboarding. Logged in at least once. Used a core feature. Invited a teammate. Created their first piece of content. All of these are defensible definitions. None of them are the same thing. And the definition you choose has real consequences for how you measure your product's health and how you respond to what you see.
We built Onboardvue around activation as a first-class concept because we believe the vagueness in how PLG teams define it is one of the root causes of preventable churn. This post is about what we think a good activation definition looks like, why most common definitions fall short, and what changes when you get it right.
The problem with activity-based activation
The most common activation definitions are activity-based. The user did something: logged in, completed the setup wizard, visited a key feature screen. These definitions are measurable, easy to instrument, and typically show high rates, which makes them feel reassuring to report.
The problem is that activity doesn't predict retention. A user can complete your onboarding flow without understanding what the product does. They can visit your core feature screen and leave without getting any value. Completing an activity means you succeeded in getting a user to a moment. It doesn't mean the product produced something useful for them at that moment.
We see this play out when teams run cohort analysis. They look at "users who activated" (by their activity-based definition) versus "users who didn't activate" and find the difference in retention rates is smaller than expected. Sometimes it's nearly negligible. That's usually a sign the activation event is measuring activity rather than value. If activation doesn't predict retention, the definition is wrong.
What a value-based activation definition looks like
A value-based activation definition names the specific moment when the product delivered something to the user that they couldn't easily have gotten elsewhere. The emphasis is on what the product produced, not what the user did.
For a data analytics product, that moment might be "user saw a report with their own data in it for the first time." For a collaboration product, it might be "user received a response from a teammate on content they created." For a workflow automation product, it might be "user ran their first automation that saved them from a manual step."
Notice the common structure: each of these definitions ends with the product having done something valuable, not just with the user having completed an action. The user action is a prerequisite. The value delivery is the activation event.
Getting to this definition requires a qualitative step. You can't derive it from analytics alone. You need to understand what users came to your product to accomplish and what the product needs to produce for them to feel like they got it. That understanding comes from interviewing users, specifically users who stayed long-term and users who churned early. The contrast between those two groups usually points to the value delivery moment clearly.
The instrumentation gap
Once you have a value-based activation definition, the instrumentation question follows. Can you actually measure when a user reaches this moment?
Sometimes the answer is straightforward. If the activation event is "user's first automated report delivered to their inbox," you have a server-side event that fires when the report is sent. You can instrument that cleanly.
Often it's more complex. If the activation event is "user understood the data," you can't instrument understanding directly. You need a behavioral proxy: a sequence of actions that strongly suggests the user processed and engaged with the information rather than just loading the screen. That might be: user viewed a report, changed at least one filter, and spent more than 90 seconds on the page. That's not a perfect measure of understanding, but it's a reasonable behavioral proxy for it.
The proxy design is where most activation definitions start to drift back toward activity metrics. The temptation is to simplify: "viewed report" is much easier to instrument than "viewed report AND filtered AND spent 90 seconds." Simpler instruments produce cleaner data but often the wrong data. Spend the extra time to get a behavioral proxy that actually matches the value moment, even if it's more complex to instrument.
Activation is not binary
One of the most useful reframes we've applied in our own product thinking is treating activation as a spectrum rather than a binary state. A user hasn't simply "activated" or "not activated." They're at some position along a continuum from "just signed up" to "getting full value from the core product loop."
This framing matters because it changes what you do with the information. A binary activated/not-activated model produces a list of at-risk users: everyone who hasn't hit the activation event. A spectrum model tells you where each user is on the journey and what they need to move forward.
In Onboardvue, we track users across multiple activation milestones rather than a single event. The first milestone is setup completion. The second is the initial value event. The third is the first return to the value event within 7 days. The fourth is consistent value event frequency over 30 days. A user at milestone two is in a very different state from a user at milestone four, and the nudges appropriate for each are completely different.
This approach also makes your activation funnel more useful for identifying bottlenecks. If you see high conversion from milestone one to two but significant drop between two and three (reached value once but didn't return), that tells you something specific about the product experience after the initial value moment. Binary activation would obscure that.
Why the definition matters for churn specifically
Churn prediction and activation definition are more tightly coupled than most teams realize. If your activation event is a weak predictor of retention, it's also going to be a weak input into any churn model you build. A churn model trained on "completed onboarding" as its activation feature will have worse predictive accuracy than one trained on "first delivered value event," because the latter is more directly causally connected to the user's decision to continue using the product.
We found in building our prediction model that switching from activity-based activation signals to value-based ones improved the precision of our churn risk scores noticeably. Not because the model architecture changed, but because the input signal was more meaningful. Garbage in, garbage out applies to churn models as much as anywhere else in data work.
The practical implication: your activation definition is not just a reporting choice. It's an input to every downstream analysis that depends on knowing whether a user "got it." If that input is imprecise, your retention analysis, your churn prediction, and your onboarding optimization will all be working from a weak foundation.
Testing whether your definition is right
The empirical test for a good activation definition is retention correlation. Take your candidate activation event and measure the 60-day retention rate for users who hit that event versus users who didn't (within the same signup cohort and user segment). If activated users retain at significantly higher rates than non-activated users, your definition is working. If the retention difference is small, your activation event is not capturing a meaningful moment.
We've seen teams run this test and find that their current activation definition explains only a modest portion of retention variance, while a revised value-based definition explains dramatically more. That's not a failure to celebrate, but it is a clear signal that the product team has been optimizing against the wrong metric, and that improving the definition is likely to generate more useful onboarding improvements than optimizing the current definition harder.
Good activation is not a state you can declare once and move on from. It's a working hypothesis that needs to be validated against retention data on a regular basis, especially as the product changes and as you acquire new user segments with different jobs to be done. The definition that worked for your first 200 users may not be right for your next 2000. Treat it like a product decision that needs occasional revision, not a one-time definition you carved in stone during your founding year.