Picking your activation metric from research, before you have users
The moment a practitioner glances at the screen and moves on is the moment the job was done. That glance is your activation event, and it is visible long before you have analytics.
The short answer: find the point in practitioner videos where someone confirms the task worked, express it as one event with a time window, and build onboarding to reach that event as early as possible. Treat it as a hypothesis until real retention data agrees with it.
Most early products have an activation metric by accident. Someone adds analytics the week before launch, picks an event that is easy to fire, and the team spends six months optimising a number that has no relationship to whether customers stay. The usual excuse is that there was no data to choose from. There was. It was sitting in every tutorial where a practitioner finished the job your product is meant to do.
The confirmation moment
Watch the end of any workflow video and you will see a small, consistent behaviour. The creator checks something. A total matches. A preview looks right. A message shows as delivered. They rarely remark on it, because to them it is just what finishing looks like.
That check is the practitioner’s own definition of success, and it is a better candidate for activation than anything a product team would invent. It has three useful properties: it is observable, it comes at the end of real work, and it is the same across people who otherwise work very differently.
For each video: what was checked, how long after starting, and what they did next. “What they did next” matters because it tells you whether the confirmation ends the job or hands it to someone else, and a hand-off is often a stronger activation event than the check itself.
The confirmation moment is the outcome clause of a job statement seen from a different angle. If you have already written those, the candidates are in hand — the method is in writing jobs-to-be-done statements from video research.
Sorting candidate metrics
| Candidate event | What it measures | Verdict |
|---|---|---|
| Account created | Your acquisition | Not activation; no value delivered yet |
| Onboarding checklist completed | Your funnel design | Weak; can be finished with nothing achieved |
| First item imported or connected | Setup effort invested | Useful leading step, not the outcome |
| First finished output the user would show someone | The job done once | Strong primary candidate |
| Output shared, sent or handed to another person | The job done and relied on | Strongest, but may come too late to steer by |
The pattern in the table is a trade between proof and speed. Events near the top are fast and prove little. Events near the bottom prove a great deal and may take a week to occur. The working choice is usually the earliest event that a practitioner would still describe as “it worked”.
Adding the time window
An activation event without a window is not a metric. “Sent a first report” becomes measurable only as “sent a first report within three days of signup”. The window should come from how the work is actually paced, and video is a reasonable guide to that.
If practitioners do the job daily, a new user who has not reached the event in the first session or two is unlikely to. If the job is monthly, a seven-day window will mark almost everyone as failed for reasons that have nothing to do with your product. Reading the natural cadence of the work from the footage — the same cadence that determines how long a trial should run — is covered in designing a free trial from evaluation behaviour.
Vanity activation, and how to recognise it
- ✗It can be reached by clicking through without real data
- ✗The team chose it because it was easy to instrument
- ✗It rises whenever onboarding gets shorter, regardless of retention
- ✗No customer would describe it as the product working
- ✓It requires the user's own data or own work
- ✓It matches what practitioners check for in the wild
- ✓A user could explain it to a colleague in one sentence
- ✓It is uncomfortable to look at in the first month
The last item is not a joke. An honest activation rate in an early product is usually low, and the temptation to redefine the event until the number looks respectable is strong. A flattering metric removes the one instrument that would have told you onboarding was broken.
The steps before the event
Once the event is chosen, the same videos give you the path to it. List every step a practitioner takes between starting and confirming. Each is a place a new user can stall, and the ones where creators slow down, look something up or apologise for the complexity are the places your users will stall too.
Those stall points are the short list of intermediate events worth tracking. Three is usually enough: the first real input, the hardest step, and the event itself. The reasoning for building onboarding around observed stalls rather than a feature tour is in designing onboarding from where people stall.
Removing steps is often more effective than explaining them. If the hardest step is assembling starting material, shipping that material ready-made can move the activation rate more than any tooltip — the approach in shipping starter templates from practitioner workflows.
Testing the hypothesis once users arrive
A research-derived activation metric is a starting position. The test is simple in principle: among users who reached the event in the window, do noticeably more return in the following weeks than among those who did not? If the two groups look alike, the event is not the value moment, however sensible it seemed.
Early on, the sample will be too small for confident conclusions, and it is better to say so than to read meaning into a dozen accounts. What a small sample can do is flag surprises: a user who never reached the event and stayed, or one who reached it immediately and left. Those accounts are worth a conversation, and the pre-launch equivalents of such warning signs are described in churn signals before you have customers.
Keep a record of why the metric was chosen and what evidence supported it. When the data eventually disagrees, the record shows whether the research was wrong or the product failed to deliver what the research described. Those are different problems with different fixes, and the review habit is laid out in checking whether the research held up after launch.
A worked example
Suppose the research covers freelance bookkeepers closing a month for a client. Across the corpus, the consistent final act is sending the client a short summary and seeing it marked as sent. Uploading statements happens first, categorising transactions is where creators slow down and complain, and the summary is the visible end.
The activation hypothesis writes itself: first client summary sent within the user’s first month-end. The intermediate events are a first statement uploaded and a first month fully categorised. The window follows the work, which is monthly, so a seven-day target would have been the wrong instrument. None of that required a single registered user.
What this costs
The corpus for this is the same one used for job mapping and onboarding design, so it is rarely an extra expense. As of October 2026, Hobby is $19 a month with 25 videos and 2 projects, Pro is $59 with 80 videos and 8 projects, and Studio is $199 with 250 videos, 20 projects and 3 seats, each with a 7-day free trial — see the pricing page.
Synthesise a corpus of workflow videos and ask what practitioners check to confirm the task worked. Sources and timestamps included. 7-day free trial.
Closing thought
The activation metric is a statement about what your product is for. It is worth deciding deliberately, and the people who already do the job by hand have been demonstrating the answer in public for years.
Frequently asked
What is an activation metric?
The earliest observable action that tells you a new user has received the value they came for. It sits between signing up and paying, and it is the event your onboarding should be built to reach as quickly as possible.
Can I really choose an activation metric before I have users?
You can choose a well-founded hypothesis. Practitioner video shows what people check to confirm a task worked, and that confirmation moment is the natural candidate. You then test it against real retention once you have enough users for the comparison to mean something.
Why not just use 'created an account' or 'completed onboarding'?
Because both measure your funnel rather than the user's outcome. Someone can finish every onboarding step and receive nothing of value. An activation metric should be something a user would describe as the product working, not as a form being finished.
How specific should the metric be?
Specific enough to be a single trackable event with a time window: for example, a first report sent to a client within three days of signup. A vague target like 'engaged with the product' cannot be instrumented, and what cannot be instrumented will not be improved.
What if research suggests several possible activation moments?
Pick the one that happens earliest and still represents a finished outcome. Later moments are better proof of value but too slow to steer onboarding by. Keep the others as secondary events so you can check which one actually predicts retention.
How many events should I instrument at launch?
Fewer than feels comfortable. The activation event, the two or three steps leading to it, and the return visit are enough for a first version. A long tracking plan built on guesses produces a dashboard nobody reads.
What does the research cost?
As of October 2026, Hobby is $19 a month for 25 videos and 2 projects, Pro is $59 for 80 videos and 8 projects, and Studio is $199 for 250 videos, 20 projects and 3 seats, with a 7-day free trial on every plan.