Onboarding metrics are usually about reducing friction: time to first login, completion of a setup checklist, percentage of fields filled. They are easy to measure and easy to move. They are also weakly correlated with retention, because completing setup is not the same as understanding why your product matters.
This is how teams end up celebrating a 90% checklist completion rate while week-four retention stays flat. The checklist measured obedience, not comprehension. Users did what the product asked, never felt the point, and left quietly. Optimising friction without optimising for the value moment is polishing the hallway to a room nobody enters.
The metric that matters
Time-to-aha: the time from sign-up to the moment the user feels the core value of the product, the moment that makes them say "oh, this is useful." That moment is different for every product, and it is rarely setup completion. For a workflow product it might be the first time work moves through it. For an analytics tool, the first chart that surprises them.
The reference points worth knowing: the famous activation milestones (friends added in a social product, messages sent in a team chat, documents created in a collaboration tool) were all discovered the same way, by correlating first-session behaviour with long-term retention, and none of them were "completed setup." Your equivalent exists. It is usually one action, it is countable, and once named it becomes the single number the whole onboarding can be accountable to.
How to find your aha
Watch your most engaged users. What did they do in their first session that the churned users did not? That action is your aha. Make it the goal of onboarding, not the setup checklist.
The practical method, even at small scale: take your twenty most retained accounts and twenty that churned inside a month, and diff their first sessions. Look for the action with the biggest gap between the two groups, then sanity-check it in three user conversations ("when did this click for you?"). Beware the two classic traps: correlation theatre (power users do everything more, so pick the earliest divergent action, not the biggest) and proxy drift (defining aha as something easy to force, like "viewed the dashboard," rather than something felt, like "saw their own data in the dashboard"). If forcing the action does not move retention, you named the wrong moment; go back to the diff.
What to redesign
Strip onboarding of anything not on the path to aha. Move setup steps to after the aha, not before. Pre-fill aggressively. The user should reach the aha within minutes, sometimes seconds, with as little ceremony as possible.
The redesign has a mechanical recipe. List every screen between sign-up and the aha action and delete or defer anything that does not directly serve it: email verification can wait, the invite-your-team step can wait, the personalisation survey can almost always wait. Use sample data, sensible defaults, and imports so the first meaningful moment happens with the user's own context wherever possible, and with realistic demo context where not. Every step you defer is a percentage of users who survive to the moment that retains them; teams that run this exercise honestly usually cut their step count in half and watch activation follow.
Instrument it like a first-class metric
Report the median, not the mean (one stuck user should not hide a fast funnel), track the percentage of new sign-ups reaching aha within their first session, and watch the number weekly next to retention. When time-to-aha drops and retention does not move, you have the wrong aha; when both move together, you have found the one metric onboarding should be accountable to, and every future onboarding argument becomes an experiment against it.