Short answer
Chronic activation failure is treated as a UX problem and is usually a decision problem: the company never chose the single first outcome the product must deliver, for which user, in which first session. Until that is decided, every onboarding experiment optimises the path to an undefined destination.
Signups are healthy. Week-one retention is not. The team has shipped four onboarding iterations this year: a product tour, a checklist, better empty states, a welcome sequence. Each moved the number a little. None moved it enough.
And the honest answer to “what should a new user accomplish in their first session” differs depending on who you ask. That disagreement is not a detail. It is the reason the four iterations underperformed, and it will defeat the fifth one too.
The three undecided questions
Teams iterate around these rather than answering them, and the reason is structural rather than lazy: answering any of the three requires saying no at the strategy level, and nobody owns that. So the work moves to the layer where progress is possible without permission, which is the interface, and the interface is not where the problem lives.
Who exactly is the first user?
The admin who bought, or the operator who works? Their first-session needs are close to opposite: one wants configuration, permissions and proof this was a good purchase; the other wants to do their actual job faster today. An onboarding flow serving both serves neither, and most flows are quietly trying.
What is the one outcome that proves value?
Not “explores the features”. The specific moment a user would describe to a colleague afterwards. Most teams have never written this down, which means the funnel they are optimising has no defined end state and every metric improvement is measuring motion rather than arrival.
What are we willing to remove?
Activation usually improves by subtraction: fewer choices, one obvious path, features hidden rather than added. Every removal has an internal defender, which is why this is a decision problem rather than a design one. Designers can propose the deletion; only authority can execute it.
“Ask five people in the company to finish the sentence “a new account is successful when they blank within blank.” I have never seen a company with an activation problem get five matching answers, and I have never seen one without an activation problem fail to.”
The test that takes ten minutes
Ask five people separately: the founder, a product manager, an engineer, someone in customer success, and someone in sales. Give each the same sentence to complete, and collect the answers before anyone compares notes. The result is unambiguous in a way that very few product diagnostics are, which is why this is worth doing before commissioning any further onboarding work.
- Five matching answers: you have an execution problem, and the funnel work will pay off. Go and do it.
- Five different answers: you have found the real bug, and it sits upstream of every tooltip anyone could ship.
- Answers that match but describe a login rather than an outcome: the destination is defined too shallowly, which behaves exactly like having no destination.
What good looks like in 90 days
The sequence matters more than the components. Decide the promise, get evidence about where it breaks, clear one path to it including things you remove, and only then redefine the metric. Teams that redefine the metric first end up optimising a number that still does not correspond to a customer experiencing anything, which is how a company ships four iterations and moves nothing.
- One written activation promise per segment, in plain language, agreed by the five people from the test.
- Evidence from roughly 10–15 focused interviews with recent customers about where the promise breaks in practice.
- One path ruthlessly cleared to that first outcome, including features hidden rather than added.
- The metric redefined around the promise rather than around logins, and only then handed to the growth tooling.
If the interviews reveal that the promise itself is unclear across the whole customer base rather than just in onboarding, the problem is larger than activation and our piece on revenue without product-market fit is the more useful diagnostic.
Fifth onboarding iteration and still flat?
Thirty minutes, no pitch deck. You will leave with the activation promise written down.
Book a Product Strategy SessionFrequently asked questions
Our activation differs wildly by segment. Is that normal?
Normal and highly diagnostic. It usually means one segment is who the product is genuinely for and the others are who marketing happens to attract. That mismatch is a strategy decision hiding inside a funnel chart, and no amount of onboarding work resolves it, because the flow is being asked to compensate for the wrong people arriving. Fix the targeting or fix the product, but recognise which one you are choosing.
Is this specific to product-led growth?
No. Sales-led products have identical physics on a longer timeline: month-three usage predicts renewal, and month-three usage is largely determined in week one. The difference is only that a sales-led company can carry a bad activation experience for two quarters on relationship strength, which delays the diagnosis and raises the eventual cost.
Should we add an onboarding specialist to the team?
Only after the promise is defined, otherwise you are hiring someone to optimise a route to an unnamed place, and they will leave frustrated within a year. If the seat above them that defines the promise is empty, our guide to capacity versus direction covers which hire actually unblocks this.
How many interviews does this need?
Fewer than a full discovery programme, because the question is narrow. Roughly 10–15 focused conversations with recent customers will usually surface the dominant themes about where a specific promise breaks. Speak to people who activated and people who did not; the contrast between those two groups is where the answer lives, and interviewing only the happy ones is the most common way to run this and learn nothing.
Choose the promise first. The funnel work becomes obvious afterwards, and the same team that shipped four inconclusive iterations will usually get the fifth one right in half the time.