All articles

The second-workflow problem: why activation dies after the first win

Signup is instrumented to death. The moment that decides retention happens later, when a user tries the real thing your product is for — and nobody is watching.

Maxence Rabrais — August 22, 2026 — 7 min read

Every onboarding review I have sat through follows the same script. Someone shows the signup funnel, points at the drop between "created account" and "activated," and the room agrees to fix the first five minutes. Shorter form. Better empty state. A product tour. Then the numbers barely move, because the user who leaves was rarely lost in the first five minutes. They were lost in week two, trying to do the actual thing they came for — and nobody was watching that moment.

I want to name that moment, show you how to find it in your own data, and argue that it is the single highest-leverage place to put help inside a product.

Two workflows, one funnel

Every product has a first workflow and a second workflow, and they are not the same kind of thing.

The first workflow is setup. Create a workspace, invite a teammate, connect a source, import a file. It is linear, it is the same for everyone, and because it is the same for everyone it is easy to tour, easy to instrument, and easy to optimize. This is where most onboarding effort goes, and honestly, most products are fine here.

The second workflow is the job. Build the first real report. Ship the first campaign. Run the first reconciliation. It is the thing the user would describe if you asked why they signed up. It is multi-step, it branches on their data, it depends on choices they made during setup, and it differs between a two-person startup and a 400-seat enterprise.

The funnel treats both as one number. "Activated" is usually defined by the first workflow — because that is the one that emits clean events — and so the dashboard turns green at exactly the moment the real risk begins.

Why the second workflow is where users leave

It is not that the second workflow is harder, though it usually is. It is that it is unsupported in a specific way.

The tour is gone. Product tours fire once, on first visit, on a known path. The second workflow happens on day four, on a page the tour never touched, starting from a state the tour author did not anticipate.

The docs are generic. Documentation explains the feature. It cannot explain the feature for this account, with this data, on this plan, after the user skipped step three of setup. The more your product adapts to the customer, the less a general doc can say.

Nobody on your side knows it is happening. A user stuck in setup sometimes writes in. A user stuck in the second workflow usually does not: they assume it is their fault, they plan to come back to it, and they do not. There is no ticket, no session replay anyone looks at, no alert. The account goes quiet and churns at renewal with a clean support history.

The second workflow is where the product stops being a demo and starts being work. That is also the moment most products stop helping.

Finding it in your data

You almost certainly already have the events; you just have not drawn the line in the right place. Here is the exercise I would run this week.

1. Write down, in one sentence, the job a new customer signed up to do. Not a feature — a job. "Send a campaign to a real list." "Close the month."
2. Map that job to the smallest sequence of events that proves it happened. Three to six events, ending in something that exists in the world: a sent email, a closed period, a published page.
3. Measure the share of users who complete that sequence within 14 days of finishing setup. Call it second-workflow completion.
4. Put it next to your existing activation number.

The gap between those two numbers is your real onboarding problem, and for most products it is the largest number in the funnel.

1st

workflow is setup — linear, toured, instrumented

2nd

workflow is the job — branching, untoured, where users leave

14 days

a reasonable window for second-workflow completion

What the number usually looks like

Every team I have asked to run this has found the same shape: activation in the 40–60% range, second-workflow completion somewhere between 10 and 25%. Half of the users your dashboard calls activated never do the thing they came to do. That is not a documentation problem or a UX polish problem. It is a help-at-the-right-moment problem.

Why the obvious fixes underperform

More tours. Teams try to tour the second workflow and discover why nobody does: it branches. A tour that works for a blank account breaks for an account with imported data; a tour for the Growth plan points at a button the Launch plan does not have. You end up maintaining a tree of tours that rot every time the UI changes.

Better docs. Docs are the right answer for reference and the wrong answer for a stuck user. The stuck user has to leave the product, find the right page, translate general instructions into their specific screen, and come back — all while holding their half-finished intent in their head. Most do not.

A chat widget. Closer, because the help is at least in the product. But a bot that has read your docs answers in prose, and prose still has to be translated into clicks. It also cannot see the user's screen, so it cannot know that the button it is describing is disabled on their plan.

What actually moves it

The customers who complete the second workflow at high rates are the ones with a named human Customer Success Manager. That is not a coincidence, and it is not because the CSM knows the product better than the docs. It is because the CSM does three things the docs cannot:

• Sees the user's actual situation — their screen, their data, their plan — and gives the specific next step, not the general one.
• Walks them through it, live, instead of describing it. Pointing at the button beats naming the button.
• Notices silence. When an account stalls in the second workflow, the CSM reaches out before the user has decided the product is not for them.

The economics of a human CSM only work above a certain contract size. The reason I find this problem so interesting is that all three behaviors are now things software can do: read the rendered interface, move a cursor to the element, watch usage and speak first. That is the whole thesis behind Barkan — a Customer Success Manager for every account, not just the big ones — and the second workflow is precisely where it earns its place.

A practical plan for the next 30 days

1. This week: define the second workflow and measure completion. Expect an uncomfortable number.
2. Week two: read ten sessions of users who started the second workflow and did not finish. Write down the exact step where each one stalled. It will cluster.
3. Week three: put help at the top two stall points — in the product, on that screen, specific to the user's state. If you do not have a guidance layer, a contextual prompt that opens a chat with a human is better than nothing.
4. Week four: re-measure. Second-workflow completion is the number; everything else is a proxy.

The first workflow gets the user into the building. The second workflow is whether they stay. Instrument it, staff it, and stop congratulating yourself on the funnel that ends at the door.