Onboarding

Back when software was sold once and done, a person taught you how to use it. Rollout training was a line item in the contract, and after signing, someone drove to the customer's office and walked through the screens for two days. The cost made sense because a single contract ran into tens of thousands of dollars and the money was already collected.

That stopped working once signing up came down to typing an email address. People now create an account with no contract and no payment, and the question of whether the product is worth twenty dollars a month is still open. You cannot attach two days of training to three thousand people who haven't decided yet. Their arrival isn't scheduled either. A training calendar built around contract dates has nothing to offer someone who signs up at two in the morning. So whoever presses the signup button ends up alone in front of an empty screen.

An empty screen is harsher than it sounds. To the people who built the product, what to press first is so obvious that it doesn't register as something requiring explanation. To someone arriving for the first time, the eleven buttons on the screen all look the same size. They know what they came to do and not how to start doing it here, and that state doesn't last. Most close the tab within three minutes, and almost none of them write in to say what stopped them, so the team hears silence rather than a complaint.

At Pagedeck the gap showed up as the distance between two numbers. Three thousand new signups a month, and nine hundred people who created at least one document in their first week. Mia opened a handful of session recordings from the other twenty-one hundred and they all had the same shape. The screen after signup showed a document list, the list was empty, and under it sat one line of gray text reading "No documents yet." The cursor circled the screen once and the tab closed. Nothing was missing from the feature set, and none of those people had run into an error. They left because nobody pointed at the first step.

Onboarding is the work of making the product do that pointing. Under activation, the group above it, sit the things that happen to a user: the aha moment where someone realizes the product is useful to them, and time to value, which measures how long reaching that took. Onboarding is the odd one out. The rest are events on the user's side; onboarding is what the team lays down in advance so those events can happen. That's why a bad activation rate usually turns into work on onboarding.

The terms under this heading fall into two groups. Modes split on whether a person is involved, and mechanisms are the in-product devices that push the first step along. Mode gets decided first, from who the customer is and what they pay; mechanisms are the choice of device inside that mode, and the order doesn't reverse. Assign a dedicated person to a twenty-dollar-a-month customer and no arrangement of screens fixes the arithmetic afterward.

Modes split into two.

1. Self-serve onboarding — first setup gets finished through product screens and automated email, with nobody attached. When revenue per user is small, this is the only option available. An hour of a specialist's time spent on a user paying twenty dollars a month eats a large share of what that account will produce in a year. In exchange, screens built well once serve three thousand people and thirty thousand at the same cost.
2. Assisted onboarding — a specialist works alongside the customer until setup is done. On a fifty-thousand-dollar annual contract, creating a hundred accounts at once, migrating material out of the previous tool, and designing permissions department by department does not finish through on-screen prompts. It usually runs as four stages — kickoff, configuration, training, a check-in a month later — across three to six weeks. The real bottleneck is rarely the product. It's the materials inside the customer's own organization: the person who can export the old data, the person who knows the department structure, and the person who can clear security review are three different people, and schedules that slip from three weeks to six usually slip because those three calendars don't line up.

Mechanisms are the devices used inside the screens, and there are three.

1. Product tour — callouts layered over the screen, pointing at what to press in order. Step count decides the outcome. A six-step tour gets closed at the second callout; a three-step tour keeps noticeably more people to the end. So the real work is less about what to put in the tour than what to take out. Leaving one first action and saving the rest until after that action is finished does better on both completion and actual use.
2. Onboarding checklist — the first setup items laid out as a list with a visible count of what's done. Three of five boxes filled makes people want to fill the remaining two, and the device runs on that. Some products check the first box the moment an account is created, for exactly that reason. What goes on the list decides whether the device is worth anything, though. Pad it with items like uploading a profile photo, which have nothing to do with what the product is for, and completion rates climb while drifting loose from activation.
3. Empty state onboarding — turning a screen with no data yet into a prompt. Instead of a gray line reading "No documents yet," a large button for creating the first document sits there, with three samples underneath: meeting notes, a project brief, something similar. Every user passes through this screen, which is why changes here show results faster than anywhere else. It is also the cheapest of the three to build, since the screen already exists and only its contents change.

Whichever mode is chosen, the sequence for building it is much the same.

1. Settle on one aha moment. For Pagedeck it's written as a single action: inviting a colleague and editing the same document together.
2. Strip the path from signup to that action down to the steps that genuinely must happen. Profile photos and notification preferences usually come out here.
3. Attach one mechanism per remaining step. Empty state prompts on the first screen, a checklist when steps span several sessions, a tour for features people can't locate on screen.
4. Measure how many people drop out at each step and fix the leakiest one first, then measure again before touching anything else.

Onboarding doesn't happen only inside the product. The email that goes out the day after signup, the nudge sent only to people who have done nothing by day three, the message that shows the next feature to someone who created a first document — all of it belongs to the same activity. Screens can only hold people who are currently present, and calling back someone who closed the tab has to happen outside them. This is also where the two groups meet in practice. A mode decides who is allowed to receive a human follow-up, the mechanisms decide what the product says while nobody is watching, and the messages sent between sessions carry whichever of the two the account qualifies for.

"Day-one drop-off is 60 percent. New features move to next quarter and we rebuild onboarding first."

The losses come from getting lost on the first screen rather than from a missing feature, so that gets fixed first.

"We cut the tour from six steps to three and completion went from 22 percent to 61."

Sharing a result that came from removing guidance rather than adding more of it.

"Onboarding can't fix this one. The settings screen itself has to go."

When a step is hard to clear even with good guidance, the answer is removing the step, not explaining it better.

At Pagedeck this work doesn't finish in one pair of hands. Mia pulls drop-off by step, and that settles where the largest hole is. Once it's confirmed that half the users leave at the empty document list, Noa redraws that screen: gray helper text out, one button and three samples in.

Theo added one more piece. Every new account gets a sample meeting-notes document created automatically. A list that is completely empty and a list with one document in it read as entirely different screens to someone seeing them for the first time. Three people did three different jobs, but they fixed one thing: knowing what to do within thirty seconds of signing up.

After those three changes, first-week document creation went from nine hundred to the mid fifteen hundreds. No new features shipped. The same product served the same people and the number moved, which is how onboarding became its own line in the next quarter's plan rather than something appended after the feature list.

Reading onboarding as the popup that appears right after signup is the most common mistake. That popup really is built under the name onboarding, so the impression is earned. But onboarding covers the entire stretch from signup to first value, and that stretch holds the email that arrives the next day, the empty screen, and the samples placed in advance, along with the popup. Polishing the popup alone rarely moves anything, because the place people leave is usually the screen behind it.

Believing every feature has to be introduced is the other one. A lot of work went into those features, the urge to show all of them is real, and nobody enjoys hearing "I didn't know it could do that" six months later. But a person arriving for the first time came to solve one problem, not to memorize a feature list. Introduce twelve things in order and the close button gets pressed around the third, and nothing lands at all. The job in this stretch is letting someone finish the one thing they came for; the other eleven come out one at a time after they return. How many features were introduced is easy to count and whether the person finished their own task is hard to count, which is part of why this mistake survives review after review.