Setup, aha and habit: three moments to design for

Onboarding is easier to design when you split it into three moments.

• Setup moment: what must happen before any value is possible. For the Mangrove app: an address inside the delivery area, an account, a way to pay. Every setup step costs some users, so keep only what the first value needs.
• Aha moment: the first time the user gets the core value with their own eyes. For the Mangrove app it is the first order arriving inside 30 minutes, complete, with nothing cut. Customers often say some version of "30 minutes, really?"
• Habit moment: the repeated behaviour that predicts someone will stay. For a weekly grocery habit, that might be a second order within 14 days of the first.

To manage onboarding you need one number that says a new user has made it. That is an activation metric: an early event that people who stay are far more likely to reach than people who leave. Slack's founder Stewart Butterfield described theirs in a 2015 interview with First Round Review. From watching which companies stuck with Slack, they decided that any team that had exchanged 2,000 messages had "really tried it", and 93% of those teams were still using Slack.

Find a candidate the same way. Take users who joined three to six months ago, split them into those still active and those gone, and compare what each group did in its first days: how many orders, how fast the first delivery came, which features they touched. Look for an event with a big gap between the two groups that most active users reached early.

Then be careful. An activation metric found this way shows a link, not proof of cause. People who place two orders in 14 days may be people who needed groceries more anyway. Before you push everyone toward the event, test it: give a random half of new users a nudge toward it and check whether their retention, not only the event, goes up (Experiments). Also check that the event fits the product's rhythm. "Seven orders in seven days" would look great on a dashboard and make no sense for groceries.

Time to value and the path to the first order

Time to value is how long it takes a new user to reach the aha moment. The shorter it is, the more people get there, because every step and every day of waiting gives them a chance to leave. Measure the median and the slowest tenth of users, and the drop-off at each step on the way (Funnels, activation and adoption covers step conversion in detail).

The diagram below shows one month of new installs of the Mangrove app in one established city. Read it from top to bottom and ask, at each step, why people stopped.

• 31% of installs never got past the location step. Some lived outside the delivery area, which is fine. Others typed an address the app could not match and gave up.
• The app asked for sign-up before showing a single product. 19% of those in the delivery area left at that screen. Duolingo faced the same choice. Gina Gotthilf, who led its growth, told First Round Review in 2017 that moving the sign-up screen back a few steps, so people could try a lesson first, led to about a 20% increase in daily active users.
• Only 70% of sign-ups built a cart. The first screen was a grid of 30 categories, with nothing to say what usually sells in that area or what could arrive in 30 minutes.
• The aha moment was fragile. Of 2,900 first orders, about 400 arrived late or with an item cut for being out of stock. Customers whose first order was on time and complete placed a second order within 14 days about twice as often as those whose first order went wrong.

That last point matters most. Onboarding does not end at checkout. For a delivery product, the operations behind the first order are part of the first-time experience. A PM who only polishes the sign-up screens and ignores the first delivery has optimised the wrong half.

Remove, defer, default: designing the path

Go through every step between install and the aha moment and put it in one of four groups.

• Remove what the first value does not need. A birthday, a profile photo, a survey about food preferences: none of these gets groceries to the door.
• Defer what is needed later. Sign-up can wait until checkout; the second address until the second order; notification permission until there is an order to track; a saved card until the customer chooses to save one.
• Default what you can guess well. Detect the location and confirm it with one tap. Open on "Popular near you in 30 minutes" instead of 30 categories. Let a first order be paid by cash on delivery, which about 35% of Mangrove orders already use, so nobody has to trust the app with a card before it has delivered once.
• Keep what protects the user or the business: a phone number so the rider can call, age checks where the law needs them, a clear delivery fee before payment.

Two more tools do a lot of work.

Empty states. A new user's screens are empty: no order history, no favourites, an empty cart. An empty state is a chance to guide. Kate Kaplan of the Nielsen Norman Group (2021) gives three guidelines: say clearly what is going on (loading, an error or nothing yet), use the space to teach, and give a direct path to the key task. An empty cart that says "Your cart is empty" does none of these. "Popular in your area, at your door in about 25 minutes", with one-tap adds, does all three.

Tips in context, not tours. Nielsen Norman Group's 2020 review of mobile onboarding advises skipping onboarding screens where you can, noting research in which swipe-through tutorials did not improve task performance. Brief tips at the moment a feature becomes useful work better: the first time a customer's item is out of stock, show how substitutes work, then never again.

Finally, set honest expectations before the first order. Show the real delivery time for that address before sign-up. A first order promised in 30 minutes that takes 50 does more damage than a first order promised in 50 that takes 45, because the aha moment is measured against the promise.

Onboarding a business is different from onboarding a person

When a corner shop joins the Partner app, the "user" is several people and the path includes steps nobody can do on a phone.

• Several people, several jobs. The owner decides to join and cares about prices and credit. An assistant may place the orders. Someone else pays the invoices. Each needs their own first value, and the owner's "aha" (credit for slow-selling stock, prices in one place) is different from the assistant's (an order placed in two minutes).
• Offline steps. Verifying the shop, checking a trade licence, agreeing delivery days, and often a sales rep's visit. These are part of onboarding even though no screen shows them, and they are usually where the days go.
• Money and trust. Credit terms, a payment method and the first invoice. A business's first value is not real until goods arrive and the bill is right.
• Activation is a business event. For a partner shop: the first order delivered and paid, and a second order placed in the app, not by phone, within 14 days.

Map these as a service path, with each person's steps and the waiting between them, and measure elapsed days, not screens. In the diagram below, a shop that signed up waited about nine days for its first delivery, and most of those days were spent waiting for people: a rep's visit, a document check, a credit decision. No screen redesign would have fixed that. The fixes were moving the verification into the rep's visit, letting the first order go out on cash or wallet payment while credit was reviewed in parallel, and offering a starter list.

The same thinking applies to internal tools. A new store employee's "onboarding" into the Store app is their first shifts. With staff turnover, a chain of 320 stores is onboarding new people every week. Design for the first shift: a training mode with practice deliveries, one task at a time (receiving before stock counts), a named champion in the store to ask, and an activation metric such as "first goods received note (GRN) completed without help in the first week". Designing for frontline work and Change management and adoption go further.