Signing up a restaurant owner, alone and online
This is the journey that lets a merchant sign up alone, online, in a single sitting. It has six steps, and the last one you fill in is the one nobody understood.
A six-step online sign-up
Keep scrolling: the journey moves with you
A restaurant, a bakery, a grocery store accept meal vouchers, and Edenred reimburses them. But first they have to sign up, which until then went through a sales rep or a paper form.
The project was to let them do it alone, online, in a single sitting. It all holds together, except the last step: every payment terminal carries an identifier, the MID, without which the customer’s card is declined. And no merchant knows that word.
The journey has six steps, some of them spread over several screens, twelve fields and four documents to dig up: a SIRET number, a Kbis company registration, bank details, a card receipt. That is a lot for someone working alone, and it is what fed the first decision.
I was the product designer on this journey: I prepared and ran the user tests, then designed the screens and all their states (UX/UI). I own a bias in how we recruited: participants were owner-managers of bakeries, restaurants and grocery shops already signed up with Edenred. The first sessions challenged the whole journey, the later ones a single page. The work stretched over about a year, alongside other projects.
- 6test sessions
- 2countries, France and Poland
- 1 yearof work, Nov. 2023 → 2024
Trimming the journey, and taking the MID off the critical path
Merchants go through six steps, on their own, dealing with administrative topics. In those conditions, every unnecessary screen becomes a chance to drop out.
The first test was rather encouraging: participants found the journey quick and easy, and the design felt playful to them. Two things snagged, though. The interstitial screens were misleading: one participant waited several seconds in front of an illustration, thinking the page was still loading. And the extra information page, nobody understood it.
From the tests to the decisions
Two paths were open to us. Keep the interstitial screens, because they’re beautiful and they set a rhythm. Or remove them, and ask every remaining screen to carry more.
I proposed to take the MID off the critical path. You can create your portal access without it and fill it in later. That costs something: we shipped an incomplete journey. But getting the contract signed came first and the MID could wait. The rest followed. We compressed: the interstitials merged into the screens they announced, the labels tightened, we asked for less.
Not one screen is left that merely announces the next one. Account creation, for instance, went from two pages to a single screen, with three checkboxes merged into one, and the password rules are shown before typing rather than after the error.
The contract page is the one that moved the most. I went back to it three times, and with each test the vocabulary gets lighter: the fees stop being technical lines, the benefits come first, and the commitment ends up stated in one sentence.
The three versions of the contract page
First version: two fee blocks in technical vocabulary.
An ID nobody knows how to find: the MID
One field, one acronym, and not a single merchant able to say what it stands for. No participant knew the word "MID". One participant went as far as searching Google, trying four increasingly long phrasings, without ever finding the answer. She carried on anyway, but she would have ended up calling customer service.
"If you hadn’t told me it was my terminal identifier, I would never have guessed."A participant, November 2023 test
One participant even left the journey to fetch a receipt from her bakery, then came back. Yet the MID is printed on the card receipt, the very object they handle a hundred times a day. All that was left was to send them there, instead of writing one more explanation.
Two versions had been competing since the first test. Version A: an explicit tutorial, then classic form fields. Version B: a skeuomorphic bet, where the receipt itself became the input interface. Version A won clearly, for reasons of execution more than of principle.
Two ways to enter the payment terminal identifier (the MID)
Version A won. “It’s crowded, it gets lost in the content,” said one participant looking at version B. Skeuomorphism came apart in execution: the fields drowned in the receipt, and the help sat below the fold.
I proposed to keep the best of both versions: version A’s explicit pedagogy, and version B’s familiar object. First a three-beat primer carousel you can’t skip: we say what we are about to do, we ask you to grab a card receipt from your terminal, then we show where to read the identifiers on it. Only then does the field appear, with help that never leaves the screen. So the carousel doesn’t open with a definition but with a gesture: rather than explaining what a MID is, we send the merchant to an object they already have in their hands. The order of the help screens was settled by the test too: the receipt first, the terminal second. The three beats are visible at the top of this page, in the mobile journey.
I went back to this page three times before it held. From one version to the next, the wording of the help barely moves; its place does.
Three places for the same help
The version that came out of the test, and the starting point for everything after it: you read first, you type after.
Always visible, never consulted: participants read it as decoration, not as help.
A three-beat carousel you can’t skip, then the help docks on the left for the rest of the journey.
- 1We say what we are about to do. One button, no skip.
- 2Grab a card receipt from your terminal.
- 3Here is where the two identifiers are printed on it.
- 4The field finally appears, the help never leaves the screen, and each terminal added is listed below.
The help was there from the start. It took three versions before anyone recognized it as help.
"In the end, I wouldn’t even have needed the terminal, it’s even simpler since we have the receipts, which everyone has."A merchant, once shown the carousel
What remains is a bare field, help that stays on screen, and an exit at every dead end. Each terminal added is then listed and can be removed. The help is also localized: a Polish receipt doesn’t look like a French one, so the content had to be rewritten for each country rather than translated.
"In our line of work, especially during peak hours, everything needs to be coordinated. […] The waiting time is not a problem for us, what we want is reliability."A restaurant owner, November 2023 interview
Move it, don’t rewrite it
For a long time I thought the MID tutorial needed better writing. It started working the day I moved it.
We compressed the rest of the journey so that one single moment could afford a carousel, permanent help and country-specific content. That is where I understood what lightening is for.
The test picked neither version A nor version B. It gave the reasons, and we had to decide with them.
Some things I never solved. A terminal can carry two MIDs, a classic one and a contactless one, and the design still doesn’t answer that case. I also dropped the idea of announcing the MID on the landing page, for lack of a phrasing that worked in every country.
The outcome figures for this project belong to Edenred, so I don’t publish them here. What gets measured on this journey is the share of sign-ups completed without a call to support, and the drop-off on the MID step.