Making Cleo's Roast Shareable in the UK
Cleo's roast mode made fun of people's spending, but when the roast ended there was nothing to share. For the UK launch, I designed the Roast Receipt, a printed-style receipt that sums up your spending habits in one joke. In an A/B test, 23% of people who opened their receipt tapped share.
Overview
Cleo already had a roast mode in the US that people used every day. It had no share loop, and nobody had worked out how such an appealing feature could bring in new users. The UK launch was our chance to build that loop from scratch.
Impact
23%
of people who opened their receipt tapped share, in an A/B test with a few thousand users
Context
A UK launch aimed at a younger, chat-first audience.
We expected UK users to skew younger than our US base. In the UK, Cleo was mostly a chat product, so the roast had to show people Cleo's value quickly. The roast also needed a connected bank account, and bank connect was a known drop-off point. Earlier tests had tried to ease it by keeping the explanation short and adding social proof for referred users.
The problem
People had to get through a lot before they could share anything.
A popular feature with no way out
The roast ended in the chat. There was no asset to post and no route from a roast to a new user.
Bank connect was the wall
Getting a roast required connecting a bank. Our data showed this was where many people dropped off, and earlier fixes hadn't solved it.
Thin data, generic roasts
Many younger users didn't have a steady income, and data enrichment could take days for new users. Few people had enough income and 30 days of spending for a specific roast, so many got a more general fallback.
Finding the format
The receipt wasn't the first idea. It came from asking what you actually get at the end of spending money.
A story of your spending
Our first instinct was three or four story slides, each with something to share. But there's a fine line between a funny take on someone's habits and laying those habits out in full. Spelling out spending patterns raised compliance questions we didn't want to take on at launch, even though other companies had done it.
A single social card
Simple, and easy to post anywhere. It would also have looked like every other card in the feed. We needed something that stood out.
A printed receipt
I wanted to take a very digital experience and make it feel physical. After you've spent money, the thing you're handed at the end is a receipt. Not the email kind, the old printed kind. I pushed for this shape because I believed it was the right one for Cleo, and it's what we shipped.
What goes on it
The joke is about a habit, never an amount.
Each receipt is built around an archetype: a word plus the retailer you use most, generated from your spending data. Netflix Nepo Baby. Spotify Sommelier. Deliveroo Discipline. Amazon Prime Suspect. It's specific enough to feel personal, but it doesn't show anyone what you earn or spend. I drafted the receipt's structure and the first round of content. Our content designer then standardised it and wrote the prompt behind it, setting the guardrails for what the model could and couldn't say. Content and research helped us decide where a joke became too exact.
Making it feel printed
Designed to be posted, and to feel like it came out of a till.
I designed the receipt for Instagram Stories first, and checked that the archetype and the Cleo logo stayed the focus at every size, down to a thumbnail in someone's feed. I tried a typewriter typeface and a real printed-paper texture to make it feel like an object. In the app, the receipt prints out line by line with haptics. We ran internal dogfooding parties to see how varied the model's output was and how the receipt held up across devices. The main changes from those sessions were to the haptics and the printing speed.
Trade-offs for launch
01
"Roast me" as the only way in
- Problem
- Enrichment could take days for new users. Sending everyone straight into a roast after bank connect would mean long waits and heavy load on our systems.
- Decision
- I argued against quietly removing the roast from onboarding, because that would hide what people had signed up for. I proposed an honest holding screen that explained the wait. The launch date ruled it out, so we made typing "roast me" the only way in.
- Outcome
- Fewer people reached a roast, but people came to it in their own time and enrichment wasn't overloaded at launch. After launch, discoverability became the main thing holding the feature back.
02
Holding back the extra delight
- Problem
- The timeline limited how much of an experience we could build. We had several ideas for making the receipt a more memorable moment.
- Decision
- We scoped them down so the core receipt and share flow would ship on time.
- Outcome
- We waited for the test to win and the feature to launch before exploring them, with ideas like burning or ripping up the receipt.
Looking back
The receipt worked for the people who reached it. Getting people there was the harder problem.
Almost a quarter of people who opened their receipt went on to share it. With "roast me" as the only way in and no holding state, many people never got that far. We went minimal on the way in at launch, and I believe that held results back, so the next iteration will focus on making the feature easier to discover.