Dessert before vegetables – why nobody ever fixed onboarding by cutting screens

Someone sees a style analysis of themselves for the first time and the first thing they do is take screenshots. Not of the products. Not of the app. Of the analysis itself – the colour palette, the style vibe label, the recommendations. They send it to three friends before they’ve even finished scrolling.

That reaction tells you everything you need to know about where the feature belongs. It doesn’t belong buried four levels deep in a tab called Playground. It belongs at the front door.

The feature salad

The existing onboarding was thirteen screens. Name, handle, colour theme, gender, country, date of birth, clothing size, bra size, notification permissions, seven feature showcase slides, a selfie request, and a celebration screen. The reward for completing this ordeal: an empty Saves page.

Seven features presented with equal weight means none of them feel important. It’s a menu with 300 items – it doesn’t say “we can do everything,” it says “we have a big freezer.” The user’s unconscious reads this as: they don’t know what they’re good at.

But the real damage isn’t the length. It’s the sequencing. The app asks for a selfie with no context, no payoff, no visible reason. Upload your face to a stranger’s app and trust that something will happen later. The drop-off isn’t about impatience. The psychic contract is broken: thirteen screens of effort, zero value delivered in return.

A $200 consultation disguised as a selfie request

A professional colour analysis costs $200–300. A session with a personal stylist who tells you your undertone, your best necklines, which fabrics to reach for and which to avoid – that’s an afternoon and a significant chunk of money. The app already does all of this from a single photo. It just never told anyone.

The wine cellar existed. Nobody had built the front door.

Move the Stylist from page four of the app into page one of the onboarding. The selfie request transforms from a suspicious data grab into an exciting invitation: give us one photo and we’ll give you a personal style analysis that would cost you hundreds elsewhere. Same feature. Same screen. Same pixels. Completely different psychology.

Cialdini’s reciprocity principle explains why this works: when someone receives something of genuine value, the urge to reciprocate is almost involuntary. The original flow demanded information before delivering value – a toll road. The new flow gives away the most impressive thing the app can do before asking for anything – a VIP entrance. The selfie isn’t a cost anymore. It’s a key.

The emotional trajectory matters more than the step count

The information-gathering screens – name, gender, country – are vegetables. Necessary but joyless. Front-load them and the user’s emotional arc goes: excited (download), deflating (form, form, form), too late (the good part). The user has already decided this is boring before reaching the interesting bit.

Flip the order. One essential question first – “Show me women’s or men’s” – which isn’t even a form, it’s a preference. One tap. Then straight into the Stylist: upload a photo, watch the face-scan animation, receive a personalised style breakdown. Peak emotion. After that, ask for the name. After that, the handle.

Now the vegetables come after the dessert. And they don’t taste like vegetables anymore. The user has just been told their warm undertone makes earthy tones work effortlessly for them. Giving their name now feels like completing a profile, not filling in a form. The context has changed the task entirely.

Kahneman’s peak-end rule is useful here: people judge an experience by its emotional peak and how it ends. The old flow peaked nowhere and ended with an empty page. The new flow peaks at the style reveal and ends with a populated, personalised feed. Same number of steps. Profoundly different memory of those steps.

Fifteen seconds of theatre

The analysis takes about fifteen seconds. During this time, the user watches a face-scan animation tick through six steps: scanning features, analysing colouring, mapping face shape, building colour palette, finding style vibe, curating recommendations. Every step is real, but the pacing is theatrical on purpose.

There’s a reason for this. Sutherland tells the story of a locksmith who got fewer tips as he got better at his job – the faster he opened the door, the less people valued the service. The same applies to AI: instant results feel cheap. A visible process that appears to work hard on the user’s behalf creates perceived value proportional to the wait. The fifteen-second animation isn’t a loading screen. It’s a performance of competence.

But the performance has to carry through. If the animation shows the user their own face being studied, and then the results screen switches to a blank white page with bullet points, the emotional thread snaps. Showing someone their reflection in a mirror and then turning the lights off to describe what you saw. The user’s photo stays – blurred, as a backdrop – behind every slide of the reveal. Every insight feels tethered to the person in the photo, not to an abstract profile.

The reveal is the product

Most product teams treat analysis results as information to be delivered. A title, a paragraph, some data points. That’s a receipt, not a reveal.

Five paced slides – style vibe, colour palette, personal insights, style rules, curated products – each with its own animation and moment to breathe. The first version the user scrolls past. The second version they screenshot and send to friends. The difference isn’t the content. It’s the pacing.

George Loewenstein’s information gap theory explains the pull: curiosity is the gap between what we know and what we want to know. Each slide closes one gap and opens another. “Urban Minimalist” – what colours go with that? Colour palette revealed – what should I actually wear? Style rules revealed – where do I buy this? Products appear. The sequence manufactures desire at every step.

The shareable moment matters more than most teams realise. “Urban Minimalist” on a blurred backdrop of the user’s own photo is inherently shareable. It’s an identity label, not a data point. Nobody shares a paragraph of analysis. Everyone shares a label that makes them feel seen. That screenshot going to three friends is the viral loop that no referral programme could manufacture.

Tabs aren’t navigation. They’re narrative.

The original tab bar: Saves, Try On, Discover, Polls, Playground. Five features arranged in no particular order with no story connecting them. The user lands on Saves, which is empty, because they have nothing saved. The first experience of the app is absence.

Reordered: Discover, Saves, Try On, Polls. Four tabs, each a beat in a story. “Here’s what suits you” → “Keep what catches your eye” → “See it on yourself” → “Ask your people.” Each step has a clear “why now” and a natural handoff to the next. The Discover tab is populated from day one with the Stylist’s product recommendations. No cold start. No empty state. Day one feels like day one hundred.

Same ingredients, different recipe

The total feature count didn’t change. The Stylist already existed. The Discover feed already existed. The tab bar already existed. The selfie upload already existed. No new API endpoints. No new AI models. No new screens, strictly speaking – just existing screens in a different order, with a different emotional arc.

The engineering team’s instinct when onboarding underperforms is to build new features. The actual fix is almost always sequencing. Same ingredients, different recipe, different dish. The former is expensive, slow, and uncertain. The latter is a Tuesday afternoon.

The bottleneck is never the feature set. It’s the story the features tell when experienced in sequence. Get the story right and the product feels like it was built for the user. Get it wrong and it feels like a feature salad – technically complete, psychologically empty.