Getting more people from skin quiz to first order.
Dew is a fictional prescription skincare subscription I created to practise the work growth teams do every week: find where a funnel leaks, turn evidence into testable hypotheses, and design build-ready A/B variants inside a design system.
Role
Product designer (solo, self-initiated)
Timeline
November 2026
Scope
Landing page, skin quiz, plan, guest checkout, order confirmation
Tools
Figma, Figma prototyping
↓ Skip to the three experiments
01 · In brief
Three leaks, three tests.
Problem
Plenty of people start an online skin consultation. Far fewer finish it and place a first order.
My role
Solo product designer: research, funnel analysis, hypotheses, UI, UX writing and a clickable prototype.
Approach
Tear down four real journeys, check them against published checkout research, map the funnel, rank eight hypotheses, design three tests.
Output
Three test-ready A/B variants, a clickable five-screen flow, a brand and component system, a test plan and a walkthrough. A concept, so no live results are claimed.
Test A · Price up front
Show £24 a month on the results page. Targets the 40% of shoppers who abandon over costs they didn't expect.
Test B · Honest progress
A progress bar and a real time estimate in the quiz. Targets drop-off halfway through the questions.
Test C · Pay as a guest
Pay first, set a password after. Targets the 18% who leave when forced to create an account.
02 · Challenge
The consultation is the front door, and every question costs orders.
In D2C skincare, the quiz does two jobs at once. It has to collect enough to recommend the right products, and it has to build enough trust that someone pays at the end. Every extra question, a price that appears too late, or a surprise account wall at checkout gives people a reason to leave.
I set myself the brief a growth team would get: increase quiz-to-first-order conversion without hiding anything a customer needs to trust the brand.
How might we
How might we keep people moving from quiz to first order, without hiding anything they need to say yes?
03 · Research
Four journeys, one funnel map.
I went through four public UK prescription skincare and D2C subscription journeys end to end on mobile, from first landing to checkout, and recorded every step, question and price reveal. The pattern is consistent: a free photo consultation, a prescriber review, then a monthly plan at roughly £20 to £26 a month. So Dew is priced at £24, in line with the market, and the work focuses on when and how that price appears.
Journey teardown
Long consultations with no sense of progress. The monthly price held back until checkout. An account required before you can pay.
What the data says
40% of shoppers who abandon checkout blame unexpected extra costs. 18% leave because they are asked to create an account. 17% find checkout too long. (Baymard Institute)
Where the funnel leaks
Halfway through the questions. The results page, with no price. The account wall just before payment.
Sources: Baymard Institute, cart abandonment research (updated 2026). Villar, Callegaro and Yang, meta-analysis of 32 progress indicator experiments (2013). Public UK price comparisons, April 2026.
04 · Hypotheses
Eight hypotheses, ranked by impact and effort.
I wrote every idea as a hypothesis with a reason, so the team can argue about evidence rather than taste. Then I scored each one for expected impact and build effort, and took the best three into testing.
05 · Experiments
Three tests, designed to be built on Monday.
Test A · Show the price earlier
If we show the monthly price on the results page, more people will start checkout, because 40% of shoppers who abandon checkout blame costs they didn't expect (Baymard). Primary metric: results-to-checkout rate. Guardrail: first-order refund rate. Decision rule: ship if the uplift is 5% or more at 95% confidence.
Test B · Progress you can see
If the quiz shows progress and an honest time estimate, more people will finish it, because uncertainty about length causes drop-off. Research is mixed here: a meta-analysis of 32 experiments found progress bars can hurt completion when the early questions feel heavy, so the variant front-loads the quick questions and keeps the bar subtle. Primary metric: quiz completion. Guardrail: answer quality.
Test C · Pay first, account later
If people can pay as a guest and set a password afterwards, checkout completion will rise, because 18% of shoppers abandon when a site makes them create an account (Baymard). Primary metric: checkout completion. Guardrail: subscription activation in the first 7 days.
The words are variants too
Quiz intro
Before: "Take our skin quiz". After: "Get your plan in 3 minutes. No account needed."
Price
Before: "From £X". After: "£24 a month. Pause or cancel anytime."
Main button
Before: "Continue". After: "See my plan".
06 · Brand & packaging
Clinical enough to trust, calm enough to keep on show.
Prescription skincare often looks like a pharmacy shelf. I wanted Dew to feel calm and earthy, so the brand uses warm neutrals, a lowercase serif wordmark with a single drop, and illustrated packs in one colour per step. The prescribed formula sits at the centre with the patient's name and strengths on the carton, and each skin concern gets its own treatment.
07 · Design system
Every variant built from nine components.
Designing tests at volume only works if variants don't drift. I built every screen from a small, tokenised set: buttons, answer options, a progress bar, price blocks, product rows, form fields and badges, with tokens for colour, type and spacing. A new variant takes minutes, and engineering gets the same parts every time.
08 · Accessibility
Accessible by default, not as a final check.
Contrast
Cocoa on milk is 12.9:1. Secondary text sat at 60% opacity (3.8:1), so I raised it to 75% (5.9:1) to pass WCAG AA. Tags with white text moved to a deeper clay (5.1:1).
Touch targets
Every button, answer and field is at least 53px tall, above the 44px minimum. Whole answer rows are tappable, and the selected state uses a filled radio and a border, not colour alone.
Screen readers
Each question is a labelled radio group, progress reads as "Question 4 of 8", and every field has a visible label instead of placeholder-only text.
09 · Final design
The final design, clickable end to end.
All three tests combined into one flow, from landing page to confirmed order. I wired it as a clickable Figma prototype, the same one the test script uses, so reviewers and engineers can tap through it instead of guessing from static frames.
10 · Walkthrough
A walkthrough before testing with people.
Before recruiting anyone, I ran my usability script as a cognitive walkthrough: one task per screen, asking whether a first-time user could answer each question without help. Price visibility, quiz progress and guest checkout held up. Four issues didn't, so I fixed them before the next step, testing with five real buyers.
Price vs delivery
"£24 a month, every 8 weeks" could read as £24 every 8 weeks. The plan now says "billed monthly", and the landing page explains the 8-week bottle once.
Missing checkout fields
The guest form had no expiry, security code or delivery postcode, and a prescription has to ship. I added all three without bringing back the account wall.
Jargon and reassurance
"Why we ask" named ingredients most people won't know, so I rewrote it in plain words. The confirmation now says you get a full refund if the prescriber can't treat you.
11 · Options I rejected
Options I rejected.
Eight hypotheses went in and only three became tests. Two that lost: before-and-after photos on the landing page, because results vary and a prescription brand shouldn’t promise them, and a free first month, which was costly to build and would attract people who never meant to stay. Inside the three tests, these were the calls I made.
✕ Rejected
Holding the monthly price back until checkout
✓ Instead
£24 a month on the results page, because 40% of shoppers who abandon blame costs they didn’t expect.
✕ Rejected
A bold progress bar from the first question
✓ Instead
Quick questions first and a subtle bar, because research shows bars can hurt completion when early questions feel heavy.
✕ Rejected
An account wall before payment
✓ Instead
Pay as a guest and set a password afterwards.
✕ Rejected
A clinical, pharmacy-shelf look
✓ Instead
Calm, earthy warm neutrals, with one colour per step.
12 · Reflection
A test plan is a design deliverable.
Limits
This is a concept with no live traffic. The decision rules show how I would read real results, and I have not claimed any.
What I would do next
Test the clickable flow with five real buyers using my script. Then run Test A first, because it has the highest expected impact for the lowest build effort.
Next project