UX study for a higher conversion rate
Why shoppers leave your site.
Our Synonas walk your key journeys and find the small things that make shoppers leave.
Example studies
- Three offers, one pageUp to 25% off, 15% off a first order, 15% off by subscribing. Personas tapped them to work out which applied.
- Free shipping over £35On an £18 order, nothing says what delivery costs.
- Dogs: see policyThe dog limit and the £30 dog fee are only in House rules, two taps away.
- From £145 a nightThe total, with cleaning fee, appears only after picking dates on a small calendar.
- Excess £1,250Small grey text. What it costs to lower it only appears on the next page.
- Choose this carOpens a full-screen extras page. Its back arrow drops the search dates, so personas start again.
- Step 3 of 8Eight steps before the first check is logged. Personas tapped the counter to see what was left.
- Browser backIt leaves the setup and loses every answer, so they start again from step 1.
Where visitors leave, and why
Most leave for one of five reasons, each at its own point on the way to buying. Usually only 2 or 3 in every 100 visitors buy.
-
Arrive and browse
Not ready yet
"Just looking. I'll bookmark the site for later."
-
Read the product
Not sure
"Which size do I need? Nothing says."
-
Weigh it up
Don't trust it
"Never heard of them. Is this shop real?"
-
Basket
Costs more than expected
"£4.95 delivery? It's cheaper elsewhere."
-
Checkout
Can't do it
"It wants an account, and the code won't apply."
-
Bought
The few who make it
"Found my size, and it came with free delivery."
Tap anyone to hear why they left, or why they bought.
A look inside a report
Two excerpts from an example report: how each journey went, and the top bottlenecks with the fix to try first. A full report also has the routes personas took, the walks behind each finding and the fixes we tested. Modelled on a real UK skincare store, name removed. Counts are illustrative.
Journeys · 150 Synonas each
Top bottlenecks
The page says delivery is free over £35, but never what it costs below that
"It's £18. Is delivery another £4? I'm not typing my address to find out."
Try first: put the delivery price for orders under £35 next to the free-delivery line.
A 15%-off email popup covers the product on the first visit
"I can't see the cream. Where do I close this?"
Try first: show the offer after the first add to cart, or as a bar that doesn't cover the page.
Buy a moisturiser for sensitive skin · the pages they went through
Each ring is the next page personas reached after the homepage, out to the cart. Amber shows where they left; the outer ring shows how it ended. Hover or tap a segment to follow a route.
8 of the 58 who searched for "eczema cream" landed on an empty results page. 3 of them left the store.
How it works
You give us a web address. We do the rest on your live site, without touching your code or your customers' data.
- YouSend your site and emailThe homepage is enough.
- Personas built for your siteWho your visitors are, why they came, what device they use and how patient they are.
- Tasks and success criteria setYour key journeys, such as "find a moisturiser and reach payment", and what counts as done.
- Your live site checkedWe make sure each page opens for a fresh visitor. If a login or a blocker stops us, we email you.
- Measured walksEach persona walks on its own and says what it thinks. Every step, pause and exit is recorded.
- Analysis and findingsWe count where personas stop, group the reasons and rank what to fix first.
- 8hYou get a link to your dashboardBottlenecks, the walks behind each one and the fix to try first. Within 8 working hours.
We couldn't send your request. Please try again, or email info@synona.app with your site address.
We're on it.
- Site
- Report to
- Due by
Free checks run straight away for sites with 100 or more orders, bookings or sign-ups a month. We'll look at your site and email you within 1 working day.
Your Synonas start on your homepage shortly. If a page needs a login or blocks visitors, we'll email you before the deadline rather than send an empty report.
Worried about a particular journey, or want to add one? Email info@synona.app with your site address and we'll include it.
They aren't real people. Here's why the results still hold.
How we build each persona, what the research says about AI personas, and how we calibrate them to your product analytics.
How a persona is built
- Device
- Laptop, Chrome
- Arrived from
- Google search, second visit this week
- Goal
- Buy one jar as a gift that arrives by Friday
- Budget
- About £25, delivery included
- Patience
- Gives up after 2 dead ends
- Reading
- Skims: small print is blurred for him
- Trust
- Not a sceptic (about 1 in 8 are)
- Competitors
- Sees 2 other shops' delivery prices
- Your audienceDevice, traffic source and first-time vs returning come from your product analytics, so the personas have the same mix as your visitors. Without analytics, we use a typical mix for your kind of site.Device · arrived from
- A reason to be thereEach persona gets one goal from your key journeys, with a deadline or budget that makes it real: a gift by Friday, four nights under £600.Goal · budget
- Human limitsSome skim and miss small text, some read slowly, some give up after one dead end. About 1 in 8 checks you're a real business first. The mix is fixed before the study starts.Patience · reading · trust
- The outside worldOn monthly plans, about 1 in 3 sees your competitors' price, delivery cost and date, then decides whether to stay.Competitors
- A walk on its ownA fresh browser on your live site, with no memory of the other walks. Different AI models play different personas, so no single model's habits decide the result.Every persona
What research says
Research on AI personas points both ways. We built around both.
Where they hold up, we lean on them. Where they fall short, the method compensates.
| Verdict | What research found | What we do |
|---|---|---|
| Holds up | AI copies of 1,052 people matched their survey answers 85% as well as the people themselves did.Park et al., Stanford, 2024 | Every persona gets a full profile: goal, device, patience, budget and how they arrived. |
| Solved | AI guessed a real shopper's next click 1 time in 8.Lu et al., 2025 · 31,865 real sessions | We don't bet on one click. We count how many of 150 walks stop at each step, with a range on every count. |
| Solved | AI testers took the straight path and flagged 3 of the 10 trouble spots people hit.Zhong et al., 2025 · GPT-4, Gemini 2.5 Pro | Each persona knows who it is and why it came. In the same study, that brought AI much closer to people. |
| Solved | AI flattens groups of people towards an average.Wang et al., 2025 | We mix skimmers, slow readers and impatient visitors on purpose, and match the mix to your analytics. |
Calibrated to your product analytics
Calibrated to your product analytics.
On monthly plans, connect your analytics read-only. Your traffic sets who the personas are, and your funnel sets how they behave, before the study counts anything.
Your traffic becomes the persona mix
Each segment of your visitors gets the same share of the 150 personas.
Their journey is matched to yours
A calibration run compares each step with your funnel. Where personas give up too early or too late, we adjust their patience and run it again.
Follow one persona: Priya's walk
One of the 150 walks behind bottleneck #1 in the report above. Priya says what she thinks at every screen, and stops where a real visitor like her would.
-
0:00
Homepage
Two offer bars, then a 15%-off email popup over the page. Looks for the close button and taps "No thanks".
Up to 25% off, 15% off… Which discount is mine? I just want a moisturiser. -
0:38
Skin-concern menu
Picks "Sensitive skin" and gets 11 products.
-
1:12
Soothing Oat Moisturiser 50 ml · £18
Reads the price and "Free shipping over £35".
I'm £17 short. What does delivery cost for one jar? -
2:05
Same page, further down
Finds benefits, FAQs and over 1,200 reviews. No delivery price for orders under £35. Adds to cart to find out.
-
2:41
Cart
Sees "Add £17.00 more for free shipping", but still no price for delivery itself.
So it isn't free. I'm not typing my address to find out if it's £3 or £5.Gave up at the cart. She was one of 106 personas who left without finding the delivery cost.
A study repeats this across your main buying journeys, with personas who differ in goal, device and patience. We count where they stop, group the reasons, and rank what to fix. You send a web address; the report reaches you within 8 working hours. Get your report.
Why 150 independent agents, not one AI session you steer.
Each of the 150 personas runs as its own AI agent. Point Claude Code or Codex at your site instead, and it walks the straight path you describe, once, reading every word. Your visitors don't.
They leave the straight path
The bugs live in the detours.
Example: "Set up a new kitchen account" in a food-safety app, walked by 150 agents. Each dot is one agent, placed by how many detours it took. One steered AI session takes the straight path, patiently finishes setup, and reports no problems.
Detours a steered session skips
Agents who did it at least once, out of 150
They see the page, not the code
Your code says it. Your page doesn't show it.
A coding session mostly reads your code, where the small grey excess line is as clear as the price. Personas look at screenshots of the rendered page, and some see it the way a hurried visitor does: small text blurred, or not there at all.
One Claude Code session
<h1>Ford Fiesta or similar</h1> <p class="price">£34 / day · £102 total</p> <p class="fine">Excess £1,250 · Basic cover included</p> <button>Choose this car</button> .fine { font-size: 11px; color: #c4c4c4; }
Excess: found
150 personas look at a screenshot of the page
Careful readersRead every word · e.g. Marta, 34
Ford Fiesta or similar
£34 / day · £102 total
Excess £1,250 · Basic cover included
Manual · 5 seats · collision damage waiver included
41 of 60 noticed the excess
Ford Fiesta or similar
£34 / day · £102 total
Excess £1,250 · Basic cover included
Manual · 5 seats · collision damage waiver included
9 of 54 noticed it
HurriedNo small print · e.g. Dan, 27Ford Fiesta or similar
£34 / day · £102 total
Excess £1,250 · Basic cover included
Manual · 5 seats · collision damage waiver included
0 of 36 noticed it
The mix of careful, skimming and hurried personas is set before the study, from your analytics where we have it. A code review would mark this line as done. 100 of the 150 personas chose the car without noticing the excess, and 71 stopped at the extras page when it asked them to pay to lower it. Illustrative
They can be counted
One walk is a story. 150 are a number you can rank.
One AI session tells you what can go wrong. Many agents tell you how often, so you know which problem to fix first.
Pricing
Start with the free check. Subscribe only if it finds something worth fixing. A one-off CRO audit typically costs £900 to £8,000.
| UX checkFree | Monthly£199 /month | AgencyContact us | |
|---|---|---|---|
| Studies included | 1 check per site | 2 full studies a month | Agreed per month across your client stores |
| Extra studies | — | £79 each | Included in your agreed price |
| Buying journeys per study | 3 core journeys | 5 core + up to 3 you choose | 5 core + up to 3 you choose |
| Synonas | 50 per journey | 150 per journey, plus a mobile-only group | 150 per journey, plus a mobile-only group |
| Competitors' offers in the mix | — | Up to 5, chosen by us or named by you | Up to 5 per client store |
| Calibrated to your analytics | — | GA4, read-only. PostHog, Amplitude and Mixpanel on request | GA4, read-only. PostHog, Amplitude and Mixpanel on request |
| Bottlenecks in the report | Top 3, with the fix to try first | All, ranked, with the fix to try first | All, ranked, with the fix to try first |
| Persona walks behind each finding | Yes | Yes | Yes |
| Your fix tested before you build it | — | Up to 2 fixes per study, on a copy of your page, with fresh personas | Up to 2 fixes per study |
| Report | Web page | Web page + PDF | Unbranded PDF for your client |
| Get my free check | Start with the free check | Contact us |
A study is one store's journeys, walked by a fresh group of Synonas. The free check and the monthly plan start with the same form: your store's website and your email. You only subscribe after you've seen the free check. Agencies can try a free check on one client store first. Every report arrives within 8 working hours. Prices exclude VAT.
Questions
Who is the free check for?
Live shops, booking sites and apps with real orders, bookings or sign-ups. It runs straight away for sites with 100 or more a month. Smaller sites go on a short review list, and we email you within 1 working day.
Who are the Synonas?
They are AI models that role-play your customers: a first-time visitor on a phone, a returning buyer comparing two products, a gift buyer in a hurry. Each journey sends 150 of them (50 in the free check), and about 1 in 8 is sceptical and checks you're a real business before adding anything to the cart. Each persona browses your real pages and explains what it's looking for and why it stops.
What can't Synonas tell you?
They show where friction is likely and why. They can't tell you how much real demand there is, how your customers feel about your brand, or exactly how much revenue a fix will add. Treat each finding as a strong hypothesis, and confirm the fix with an A/B test or your own analytics.
Why do you ask whether I work at the brand or an agency?
The report describes weaknesses in a store, so it's meant for the people who own it or work on it. If your email isn't at the store's domain, we just ask which one you are. Any email address works, including Gmail.
What counts as 8 working hours?
Working hours are Monday to Friday, 9:00 to 18:00 UK time. A request sent on Friday afternoon arrives by Monday afternoon. You'll see the exact deadline as soon as you submit.
Which journeys do you study?
By default: finding a product, choosing a variant, checking delivery and returns, adding to basket and reaching checkout or booking. We add research journeys for visitors who aren't ready to buy, such as reading reviews on one topic or saving a shortlist to share. If there's a journey you're worried about, email info@synona.app with your site address and we'll include it.
What if my store is password-protected or blocks bots?
Synonas need to reach your pages like any visitor. If your store is behind a password, or bot protection (Cloudflare, for example) stops them, we'll email you before the deadline to agree how to get through, rather than send an empty report.
How do competitors come into it?
Shoppers compare. On monthly plans we find up to 5 competitors selling the same kind of product, or use the ones you name. About 1 in 3 Synonas see their price, delivery cost and delivery date during the journey, then decides whether to carry on with you or leave. The report shows how many left for a competitor and what tipped them. We only read competitors' public pages.
What data do you collect?
Only what any visitor sees on your public pages, plus the website and email you give us. We never touch your customers' data or your admin. On monthly plans you can connect your analytics read-only, and we use it only to calibrate the personas to your real visitors: GA4, with PostHog, Amplitude and Mixpanel on request.
We're an agency. Can we send several client stores?
Yes. Try a free check on one client store first: submit it with your agency email and choose the agency option. For several stores, email info@synona.app and we'll agree volume pricing. Each report goes only to you. We don’t contact the brand unless you ask us to.
See your site through your shoppers' eyes.
Free UX check, report in 8 working hours.