You're spending on ads. Traffic isn't the problem. A synthetic shopper study finds exactly where your funnel is leaking and why, before you spend more on fixes that might not move the needle.
Paid acquisition is working. Conversion is below where it should be. The tools you have tell you what, not why.
Six to eight weeks. £8,000 minimum for recruitment, moderation, and analysis. The findings are excellent. The timeline and cost don't suit every situation.
Shows where people click and where they stop. Doesn't explain why they hesitated, what they were uncertain about, or why they left. Quantity without reasoning.
£3,000–£10,000 per month. Appropriate at scale. Overkill if you need a clear picture of where friction sits before you know what to test.
Fast enough to inform the next sprint. Affordable enough to repeat quarterly. Evidence-led enough to act on. That's what a synthetic shopper study delivers.
"I can't see how much shipping will cost until I've entered my address. I'm not doing that."
Heatmaps show you the what. Session replays show you the when. Neither tells you the why. That's the gap a synthetic shopper study fills.
We deploy shopper agents across your complete funnel, from homepage through checkout. Each agent is assigned a distinct persona based on your brand's buyer types. Sessions run across desktop and mobile simultaneously.
Personas are defined per study based on your actual buyer types, not a fixed template. You provide 2–3 descriptions, or we infer them from your brand positioning.
Each shopper narrates an inner monologue as they browse: what they're looking for, what they're uncertain about, what's stopping them. An AI moderator probes for reasoning at key decision points, just as a human moderator would in a think-aloud study.
All sessions are synthesised into a structured PDF research report. Findings are tiered as Primary Blockers, Conversion Drags, or Watchlist items, each with annotated screenshots, verbatim quotes, cross-validation data, and ranked recommendations.
A sample study output showing the format, evidence standard, and finding structure. Brand and site details are fictionalised for illustration.
Every finding in the report has to clear a bar before it's classified and reported.
No finding enters the report unless it is grounded in observable UI: a missing element, a broken pattern, a hidden cost, a confusing label. Opinions without screenshots are not findings.
A friction one shopper encounters is not a pattern. Primary Blockers require evidence from at least 5 sessions across at least 2 independent persona types. This is stated in every report, next to each finding.
Primary Blocker: causes abandonment or prevents task completion. Conversion Drag: creates friction but doesn't always block. Watchlist — directional signal requiring more sessions before acting.
A bargain hunter and a brand-loyal shopper evaluate the same page differently. Persona stratification surfaces friction that aggregate session data would flatten, and tells you which segment it matters to most.
Every finding is reported with its denominator: "9 of 18 shoppers". Never "50% encountered this issue". Watchlist findings are explicitly marked directional, not conclusive. Confidence is earned, not asserted.
Synthetic shopper studies identify pattern-level friction cheaply and fast. They answer one question well: where is the funnel leaking, and why? That's the right question before you know what to test.
A structured PDF research document. Not a slide deck, not a dashboard export. Length and depth vary by site complexity and findings volume.
Not a replacement for human user testing, but a complement. Faster, cheaper, suited to a different set of questions.
| UserSimulations | Traditional user testing | Heatmaps / replays | CRO retainer | |
|---|---|---|---|---|
| Time to insight | 5–7 days | 4–8 weeks | Continuous, no synthesis | 4–6 weeks first test |
| Cost | £1,500 per study | £6,000–£15,000 | £200–£800 / month | £3,000–£10,000 / month |
| Sample size | 25 sessions | 6–12 participants | Hundreds of sessions | Varies |
| Answers "why" | Yes, narrated and moderated | Yes, think-aloud | No | Depends on method |
| Evidence type | Screenshots + verbatims + cross-validation | Video + observer notes | Click + scroll data | A/B test results |
| Repeatability | Run quarterly, compare over time | Expensive to repeat | Continuous, undirected | Retainer-paced |
| Best for | Pattern-level friction diagnosis | Emotional depth, longitudinal | Volume signals only | Ongoing experimentation |
Founders and heads of ecommerce spending on paid acquisition who need to understand where conversion is leaking before they know what to test.
Your site URL and your top 2–3 customer personas, or a brief description of who buys from you. We'll infer the rest. No analytics access, no developer involvement.
No presentation call required. We send it to your inbox and you share it with your team when you're ready. Happy to talk through findings if that's useful.
No. Every finding is grounded in observable UI evidence: a missing element, a confusing label, a hidden cost. Findings are observed and documented with screenshots. A Primary Blocker requires evidence from at least 5 sessions across at least 2 independent persona types before it's classified. We report what synthetic shoppers encountered, not what we think they might encounter.
No, and we wouldn't claim it does. Human user testing is better for deep emotional nuance, longitudinal research, and post-purchase behaviour. Synthetic shopper studies answer one question well: where is the funnel leaking, and why? Think of it as the research you do before you know what to test.
Primarily DTC ecommerce brands on Shopify doing £30,000–£500,000 per month in revenue. UK and US-based, premium consumer positioning. The method works best for brands with a real product catalogue and a conversion problem they haven't been able to explain with analytics alone.
5–7 days from kickoff. Kickoff means we have your site URL, your customer personas, and any specific funnel areas you want prioritised. The report is delivered as a PDF. No meeting required, though we're happy to talk through findings if useful.
Your site URL and your top 2–3 customer personas, or a brief description of who your buyers are, and we'll infer personas from your brand positioning. No analytics access, no developer involvement, no internal data required.
Yes. The Daybreak Coffee Co. sample above is a real study output. Download it and read the evidence standard, the finding format, and the recommendations. If it looks right for what you need, get in touch.
Send us your site URL and a brief description of your brand. We'll follow up within one business day.