What Signals finds on a product page. One scan, explained.
Illustrated with a fictional store, and checked against the shipped product on 11 September 2026. Signals is in Beta, so details change; the Signals page carries current behaviour.
Illustrative example. Demo store is a fictional flooring shop at demo-store.example, and every figure on this page is made up. The sections, labels, score rules and limits are exactly how Signals presents a real scan.
When people hear AI hypotheses, they picture a list of generic suggestions. This shows one scan of a product page instead: what Signals looked at, the idea it ranked first, how that idea was scored, what would prove it wrong, and what the scan could not see.
How to read the heat view
Scroll reach. The share of mobile visitors who scroll at least this far: all of them see the top, 66% reach the end of the gallery and 38% reach Add to basket.
Taps. The warmer the spot, the more taps it gets. The strongest is the third gallery thumbnail, the most tapped element on mobile; a second sits on the free sample button.
The fold. The dashed line is the end of the first mobile screen. The price and Add to basket sit about one and a half screens down.
At a glance
1. Page context
| Store | Demo store, a fictional flooring shop at demo-store.example. The .example top level domain is reserved for examples and belongs to nobody. |
|---|---|
| Page scanned | demo-store.example/products/oak-herringbone-laminate |
| Scan | Quick scan, two ranked ideas. This page shows the first. |
| Captured by | The ABTestly browser extension: a desktop, a mobile and a throttled mobile screenshot, plus the page structure. |
| Analytics | GA4 connected to the site. |
| Context given | Primary goal: Add to cart. A short business description in Brand context. No problem statement, niche, order value, monthly sessions or customer voice. |
2. Evidence examined
Every scan reads the page itself: its structure, headings, forms, calls to action, schema markup and trust signals. It then pulls whatever else the site has connected. The product lists each source and whether it fed this scan:
4 of 5 signals fed this scan.
| Source | Status | What Signals shows |
|---|---|---|
| Page capture | Used | A screenshot and the page structure, captured with the browser extension when you started the scan. |
| GA4 analytics | Used | Traffic and conversion events from your connected property. |
| Page speed | Used | Real visitor speed from Google's Chrome UX Report for this page. |
| Heatmap | Used | Clicks, scroll depth and friction from 46,300 page views on this page. |
| Customer voice | Not added | Customer voice is not added in Brand context. Add it there to inform future scans. |
The finding the first idea rests on, laid out as the product reports it:
- On mobile, the price and the Add to basket button sit below a tall image gallery and a free sample panel, about one and a half screens down the page.
- GA4 shows the site's mobile visitors convert at 2.1% against 6.4% on desktop, and mobile brings 71% of the site's sessions.
- The heatmap shows 53% of visitors, across all devices, scroll far enough to see the Add to basket button, and mobile visitors tap the gallery thumbnails far more often than desktop visitors do.
3. Hypothesis
Put the Price and Add to Basket in the First Mobile Screen
Because we found On mobile, the price and the Add to basket button sit below a tall image gallery and a free sample panel, about one and a half screens down the page.
We believe Moving the price, stock status and Add to basket button above the image gallery on mobile, so a visitor can act without scrolling.
Will result in +8 to 14% in add to cart
The +8 to 14% is the model's stated expectation for this hypothesis, not a forecast. Only the test measures it.
Moving the price, stock status and Add to basket button above the image gallery on mobile, so a visitor can act without scrolling.
- Place the price, stock status and Add to basket directly under the product title on mobile.
- Keep the gallery immediately below, starting with the room shot visitors click most.
- Leave the free sample panel below the gallery, unchanged.
As it is
Proposed variation
4. Why this hypothesis was ranked
This idea ranked #1 of 2 on the scan. Signals ranks by PXL score, then by how much evidence backs the idea, and breaks ties on effort, cheapest first. The model answers a fixed set of yes or no questions about the idea, each with a short rationale; ABTestly turns the answers into the score and overrides any answer the evidence on that scan cannot support. The questions below are grouped by what earns the point, with each answer in the product's own words; opening a group shows its rows.
4 + 4 + 1 = 9 of the 10 points this scan could award.
Page questions
4 of 5
| Question | Result | Points | The product's reason |
|---|---|---|---|
| Q1 Above the fold, in the primary viewport | Credited. | +1 | In the mobile screenshot the new placement sits inside the first screen, where today the button is about one and a half screens down. |
| Q2 Noticeable within 5 seconds | Credited. | +1 | The price and button would be the first thing a mobile visitor sees under the title. |
| Q3 Adds or removes an element | Declined. | 0 | This moves existing elements rather than adding or removing one. |
| Q4 Addresses a primary motivational variable | Credited. | +1 | It removes the scroll a visitor needs before they can act, the main effort barrier on mobile. |
| Q10 Tied to the primary programme goal | Credited. | +1 | Add to cart is the primary goal set on this scan. |
Evidence questions
4 of 4 possible on this scan
| Question | Result | Points | The product's reason |
|---|---|---|---|
| Q5 Informed by user testing or interviews | Not scorable on this scan: | 0 | ABTestly does not run user testing or interviews. |
| Q6 Informed by customer qualitative voice | Not scorable on this scan: | 0 | customer voice is not added in Brand context. |
| Q7 Informed by the first party heatmap | Credited. | +2 | The heatmap shows only 53% of visitors reach the Add to basket button, and mobile visitors tap the gallery far more than desktop visitors. |
| Q8 Informed by GA4 funnel data | Credited. | +2 | GA4 shows mobile converting at 2.1% against 6.4% on desktop across the site. |
Q5 never scores because ABTestly does not run user testing; Q6 could not score because no customer voice was added.
Traffic question
1 of 1
| Question | Result | Points | The product's reason |
|---|---|---|---|
| Q9 On a high traffic page, above the qualification threshold | Credited by the system: | +1 | GA4 shows at least 1,000 monthly sessions on this page. |
At about 78,000 mobile sessions a month on this page and a 2.1% mobile conversion rate, a test can detect a 15% relative lift in roughly four weeks.
4 weeks
Fogg B=MAP: Ability Sweller: Cognitive Load Fogg's behaviour model holds that an action happens when motivation, ability and a prompt meet; putting the button in view removes the scroll that lowers ability on mobile, and the price beside it answers the first question a visitor has.
Fogg, A Behavior Model for Persuasive Design, Persuasive Technology conference, 2009.
The second idea on the scan, Show Delivery Cost and Dispatch Date Beside the Price, scored 6/10 and reads Below threshold. The product explains it: Scored 6 of 10 possible on this scan at Full evidence, short of the 7 needed. The evidence tier qualifies; the score is what falls short.
Each evidence tier sets its own bar: the first idea needed 9 at Verified, the second 7 at Full evidence.
5. Evidence grade or score
- Readiness
- Test ready
- PXL score
- 9/10against the most this scan could award
- Evidence tier
- VerifiedMeasured
Test ready. Scored 9 of 10 possible on this scan at Verified, clearing the 9 needed.
This scan could award at most 10 points: 5 for the page questions every scan can answer, 1 because GA4 shows at least 1,000 monthly sessions on this page, 2 because GA4 contributed and 2 because the heatmap contributed alongside it. Customer voice was not added, so its 2 points are not in the maximum.
Test ready needs two things at once: a second, independent signal that agrees with the page finding, and a score that clears what its evidence tier requires. An idea backed by page structure alone can reach the backlog but never Test ready.
Test ready means the evidence clears the bar for its tier. It is not a prediction that the change will win, which only the test can show.
6. What could falsify it
The adversarial pass argues against this idea. Rigour, not doubt.
Strongest failure mode Flooring is bought on looks. Pushing the gallery down may cut the time visitors spend judging the finish, so add to basket rises while sample orders or returns get worse.
Most likely confounder Much of the mobile traffic may come from paid social with lower purchase intent, which would hold mobile add to basket down whatever the layout.
Alternative hypothesis Mobile visitors are not blocked by scrolling; they order a free sample first and buy later, so the real lever is a clearer path from sample to purchase.
The single discriminating test Run the reordered layout against control on mobile with add to cart as the primary goal and sample orders as a secondary goal; if add to cart rises while sample orders hold, the scroll was the blocker.
Brand risk Showing the price before the product imagery may read as discount led for a premium range.
This argument comes from a second, adversarial pass over the ideas. It is attached when the pass returns one; if that pass cannot complete, the idea still reaches you without it, so a missing argument means missing, not survived. It never changes an idea's score, readiness or rank. Here it names the reading that would sink the idea: visitors who order a free sample first and buy later.
7. Metric and guardrails to validate
- Primary
- Add to cart
- Secondary
- Overall conversion rate
- Guardrail
- Bounce rate and page load time
Primary goal: add to cart from this product page on mobile. Supporting signals: sample orders and checkout starts, with bounce rate and page load time as guardrails.
The primary metric follows the goal set on the scan. ABTestly applies the same secondary metric and guardrail to every idea, so treat those two as a floor and add guardrails specific to your page before launch. Building an idea from the backlog makes an empty draft experiment with a control and one variant; the variation code and the goals are yours to write.
8. What Signals does not know yet
- What customers say. No customer voice was added in Brand context, so Q6 could not score.
- What visitors would tell an interviewer. ABTestly does not run user testing or interviews (Q5).
- Why mobile visitors hesitate. The heatmap shows where they stop, not why; the stress test's confounder (traffic intent by channel) and alternative (ordering a free sample first) are both still open.
- The heatmap's window. Heatmaps record on pages running an experiment, so the heatmap covers only the days in the last 30 on which an experiment ran on this page.
- Page level conversion. The GA4 rates Signals weighs are site wide by device, not this page's own add to cart rate; the test's primary goal measures that directly.
- The 90 day window. A scan pulls a 90 day GA4 window and cannot tell whether the page changed inside it.
- Whether the change wins. Only the experiment answers that.
Illustrative example, checked against the shipped product on 11 September 2026. Demo store and every figure are made up; the product's labels, rules and limits are real. How Signals works, its evidence tiers and its limits: abtestly.com/signals.