Most product page reviews end in a list of opinions. Someone says the gallery feels cramped, someone else defends the tabs, nothing is decided. This is a rubric instead: twelve checks, each with a written pass threshold, scored 0-2, ending in a number out of 24 that two auditors working separately should agree on within a point or two. It is for UX and CRO teams working on e-commerce product detail pages. Baymard Institute's benchmark of 155+ leading US and European sites found 62% of mobile sites perform at "mediocre or worse" on product page UX, and none scored perfect. Assume yours has findings. The question is which ones, and how bad.
How the scoring works
Each check is scored against a threshold written before you open the page. 2 means it meets that threshold on mobile and desktop, 1 partially or on one breakpoint only, 0 that it fails or is absent.
Audit the template, not one product: pick three SKUs that stress it differently - a simple item, a variant-heavy item, one partly out of stock - and take the lowest score per check. A PDP that only works for the easy SKU is not a working PDP.
Bands: 20-24, the template is sound and your conversion problem is elsewhere. 14-19, fix the zeros before running another test here. 0-13, the template is the bottleneck and testing elements on top of it will produce flat results.
Gate 1: can they understand what the product is?
Check 1 - Image set and zoom. Threshold: at least five images covering every side, one detail shot at material level, one that establishes scale, and zoom that resolves real detail. Baymard found 25% of sites lack sufficient resolution or zoom and 37% provide no in-scale image. Scale is the one teams skip, and the one that generates returns.
Check 2 - Image discoverability on mobile. Threshold: additional images are visible as thumbnails or a position indicator, without the user swiping to find out they exist. 76% of mobile sites provide no thumbnails, so the gallery is invisible to anyone who does not guess.
Check 3 - Description and spec scannability. Threshold: the description answers the five questions your support inbox receives most for that category, and specs appear as a scannable attribute list, not prose. Baymard scores 50% of sites as getting spec scannability wrong. Take the five questions from tickets or on-site search, not a workshop.
Gate 2: can they select the right variant?
Check 4 - Variant controls are exposed. Threshold: size, colour and other options render as visible buttons or swatches, not a dropdown the user must open. 57% of sites do not use buttons for size selection. A closed dropdown hides the options and which ones are unavailable.
Check 5 - Variant data synchronisation. Threshold: selecting a variant updates price, image, stock state and delivery estimate together, with no stale values on screen. 28% of sites fail to synchronise data across variations. This one usually exposes a front-end architecture problem, not a design problem.
Check 6 - Unavailable variant handling. Threshold: out-of-stock variants stay visible and disabled with a stated reason, and the page offers one next action - a restock alert, a nearest alternative, or a backorder. 68% of sites do not allow users to buy temporarily out-of-stock products. Removing the option silently is worst: the user cannot tell whether the size exists.
Gate 3: can they compute the real total?
This gate carries the most revenue and the most regulatory exposure. Across 50 aggregated studies Baymard puts average cart abandonment at 70.22%; among stated reasons, extra costs being too high leads at 40%. Most of those costs are knowable at the PDP and disclosed at checkout instead.
Check 7 - Delivery cost and date before add-to-cart. Threshold: an estimated delivery cost and arrival date, for the user's likely location, visible without interaction. 43% of sites do not show estimated shipping costs. If rates depend on address, show a labelled default-region estimate.
Check 8 - Price presentation and discount reference. Threshold: the displayed total includes all taxes; any "was/now" comparison uses a defensible reference price; and where instalments are shown, the total price, the number of instalments and the per-instalment amount appear together at readable size.
Two rules converge here. In the EU, Article 6a of Directive 98/6/EC requires an announced price reduction to state the prior price, defined as the lowest price applied in at least the 30 days beforehand. Turkey's advertising regulation applies the same 30-day-lowest logic and additionally governs instalment display. That clause is not a formality: the regulator singles out instalments precisely because they are prominent in Turkish e-commerce, so a PDP hiding the per-instalment figure behind a modal hides a number buyers decide on.
Check 9 - Return policy within one interaction. Threshold: the return window and who pays return shipping are reachable in a single tap, not via the footer. 44% of sites do not display or link a return policy on the product page, and an unsatisfactory policy is a stated abandonment reason for 13% of users.
Gate 4: can everyone actually operate it?
Check 10 - Target size. Threshold: every interactive control - swatches, quantity steppers, gallery arrows, review filters - is at least 24 by 24 CSS pixels, or spaced so a 24-pixel circle centred on it does not intersect a neighbour. That is WCAG 2.2 SC 2.5.8, Level AA. Swatches and steppers fail it most often.
Check 11 - Interaction responsiveness. Threshold: Interaction to Next Paint at the 75th percentile of field data for the PDP template is at or below 200 ms; above 500 ms is poor. Use field data, not a lab run: variant switching and gallery interaction are where PDP INP degrades, and synthetic tests do not exercise them.
Check 12 - Review evidence quality. Threshold: a rating distribution summary is present, review images can be browsed across reviews, and the page states how reviews are collected and whether they are verified. 65% of sites get the distribution wrong. A 4.4 built from 5s and 1s is a different product from a 4.4 built from 4s and 5s.
The scoring sheet
| # | Check | Pass threshold (score 2) | Gate |
|---|---|---|---|
| 1 | Image set and zoom | 5+ images, one in-scale, zoom adds detail | Understand |
| 2 | Mobile image discoverability | Thumbnails or position indicator visible | Understand |
| 3 | Description and specs | Top 5 support questions answered; specs scannable | Understand |
| 4 | Variant controls | Options visible as buttons or swatches | Select |
| 5 | Variant synchronisation | Price, image, stock, delivery update together | Select |
| 6 | Unavailable variants | Visible, disabled, reason given, one next action | Select |
| 7 | Delivery cost and date | Both estimated on PDP without interaction | Total cost |
| 8 | Price and discount reference | Tax-inclusive; 30-day-lowest reference; instalment trio together | Total cost |
| 9 | Return policy | Window and who pays, one interaction away | Total cost |
| 10 | Target size | 24x24 CSS px or equivalent spacing (WCAG 2.2 AA) | Operate |
| 11 | Responsiveness | Field INP p75 at or below 200 ms | Operate |
| 12 | Review evidence | Distribution shown, images browsable, collection disclosed | Operate |
A worked example
An illustrative composite built from the failure patterns above, not from a client audit: mid-market apparel retailer, three SKUs, lowest score per check.
Understand: 2 + 0 (blind swipe) + 1 (specs in prose) = 3/6. Select: 2 + 1 (delivery estimate stale) + 0 (sizes disappear) = 3/6. Total cost: 0 + 1 (instalment total only in a modal) + 1 (returns in footer) = 2/6. Operate: 0 (swatches at 20 px) + 1 (INP p75 340 ms) + 1 (no distribution) = 2/6.
Total: 10/24. That is not "run a headline test". It is a template fix scoped around the four zeros, in this order: delivery estimate, disappearing sizes, swatch target size, mobile gallery. Three of the four are engineering tickets, not design work, which is usually the actual finding. [INTERNAL DATA NEEDED: Switas median gate scores across audited PDPs, and conversion movement after Gate 3 fixes.]
What to do with the score
Fix every zero before you test anything on the page: an experiment run over a broken step measures the breakage, not your hypothesis. Ones are your test backlog - real, but with an arguable fix. Re-score after the fixes ship, on the same three SKUs.
Where this rubric breaks down
It is built for physical-goods retail. Digital goods, subscriptions, travel inventory and configurable B2B products have different decisive attributes, and Gate 3 does not transfer: a hotel or flight page has date-dependent pricing this rubric does not model. It also ignores acquisition - nothing here scores structured data completeness, which governs merchant listing eligibility. And it tells you what fails a written threshold, not why a particular user hesitated: it replaces the opinion round, not usability testing.
FAQ
How long does this audit take?
About 40 minutes per SKU once the thresholds are agreed, so roughly two hours for a template. Most of that goes into Gate 3, because delivery and instalment logic varies by product type.
Who should run it?
Two people score the same three SKUs independently, then reconcile only the checks where they differ by more than one point. Single-auditor scores drift toward whatever that auditor cares about.
Does a low score prove the PDP is costing us money?
No. It shows which steps fail thresholds derived from published research. The revenue link comes from your own funnel data. The rubric says where to look, not what it is worth.
How does this differ from an accessibility audit?
Only Check 10 is a formal conformance criterion. A full WCAG audit also covers keyboard operation, focus order, contrast and form labelling. Treat it as a signal that an accessibility audit is due, not as evidence you passed one.
Our PDP is rendered by a marketplace or SaaS platform. Is this still useful?
Yes, with the scope narrowed. Score everything, then split findings into what you control - images, description, specs, review collection - and what the platform controls. The platform-controlled zeros become a vendor conversation with evidence attached.
Do the Turkish price display rules apply if we only sell domestically?
Yes, regardless of where else you sell, including the 30-day-lowest reference price and the instalment display requirement. The EU rule applies only if you sell into the EU; selling into both, implement the stricter reading once.
Run it on your own catalogue
Score three SKUs from your own PDP template this week and see whether the number lands where you expected. If you would rather have it done properly, with findings mapped to engineering tickets and your funnel data attached, talk to us about a UX audit.
Sources
- Baymard Institute - Product Page UX benchmark, 155+ sites (18 March 2026)
- Baymard Institute - Product Page Usability benchmark
- Baymard Institute - Cart abandonment rate, 70.22% across 50 studies (22 September 2025)
- European Commission - Guidance on Article 6a, Directive 98/6/EC
- Gun + Partners - Turkish rules on discounted sale advertising
- W3C - Understanding WCAG 2.2 SC 2.5.8 Target Size (Minimum)
- web.dev - Interaction to Next Paint thresholds
- Google Search Central - Merchant listing structured data
