Foundations
Six store symptoms that signal a conversion problem worth investigating
Assess six non-ranked ecommerce conversion demand signals using reproducible store evidence, current UK controls and explicit measurement limits.
| Signal | Evidence to compare | First response |
|---|---|---|
| Repeatable access failure | Task, assistive technology and journey step | Research remediation |
| Optional charge without a recorded selection | Configured extra and customer action record | Pause the affected checkout version |
| Tag and choice mismatch | Stated purpose, recorded choice and activation | Remove or reconfigure, then review |
| Rate movement across denominators | Raw count, denominator and implementation version | Define the measure before comparison |
| Refund without an order match | Order, cancellation, return and refund records | Assign an owner and cut-off |
| Missing checkout information | Product details, total price, delivery, payment action and confirmation | Review content and applicable rights |
Treat each row as an investigation prompt, not a demand estimate or sales forecast. Evidence was reviewed on 6 September 2026 for a merchant with a verified establishment in Great Britain; UK and global sources retain their stated geography rather than being recast as England-only evidence.
Selection required a dated event tied to one seller, channel, eligible population, period, timezone and decision owner, plus a current authoritative page defining the control or measurement boundary. Search volume, supplier leads, broad retail totals, anonymous claims, forecasts, assumed benchmarks and unsupported before-and-after accounts were excluded. No selected source or organisation funded the review.
Access failure under a repeatable task
A repeatable barrier in browsing, basket, checkout or support merits remediation research when a disabled participant can reproduce the failure. Record the task, assistive technology and journey step; an automated score alone is neither a legal finding nor proof of customer demand.
Use accessible methods suited to the task, then preserve what happened and where access failed. The enacted text addresses service access and reasonable adjustments. Record the investigation and decision in a planning template, keeping the finding tied to the particular task tested.
An extra with no recorded selection
An optional charge without a recorded express customer action is a checkout control issue, not evidence that a default improves completion. Compare the configured extra with an event record showing that the customer affirmatively selected that specific option for the order before payment, using the matching checkout version.
CMA guidance on additional charges discusses optional items such as faster delivery. If priority delivery appears in the order total without a separately recorded click before payment, pause that checkout version until express selection is captured. A missing event does not prove that the extra was selected, so assess the configured option alongside the event record.
Consent records versus live tags
Treat a difference between a tag's purpose, the recorded choice and its observed activation as a control signal.
Match each cookie or similar technology, pixel, tag or script to its stated purpose, legal route and activation status before using the resulting behaviour to interpret store performance or decide on a change.
ICO guidance treats storage and access technologies under PECR and, where relevant, UK GDPR. If an advertising tag fires before an affirmative choice appears in the consent record, remove or reconfigure it pending qualified privacy review. A mismatch also limits conclusions from the data, because tracking cannot reveal behaviour it failed to record.
Rate movements under different counting rules
If an event rate changes when calculated by sessions, users or journey starts, that movement signals a denominator definition issue, not which calculation is correct. Preserve the raw event count and record the denominator and implementation version before comparing reports or judging performance.
For example, 1,000 add-to-basket events across 1,250 sessions and 1,000 users produce 80% per session and 100% per user, while the raw count remains 1,000. Google's global Analytics dimensions and metrics record gives a session key-event formula; apply it only when its definition matches the merchant implementation.
Refunds that cannot be joined to orders
A cancellation or refund that cannot be joined to its eligible order can distort the retained-order rate and revenue interpretation. Trace each order through cancellation, return and refund states using one identifier and a stated cut-off; the mismatch may call for data or process work rather than a conversion supplier.
GOV.UK information on returns and refunds describes general and distance-selling routes. For example, a refund may carry a customer reference that matches no order reference, leaving it outside the retained-order count. Give unmatched records an owner and cut-off date, then use order-state reporting to keep the trail auditable.
Checkout information missing at a test point
A failed authorised test of product information, total price, delivery, payment action or confirmation warrants a review of content and applicable consumer rights. It does not show that shorter copy or fewer choices will increase valid orders; check the UK decision gates before altering the journey.
GOV.UK online-selling information and distance-contract rules set out information and order steps for online sellers, including pre-contract information. One reproducible failure is a payment page where delivery cost appears only after a paid option is selected, leaving the total price absent at the outset. Keep the screenshot and journey version with the finding.
Keep each signal with its event, population, numerator, denominator where relevant, period, timezone, source, uncertainty and owner. If the team cannot reproduce it, mark it unresolved. A reproducible symptom and authority to investigate begin the work, but neither amounts to a favourable sales story.
