Ecommerce & Retail · Growth problem
Traffic isn't converting
A falling or stubbornly low conversion rate is one of the most commonly misdiagnosed problems in ecommerce, because it's usually treated as a single metric to optimise rather than a symptom with several possible independent causes. The right diagnosis depends on segmenting conversion by device, source, and category, then working out whether the issue sits with the traffic itself, the product experience, the checkout, or the operational reality behind the product.
Symptoms
What this usually looks like
- Overall conversion rate is falling despite no obvious change to the site
- Mobile conversion rate is materially lower than desktop
- Conversion rate varies widely by traffic source, with some channels converting well below others
- Cart or checkout abandonment has increased
- Conversion rate is weak on specific categories despite comparable traffic volume
- Bounce rate on product pages has increased
Diagnostic questions
What we would test first
- Segment CVR by device, source and category over the last two comparable quarters
- Map the checkout funnel step by step to identify the highest drop-off point
- Compare add-to-cart rate against completed purchase rate by category
- Audit PDP content depth (sizing, delivery, returns, imagery) against the highest-traffic, lowest-converting categories
- Check stock availability correlation with conversion rate at category level
- Review Core Web Vitals and load time on mobile for top PDP and category templates
Root causes
Why it happens
- 01
Traffic mix shift towards lower-intent sources
If the composition of traffic has changed — more social, more broad-match paid search, more top-of-funnel content — a falling blended CVR may simply reflect a genuine change in average purchase intent, not a deterioration of the site experience itself.
- 02
PDP quality gaps relative to purchase decision requirements
Product pages that lack the specific information a category needs to convert — sizing detail, materials, delivery timelines, return terms, sufficient imagery — will underperform regardless of traffic quality, and the gap is often category-specific rather than site-wide.
- 03
Mobile experience lagging desktop
Where the majority of traffic is mobile but conversion rate is materially lower than desktop, the friction is frequently in page speed, checkout field design, or payment options rather than in demand itself.
- 04
Stock or delivery reality undermining otherwise strong intent
A customer who reaches a product page ready to buy, only to find limited stock, an unclear or slow delivery estimate, or a delivery cost surprise late in checkout, will frequently abandon — and this shows up as a conversion problem, not a merchandising or stock problem.
- 05
Trust signals insufficient for the price point or category
Higher-consideration or higher-price categories typically require stronger trust signals — reviews, guarantees, clear returns policy, payment security — than lower-consideration ones, and a generic site-wide approach to trust content can leave specific categories under-supported.
- 06
Payment or finance options not matching customer expectation
The absence of an expected payment method, buy-now-pay-later option, or clear total cost (including delivery) before the final checkout step can suppress conversion specifically at the point of payment, which shows up in checkout funnel data rather than on-page behaviour.
Evidence
The numbers we would look at
These are the metrics that make the constraint visible, and the cuts that stop them being reassuring by accident.
| Metric | What it tells you | How we read it |
|---|---|---|
| CVR by device (mobile vs desktop) | Identifies whether the problem is experience-specific rather than a broad demand issue.Gap should be interpreted alongside traffic share by device, not device CVR alone. | Gap should be interpreted alongside traffic share by device, not device CVR alone. |
| CVR by traffic source | Distinguishes a traffic-quality issue from a site-experience issue.Compare against historical CVR for the same source, not against other sources. | Compare against historical CVR for the same source, not against other sources. |
| CVR by category | Reveals category-specific conversion problems that a blended figure would hide.Segment against category-level traffic volume and stock availability. | Segment against category-level traffic volume and stock availability. |
| Checkout abandonment rate and step-level drop-off | Localises where in the funnel customers are lost, distinguishing a PDP problem from a checkout problem.Step-by-step, including where delivery cost or timeline is first shown. | Step-by-step, including where delivery cost or timeline is first shown. |
| Add-to-cart rate vs completed purchase rate | Separates a product-page interest problem from a purchase-completion problem.By category and device. | By category and device. |
| Site speed / page load time by device | A frequently underestimated driver of mobile conversion specifically.Core Web Vitals on the highest-traffic PDP and category templates. | Core Web Vitals on the highest-traffic PDP and category templates. |
Measurement traps
What can mislead you
- Looks fineOverall CVR has dropped, so the site has a problem
- A blended CVR fall can be entirely explained by a shift in traffic mix towards lower-intent sources, with no change at all to the underlying site experience — segmenting by source is essential before concluding the site is the issue.
- Looks fineMobile traffic converts worse, so mobile UX must be improved first
- Mobile traffic frequently includes a higher share of research-stage, lower-intent visits than desktop by nature of how it's used; the gap may be partly structural rather than entirely fixable through UX changes.
- Looks fineCart abandonment is high, so checkout needs a redesign
- High cart abandonment is often driven by unexpected delivery costs or timelines revealed late in the journey, which a checkout redesign won't fix if the underlying delivery proposition itself is uncompetitive.
Outcome
What better looks like
Not a promised number. A clearer basis for the next investment decision.
- Conversion is understood and reported by device, source and category, not as a single blended figure
- Any CVR decline can be attributed to a specific, evidenced cause before a fix is commissioned
- PDP and checkout investment is prioritised by where the data shows the largest, most fixable drop-off
- The team has a shared view of which conversion gaps are structural (traffic mix) versus addressable (site experience)
Where a Growth Diagnostic would start
A three to four week senior review across demand, discovery, acquisition, conversion, retention, measurement and capability — sequenced so this problem is either confirmed as the constraint or ruled out early. Read alongside the ecommerce & retail model page for how we frame the wider system.
Related problems
Ecommerce & Retail
Ecommerce growth has stalled
Revenue has flattened, or is growing without a corresponding rise in profit. A look at where the real constraint sits.
Ecommerce & Retail
Growth stalled after a replatform
Traffic or conversion dropped after moving platforms. A structured way to separate migration issues from unrelated factors.
Consumer & Omnichannel
Store and ecommerce are competing instead of compounding
Separate P&Ls, KPIs and stock pools put store and ecommerce teams in competition for the same customer.
Questions about this problem
- Do you need access to our analytics platform and session recordings?
- GA4 or equivalent event-level data is usually sufficient for the segmentation work; session recordings or heatmaps are useful supporting evidence but not essential to reach a first diagnosis.
- How quickly can you tell whether it's a traffic problem or a site problem?
- Segmenting CVR by source against historical baselines usually gives a fairly clear initial read within the first stage of analysis, though confirming the specific site-side cause typically needs a further look at funnel and category data.
- We've already run A/B tests on the PDP without much change — what's different here?
- A/B testing optimises within an existing hypothesis; this review is aimed at establishing whether the hypothesis itself is right — for example, whether the real constraint is stock, delivery proposition, or traffic mix rather than page layout.