You can estimate the revenue a slow store is costing you with data you already have: segment mobile and desktop sessions, compare their conversion rates, and size the gap against your average order value. If mobile converts materially worse than desktop on the same products and prices, performance is the first suspect, because speed is the most common systematic difference between the two experiences. The point of the model is not precision — it is turning a vague frustration into a business case with a payback period.
Why a model beats a feeling
“The site feels slow” does not survive a budget meeting. “Mobile is 68% of sessions and converts at half the desktop rate; closing a third of that gap is worth roughly £X a month” does. Performance work competes for budget with campaigns, hires, and apps, and it wins or loses on whether the return is legible. A simple, conservative model also protects you in the other direction: if the numbers say speed is not your constraint, you can stop suspecting it and look at merchandising or offer instead.
The four numbers you need
- Sessions by device: the mobile/desktop split from GA4 or Shopify Analytics, over a recent 90-day window
- Conversion rate by device: same window, same source, segmented identically
- Average order value by device: mobile AOV often differs from desktop, so do not assume one number
- Core Web Vitals by template: mobile LCP, INP, and CLS from PageSpeed Insights or your RUM data, at the 75th percentile
None of this requires new tooling. Everything except possibly field-level vitals is already in the analytics you have; the discipline is in segmenting it honestly.
A simple model, run conservatively
Start with the device gap. Take your desktop conversion rate as the reference experience — same catalogue, same prices, same brand — and compare it to mobile. The gap will never be zero: phones are used in more distracted contexts, and some categories genuinely skew desktop. So do not model closing the whole gap; model closing a modest fraction of it. If mobile converts at 1.2% and desktop at 2.4%, ask what a lift of 0.2 or 0.3 percentage points on mobile is worth: mobile sessions times the rate change times mobile AOV.
Then stress-test the assumption that speed is the cause. Check that your mobile vitals actually fail the thresholds — LCP at or below 2.5 seconds, INP at or below 200 milliseconds, CLS at or below 0.1 at the 75th percentile. If mobile vitals pass everywhere and conversion still lags, your gap is UX, merchandising, or offer, and performance work is not the answer. The model is only honest if it can return that answer too.
A purely illustrative example of the arithmetic: with 100,000 monthly mobile sessions, a 0.1-point conversion lift is 100 extra orders; at an £80 average order value that is £8,000 a month. Substitute your own three numbers and you have the monthly value of each tenth of a point — the unit in which remediation costs should be discussed.
Finally, compare the estimated uplift against two costs: the cost of remediation, and the cost of doing nothing expressed as paid-media waste — the share of ad spend landing on slow templates that converts below its potential. For most stores buying meaningful traffic, even a conservative uplift estimate clears the cost of a focused performance sprint comfortably.
Where the model goes wrong
- Using blended conversion rate: mixing mobile and desktop hides the entire signal
- Modelling closing the full device gap: some of it is context, not speed — stay conservative
- Ignoring template mix: a slow blog and a slow product page are not the same problem
- Counting engagement gains twice: pick one primary outcome (revenue) and treat the rest as leading indicators
- Forgetting seasonality: compare like-for-like windows, not peak season against quiet season
From estimate to payback: a working checklist
- Export 90 days of sessions, conversion rate, and AOV segmented by device
- Compute the mobile/desktop conversion gap and the revenue value of a 0.1-point mobile lift
- Pull mobile field vitals for homepage, top collection, top product, and cart
- Identify which failing templates carry the most revenue-generating traffic
- Get a scoped cost for fixing those templates specifically — not a blanket “speed project”
- Set the pre-agreed measurement: which metric, which window, and what counts as confirmed
- Re-measure after the work and update the model with observed, not assumed, uplift
FAQ
How accurate can this estimate be? It is a sizing tool, not a forecast. Its job is to tell you whether the prize is worth a sprint, an audit, or nothing at all — and to give you a baseline so the post-work measurement is honest.
What if our mobile and desktop conversion rates are similar? Then speed is probably not your binding constraint, and the model has done its job. Look next at offer, merchandising, and checkout friction — or at whether both rates are suppressed by something sitewide.
Should we include desktop performance? Yes, but weight by traffic. Most DTC stores are majority-mobile, which is where the gap and the money usually are; a store with heavy desktop B2B traffic should weight accordingly.
Once the model shows a prize worth chasing, the work is iterative: fix the top template, measure, reinvest. That is the natural shape of a growth retainer — performance and CRO run as a continuous loop against an agreed baseline.
Performance and CRO as a continuous, measured programme — with the baseline agreed before work starts.
