Benchmarks · Retail

Retail website benchmarks

From 38 deep scans of retail websites, window ending 2026-09-07.

What retail websites most often get wrong, measured. These figures come from 38 deep scans of publicly reachable small-business sites in Go Voltic's research corpus, sector-labelled, scanned at up to 40 pages per site during the 30 days ending 2026-09-07. No business is named and none can be identified: aggregates only. With 38 sites the intervals are wide; read the ranges, not just the point.

The most common findings in retail sites

Finding	Share of sites	What it means
Third-party requests on first load	100%95% CI 90.8% to 100%	The first page load contacts outside services before any consent is given.
Images missing alt text	89.5%95% CI 75.9% to 95.8%	Images without an alt attribute, invisible to screen readers and image search.
Serious accessibility violations	86.8%95% CI 72.7% to 94.2%	Automated WCAG testing reports violations at serious or critical impact.
Wrong number of H1 headings	78.9%95% CI 63.7% to 88.9%	Pages rendering zero or several visible H1 headings instead of one.
Render-blocking scripts	78.9%95% CI 63.7% to 88.9%	Head scripts without defer or async that stop the page from drawing.
Meta descriptions missing	73.7%95% CI 58% to 85%	Pages without the summary line search results and AI answers lift.
Tap targets too small	73.7%95% CI 58% to 85%	Interactive elements under 44 by 44 pixels at phone width.
Slow first byte	60.5%95% CI 44.7% to 74.4%	At least one page takes over 800 ms to return its first byte on a cold request.

Share is the fraction of the 38 scanned retail sites that raised the finding. Intervals are 95% Wilson intervals computed from n = 38.

How this compares

The site-wide benchmarks draw on a much larger population of four-page scans. Deep scans read up to ten times more pages per site, so the two populations are measured at different depths and are never pooled. Compare the pattern of findings, not the raw shares.

Method

Population: 38 sector-labelled deep scans from our research sweep in the 30 days ending 2026-09-07. Sector labels are assigned conservatively; a site whose trade is unclear is bucketed as unknown rather than guessed, and unknowns are excluded here. Each finding is a deterministic check, identical to the checks in our paid reports. Sites that refused the scan are excluded and counted, never scored as zero. More: methodology.

Run the free scan on your own site