What 37 Website Audits Taught Us About Small-Business Sites

We have published the raw findings from our audit corpus before. This is the other half of that work: the small business website lessons we drew from running the same rubric across 37 reproducible audit records and watching the same shapes appear. The individual numbers are useful. The pattern behind them changed how we quote work.

Every site in the corpus was scored on the same measurable criteria, without the owner present and without any incentive to flatter. That consistency is what makes the pattern legible. These figures describe the audit evidence currently committed with the project; they are not a random national sample. Here is what the corpus taught us.

Lesson one: sites do not break, they stop

We expected to find broken websites. We mostly found abandoned ones. Only one site in the corpus was genuinely non-functional, showing a suspended hosting page where the business should have been. One had broken images. Everything else loaded.

What we found instead was arrested development. Fifteen of the 37 sites carried a copyright year behind the current one. Twelve still displayed template placeholder text that shipped with the theme and was never replaced. Fourteen carried the signature of a dated or DIY builder that may have been a reasonable choice years ago.

These are not failures of construction. They are failures of maintenance, and they accumulate silently. Nobody gets an alert when a site becomes dated. The median score across the reproducible corpus was 23 out of 100, and almost none of those sites looked broken at first glance.

Lesson two: the cheapest defects can cost the most

The finding that reshaped how we prioritize work is that severity and repair cost are close to unrelated.

Thirteen of the 37 sites showed a phone number as plain text with no click-to-call link. That is one of the costliest defects we find, and one we would fix first on almost any service-business site, because it breaks the exact moment a mobile visitor decides to become a customer. It also takes a developer about a minute to fix. A tel: link is a few characters of markup.

The same principle applies to the viewport meta tag, absent on two sites, which is a single line in the page head and determines whether a phone renders the layout sensibly or shrinks a desktop page to thumbnail size. Pinch zoom was blocked on two sites, another setting that can be removed directly.

We now open most engagements by listing the fixes that cost nearly nothing, because for some businesses those alone recover much of the available value. That is a worse business model and a better service.

Lesson three: the thing owners look at is not always the thing that matters

Owners ask us about the logo and the colors. Those are the visible surface, and it is natural to focus there.

Logo usability was the most common finding in the corpus, scoring below our threshold on 28 of 37 sites. Logo modernity fell below threshold on 18. So the instinct is not wrong. But logo usability in our rubric is mostly a technical measure: whether the mark holds up at the size it is actually displayed, whether it survives a small screen, and whether it reads against its background. It is closer to a rendering problem than a taste problem.

Meanwhile defects that directly intercept a customer, such as an untappable phone number, are often invisible to the owner. The owner rarely visits the site as a stranger on a phone at 8 p.m. That gap between what gets noticed and what can cost money is one of the clearest patterns in the corpus.

Lesson four: thin content is the hardest problem

Seven of the 37 sites had homepages under 200 visible words. This is the finding for which we have the least cheerful answer.

Every other defect on the list is a repair. Somebody can add a tag, fix a link, or update a year. Thin content cannot be repaired by a developer alone, because the missing thing is knowledge only the business owner has: what the work involves, what customers ask, and what makes this operation different from the three others in town.

That means the sites with the thinnest content are the ones where our involvement helps least and the owner's involvement matters most. We have learned to say that plainly during the quote rather than discovering it halfway through a build.

Lesson five: the distribution is wider than expected

Scores ran from 4 to 100, with a median of 23 and a mean of 27.9. The ratings split into 2 HOT, 12 WARM, 11 COOL, and 12 LOW.

The useful part of that spread is that it defeats the generalization we started with. We began this work half-expecting to conclude that small-business contractor sites are uniformly bad. They are not. A meaningful share are in decent shape, and some are genuinely good. What separates them is often not budget. It is whether anyone has opened the site critically in the last two years.

That is the summary lesson, and it is the one worth acting on. The difference between a site scoring in the low twenties and one scoring near 70 is not always a redesign. It is often attention. This is why we pair builds with ongoing care. The build is the easy part. Whether anyone tends the site afterward is what determines where it sits in this distribution three years from now.