What a website audit should actually catch.

A franchisee I know forwarded me a website audit last year. Forty-one pages. There was a score on the cover, 62 out of 100, in a font size normally reserved for stadium signage. Inside: meta description lengths, a count of images missing alt attributes, a note that three pages had more than one H1, and a recommendation to "consider compressing assets." Color-coded. Red, yellow, green. Professional-looking work.

It did not mention that the online order form on her site had been silently failing on iPhones. Not erroring. Failing quietly, the way the worst bugs do. You fill it out, you hit submit, the button greys out, and nothing arrives on the other end. Best we could tell, it had been doing that for somewhere north of eight months.

The audit crawled every page on the site and never once submitted the form.

Most audits are a crawler wearing a trench coat

Here is the uncomfortable mechanic behind the majority of free and cheap website audits: a tool crawls your site, a script formats the output, and someone adds a cover page with your logo on it. Nobody opened the site on a phone. Nobody clicked "Book now." Nobody watched what happens when you type a nine-digit phone number into a field that expects ten.

That is why so many audits read the same. They report what a machine can see from the outside, which is markup, headers, and file sizes. Those things matter at the margins. They are not what is costing you money, and they are almost never why you commissioned the audit in the first place.

The tell is the score. A single number from 0 to 100 is a sales device, not a diagnosis. It exists so the next slide can say "we can get you to 90." No operator has ever made a decision that improved their business because a number moved from 62 to 88.

Start at the money path, not the home page

Every operator site has a money path. It is the specific sequence a stranger walks to become revenue. Find the service, check the price or availability, fill something out, submit it, get a confirmation, and land in somebody's inbox or CRM where a human sees it.

That path is usually four to seven steps. It is where essentially all of the value is, and it is the part of the site that vendor audits skip, because walking it requires a person and a phone rather than a crawler and a license.

So an actual audit starts at the end and works backward. Where does a submitted lead land? Who gets notified? How fast? Does it hit the CRM, or an inbox nobody has opened since the office manager left? Then step backward one screen at a time, on a real device, until you reach the ad or the search result that brought them.

Do that once and you tend to find the expensive thing in the first twenty minutes. It is never a meta description.

The five things a real audit catches

1. Whether the money path completes on the devices people actually use

Not "is the site responsive." Responsive means the layout reflows. It says nothing about whether the third field in your form is reachable when the iOS keyboard is up, or whether the submit button sits underneath a sticky footer on a smaller screen. Submit a real test lead from an actual phone, then confirm it arrived. Then do it again from a different browser. Half the failures we find live in exactly that gap.

2. Where the form actually goes, and how long it sits there

Submission is not delivery, and delivery is not response. We have watched leads route correctly into a shared inbox and then sit for three days because nobody owned it. That is not a website problem, technically. It is still costing the same money as one, and any audit worth the invoice traces the lead all the way to a human with a name.

3. What speed is costing on the pages that carry traffic

Google's mobile research, still the most cited work on this, found that as page load goes from one second to three, the probability of a mobile visitor bouncing rises about 32 percent. At five seconds it gets considerably worse. The nuance most audits miss: it does not matter much on your careers page. It matters enormously on the one landing page absorbing your ad spend. A useful audit weights findings by traffic and intent instead of listing every page as equally urgent.

4. Whether people who need accommodations can use it at all

The WebAIM Million report for 2024 detected WCAG failures on 95.9 percent of the top one million home pages, averaging around 56 errors per page, and that figure got worse compared with the prior year rather than better. So the finding "you have accessibility errors" is close to meaningless. Everyone does. What matters is whether the failures sit on the money path: an unlabeled form field, a contrast level that makes your price disappear in sunlight, a booking widget that cannot be operated with a keyboard. Those are revenue problems that happen to also be compliance problems.

5. Who actually owns the accounts

This one is not technical and it is the one that bites hardest. Who holds the domain registrar login? Who is the primary owner on the Google Business Profile? Is the analytics property under your company account, or under an agency's? Whose credit card renews the SSL? We have sat with operators who discovered mid-rebrand that a former contractor still owned their domain and had stopped answering email. An audit that does not produce an ownership inventory has left the biggest risk on the table.

The findings that look like findings but aren't

To be fair to the format, some of this stuff belongs in an audit. It just does not belong at the front.

Duplicate H1 tags, missing alt text on decorative images, meta descriptions that run long, a handful of 404s pointing at pages you deleted on purpose in 2023. These are hygiene. Fix them on a Tuesday. None of them is why your cost per lead went up.

Two others get oversold constantly. First, "your site is not optimized for mobile," delivered on a site that is fine on mobile, because the tool flagged one tap target. Second, keyword volume tables, which get presented as a strategy and are really just a list. A list is not a plan, and a plan is not a build.

The checkout world illustrates the whole problem cleanly. Cart abandonment has hovered around 70 percent for years, and Baymard's testing attributes the recoverable share to a fairly small set of specific design failures: surprise costs surfacing late, forced account creation, checkout flows with too many steps. Those are findings you get by walking the flow like a customer. No crawler produces them.

What the deliverable should look like

Short. Ranked by money. Specific enough that someone could start Monday.

Every finding should carry three things: what breaks, roughly what it costs, and what it takes to fix. "Mobile form fails on iOS Safari, blocks an estimated 40 percent of inbound leads, about four hours to fix" is a finding. "Improve mobile usability" is a mood.

It should also be honest about what is fine. Most sites we look at have three or four real problems and a long tail of things that do not matter. Padding the list to justify the page count is how audits lose the plot, and it is how operators end up ignoring the two items that actually mattered.

Six pages beats forty-one. Always has.

Run the cheap version yourself this week

You do not need a vendor to catch most of this. Block ninety minutes. Pick up your phone, not your laptop. Pretend you have never heard of your company. Search the way a customer would, click your own listing, and walk the money path end to end with real information. Submit the form. Then go verify it arrived, and note how long that took.

Write down every place you hesitated, every field that fought you, and every price you could not find. That list, in order, is a better audit than most of what gets sold, because it came from a person trying to give you money.

Then do the boring part: open a document and list every account the business depends on, who owns it, and which email address it is tied to. Domain, DNS, hosting, analytics, Google Business Profile, ad accounts, form provider. If any line has a name on it that no longer works there, that is your first project.

That is roughly what we do on a Sweep, minus the part where we also tell you which of it is a two-week fix and which is a genuine rebuild. And whatever we find stays with you in a document you keep, in accounts under your name. When we walk out, the audit does not leave with us.