How I audit websites: customer problems, page types, traffic patterns, and what actually matters

Most website audits I've seen share the same flaw: they start with the site. They crawl it, score it, run Lighthouse, review the copy, check the heatmaps. After two weeks there's a 47-slide deck and a list of 200 issues nobody agrees on.

I start somewhere else. Before I open a single tool, I ask: what do we know about the people who visit this site, what problems brought them here, and what do they need to be convinced of before they take action? Everything else follows from that.

Customer problems are the audit foundation. The most useful thing a company can give me before an audit isn't a Google Analytics login - it's their sales call recordings, support tickets, and churn reasons. Not because I'm doing a customer research project. Because these sources tell me exactly which problems the website is probably failing to address. If "I didn't understand what this is for" appears in five sales call transcripts, that's not a messaging note - it's a diagnosis. The homepage has a clarity problem, and I know where to look before I've loaded the URL.

Page types have different jobs, and most audits treat them the same. A homepage is a trust-builder and orientation layer. A features page is for people who already get what you do and want detail. A pricing page is where doubt crystallizes - people are looking for reasons to say no. A paid landing page is a contained persuasion unit with a single exit. Auditing all of these the same way produces generic recommendations. I look at each page type against its actual job: is this page doing what it's supposed to do for the person most likely to be on it?

Traffic types tell you who showed up and why. Organic traffic from an informational keyword is a different person than direct traffic from a branded search. Someone who clicked a retargeting ad has seen your brand before. A referral from a partner blog arrives with context you didn't have to create. Treating a session recording from a cold social visitor the same as one from a bottom-of-funnel email click produces bad conclusions. I always segment traffic sources before interpreting behavior data - otherwise the heatmap is just noise.

The patterns that repeat. After enough audits, certain problems appear on almost every site regardless of industry. The homepage tries to serve too many audiences at once and ends up speaking clearly to none. The pricing page lists plan features but never addresses the real objection, which is usually "I'm not sure this is worth switching for." The mobile experience is an afterthought on a site where 60% of visits are mobile. CTAs use vague generic language ("Get started", "Learn more") when every piece of research shows specificity converts better. None of these require deep analysis to find. They just require someone to look for them deliberately.

What I actually deliver isn't a list. It's a diagnosis. Three to five structural issues, each tied to a specific page, a specific audience, and a specific hypothesis about what fixing it would do to conversion. The difference between an audit that produces action and one that produces a spreadsheet nobody opens is whether the findings are specific enough to be disagreeable. Vague findings can't be acted on - and they can't be argued with, which is how they end up in a folder somewhere.

A note on tools. I use them. Analytics for traffic patterns and funnel data. Session recordings to watch where people drop, what they re-read, and what they ignore entirely. Performance data, because a slow page that kills conversion shows up in the numbers before anyone complains. But the tools only confirm or challenge what the customer problems already pointed me toward. They're evidence, not the investigation.

If you want this kind of audit run on your site - starting with customer problems, looking at each page against its actual job, and ending with specific findings rather than a listicle - that's what the Web Growth Audit is. Fixed scope, fixed price, delivered in two weeks with no meetings and no fluff.