Technical audit of an online store: the visible storefront above the waterline and the invisible foundation below it, with speed, security, database and tracking icons

Technical Audit: What I Found on a Store That Was Crashing

Contents 9 sections · 5 subsections▼

A technical audit checks the invisible foundation of an online store: speed, security, the way you measure your sales, legal compliance, and how Google and AI read your site. Not the design. Not the copy on the product page, which is a separate conversation. The things you don’t see, but that decide whether your site actually works or only looks like it does.

I was called in for marketing. I stayed for what I found underneath.

A site can be running and sick at the same time

A crashed site at least announces itself. The most dangerous one keeps running with a problem no one can see. A nautical equipment retailer in southeastern Romania reached out to me for digital marketing. A mature store, over a thousand products, steady sales for years. On paper, everything was fine.

Then I started the technical audit.

The site crashed at five simultaneous visitors. Five. Response time climbed to nine, ten, fifteen seconds, and two entire categories simply stopped loading. And that was just the surface. A few weeks later, while I was working on stabilizing it, I ran into the thing that stopped my breath: a single database table had grown to 270 gigabytes. The real data: about fifty megabytes. The rest was garbage, written automatically at a rate of tens of gigabytes per hour by a payment plugin that was installed but never used, logging something on every single page load in the admin.

No one had noticed. The site was running. Until the day it wasn’t, and the server returned an error because the disk was nearly full.

That’s what a technical audit does. It looks for the time bomb before it goes off.

What a technical audit actually is

A technical audit for an online store is a systematic check of everything that holds the site up beneath the visible layer: infrastructure, security, speed, measurement, legal compliance, and readability for search engines and AI. It’s not an opinion. It’s a list of layers checked one by one, each with evidence behind it.

What exactly gets checked, across each of the fifteen trades that go into a website, is set out in what a properly built website means.

The difference between a serious audit and a „quick check” comes down to a single principle: I take nothing on faith. The fact that a plugin is installed doesn’t mean it works. The fact that you have a tracking code doesn’t mean it measures correctly. The fact that you have a cookie banner doesn’t mean you’re compliant. Every technical claim gets verified live, on the real site, not in theory.

On this store I checked nine layers. Real speed and performance, not the momentary score. Server and WordPress security. The way sales are measured in Google Analytics. Compliance with data protection and cookie law. The SEO structure of the thousand-plus products. How AI does or doesn’t reach the site’s content. At every layer, the same simple question: does this thing actually do what we think it does?

The answer was „no” more often than you’d expect.

Speed: why a fast site turns slow in a few months

An online store that’s fast at launch can turn sluggish within months, without anyone breaking anything on purpose. You add, one at a time, an extension, a tracking script, a pop-up, a product feed, a few unprocessed images. Each one seems harmless. Together, they choke the site.

At the nautical retailer, the images all loaded at once, none of them lazy-loaded. Dozens of scripts blocked the page from rendering before the text even appeared. And in front of the site, there was nothing to protect it from a traffic spike. At five people at the same time, it went down.

The fix wasn’t a miracle. It was hygiene. I put a content delivery network in front of the site that serves pages from a fast memory layer, close to the visitor, instead of hitting the tired server every time. Categories that used to struggle for several seconds started responding instantly from that layer. I upgraded the PHP version, cleaned out eight vulnerable or useless extensions, cut the installed themes from eight to three, and took the site’s critical health issues from two to zero.

Here I have to be honest with you, because honesty is part of the job. Perceived speed improved visibly thanks to serving from memory. But a rigorous re-measurement of Core Web Vitals, the ones Google actually uses as a ranking signal, happens after the dust settles, at thirty days. A serious audit tells you both what it fixed and what still has to be confirmed with numbers. The rest is their marketing, not mine.

The most elegant cookie banner in the world is worth nothing if it doesn’t ask for and record the visitor’s real consent. This site had a banner. It looked good. It had an „Accept” button. And it did absolutely nothing.

I tested it live, step by step, in a clean browser session. The banner was stitched directly into the theme code; it didn’t come from a real consent tool. It had a single button, „Accept,” with no equivalent reject option. And the worst part: when you clicked „Accept,” nothing happened behind the scenes. The signal that should have told the tracking systems „the person agreed, you can measure now” never fired. In practice, consent stayed stuck on „denied” indefinitely, no matter what the visitor chose.

Two problems from a single broken piece.

The first is legal. The EU rules on personal data and ePrivacy require consent to cookies to be given actively: the Court of Justice of the European Union ruled in the Planet49 case that a pre-checked box does not amount to valid consent. As for refusing, France’s data protection authority (CNIL) requires it to be as simple as accepting, on the same screen. The risk isn’t theoretical: in a 2023 report by the European cookie banner taskforce, the vast majority of data protection authorities considered that a banner with an accept button but no reject option on the same layer breaches the rules on valid consent. The banner here, with its single button, was exactly that case.

The second is a business problem. Because the consent signal never fired, the data in Google Analytics came in degraded, aggregated, unclear. The store was making decisions on numbers it couldn’t trust. I replaced the fake banner with a real consent platform, with three clear options, accept, reject, and customize, and set up the proper recording of the decision. Only then did consent start to mean something, and the data start to be collected with people’s real agreement.

Why AI doesn’t read you

More and more people no longer search on Google; they ask ChatGPT, Gemini, or Perplexity directly, and if your site blocks these systems’ bots, you become invisible exactly where attention is moving. I wrote at length about what changes when the customer stops searching and asks an AI agent instead. It’s a new layer of visibility, GEO, optimization for generative engines, and almost no one checks it in an audit.

I ran into an irony worth knowing. When I put the protection network in front of the site so it would stop crashing, its default configuration blocked a good portion of the AI bots. The very measure that made the site safer was quietly making it invisible to ChatGPT and the others. No one gets a notification for something like that. You just stop showing up, without ever learning why.

I checked ten relevant bots one by one, from ChatGPT’s to Perplexity’s to those of Google and assistants like Claude, and I unblocked access for all of them. Ten out of ten, confirmed by direct test.

Visibility in AI answers isn’t solved by that alone; it also takes content structured around questions and answers, clearly marked-up data, and a way of writing that machines can understand. But none of that matters if, at the lowest level, you’ve shut the door in their face without knowing it.

The catalog that sabotages itself

An online store with thousands of products can sabotage its own Google rankings through technical problems no human eye catches during normal browsing. You only see them when you run the catalog through an automated analysis, product by product.

On this catalog I found the classic e-commerce pattern, taken to the extreme. Over a hundred product pages had a technical tag, called a canonical, that told Google „this page doesn’t matter, count the other one instead.” Except „the other one” was, in many cases, a deleted product, or an engine with a different horsepower, or something else entirely. In effect, real products with real demand were shouting „ignore me” at Google. And Google listened.

Then there was the rest. A homepage with no structured title, no description for search results, none of the data that tells Google what kind of business you are and where you’re located. Nearly two-thirds of the products with no description for search results. Zero product-type structured data across the whole catalog, which means zero shot at the rich results with price and stars in Google.

The good news is that almost everything technical gets fixed programmatically, in bulk, not manually product by product. The honest news is that this fix was still on the to-do list as I write this, because order mattered. First I kept the site standing. Then I gave it its data back. Catalog SEO comes on a foundation that no longer collapses, not the other way around.

The order that matters: foundation first, then growth

On a site that crashes at five visitors and whose data lies, any promotion campaign is money poured down the drain. That’s the lesson I carry with me from this project, and the one I now apply to every new client, before anything else.

It’s tempting to jump straight to the fun part. Ads, posts, articles, traffic. But if you send people and AI to a site that collapses, reports false numbers, and shuts the door on the bots, you’ve built a beautiful house on sand. Foundation first. Then the walls. Then the paint. Because marketing, at its core, is measured by how solid what you’ve built really is, not by how loudly you shout.

A technical audit isn’t a document you file away in a drawer. It’s the map that tells you what order to fix things in, what it costs you now, and what it will cost you if you put it off. I won’t promise you the fix is spectacular. I promise you it’s real, verified live, with evidence behind every claim.

If you have an online store that „works,” but you feel something creaking underneath, this is the moment to look at the foundation. Before the time bomb goes off.

Frequently asked questions

What is a technical audit for an online store?

It’s a systematic check of the site’s invisible foundation: speed, security, infrastructure, the way sales are measured, compliance with data and cookie law, the SEO structure of the products, and access for search engines and AI. Unlike a content or design audit, a technical audit looks at what holds the site up, not at what’s on the surface.

How often should a technical audit be done?

At least once when you take over a new site, and then periodically, ideally once a year or whenever you add a lot of extensions, change the theme, or migrate hosting. A store that’s fast at launch can turn slow and insecure within months as scripts, plugins, and data pile up, without anyone breaking anything on purpose.

Why is a WooCommerce site slow?

The most common causes are unoptimized images, the lack of a caching system, weak or overloaded hosting, too many extensions and render-blocking scripts, and a neglected database. Speed often drops gradually, by accumulation, and you notice it only when the site becomes visibly sluggish or goes down under traffic.

No. Compliance requires that refusing cookies be as easy as accepting them, that options not be pre-checked, that the visitor be able to withdraw consent, and that their decision actually be recorded and respected by the site. A display-only banner, with a single button, that doesn’t pass the person’s choice along, does not make you compliant and exposes you to penalties.

What is GEO and why does it matter for my site?

GEO means optimizing your site for generative AI engines, like ChatGPT, Gemini, and Perplexity, so they can read you and mention you in their answers. It matters because more and more people ask AI directly instead of searching on Google, and if your site blocks their bots or has no structured content, you become invisible exactly where attention is moving.

About the author

Georgiana Manolache, marketing consultant

Georgiana Manolache, 22 years in marketing and communication. I build the message and the path that bring you customers, not traffic without sales. I apply buyer psychology through a time-tested method, supported by AI tools.

I wrote the book „Tell Them What They Want to Hear” for entrepreneurs who want to be seen, heard, and chosen by the right customers, without pouring money into channels that have no strategy.

Read more about me →

More articles on the blog →