Priorities that have never seen your site

In June 2025 Screaming Frog published a guide to enterprise SEO audits whose central claim we agree with completely: a large site does not need a longer checklist, it needs a way to decide what to do first. The guide is right, and it stops exactly where the hard part starts.

To be fair to them, they do rank what they find. The Issues tab has carried High, Medium and Low across three hundred-odd checks since version 17. What is unusual, and to their credit, is the sentence they attach to it: those priorities rest on potential impact, and they are not hard rules for what should be prioritised in your strategy, or fixed in your audit, because they lack context.

That is an honest disclaimer and it names the gap precisely. A severity written into a catalogue cannot know that your four broken links are all in one menu, that your slowest template is the one nobody lands on, or that the thing at the top of your list is a filter you switched on deliberately. Severity is a property of the finding. Priority is a property of your week.

The first thing we do with a finding is decide which of three lanes it belongs in, and the lane is a claim about consequence rather than about difficulty.

A leak costs you traffic now: a menu link that answers 404, a filter space that has fanned out into fifty thousand crawlable variants, an add-to-cart link search engines are still following. Hygiene is real, worth doing, and will not change your week: a missing ETag, an absent Open Graph tag, a security header. Later is template work you should schedule rather than squeeze in.

Only the leak lane can take a slot in the week. That one rule removed most of the noise, because findings that pile up in the thousands are almost always hygiene, and the ones that actually cost money are almost always few.

• Core Web Vitals and slow responses moved to hygiene. They are real, they are developer work, and the speed gauge on your overview already shows them.
• A difference between the raw and the rendered page moved to hygiene too, unless the H1 itself only exists once scripts run. That is the difference between a widget loading late and Google seeing an empty page.
• A stale sitemap entry moved to hygiene, and folds into the orphan-pages card when it sits on the same URLs, because one dump of a catalogue is one problem and not two.

This is the decision that took us longest and changed the most. A card is sized by the number of edits it needs, not by the number of pages showing the symptom.

On a mattress shop we read recently, four links in the main navigation pointed at pages that no longer existed. Each of those four appeared on 4,344 pages. Counted the obvious way the card reads nine thousand eight hundred and eighty pages and around four hours, and the owner never opens it, because nobody starts a four-hour job on a Monday. Counted properly it reads four broken link targets and about twenty minutes, which is the truth. It was the first thing they fixed.

The same arithmetic runs everywhere. A faceted listing generating fifty thousand crawlable URLs is one line in robots.txt, so the card says how many listings leak and that the fix is a single edit. A title pasted across forty products is one template decision. Once you count the edit rather than the symptom, a site with sixty-three thousand raw findings turns into five decisions.

The week is capped at five cards. The cap is not a design flourish, it is the product's only real promise: that opening the app on a Monday gives you something you can finish.

Cards landing in the same place and in the same hands are bundled, so three separate findings about titles become one sitting in the CMS rather than three trips to the same screen. The ordering inside the lane is not the catalogue's either. It weighs how many edits the card needs, how many pages it reaches, what Search Console says those pages earn, and what kind of site this is, because a leak on a product page matters more to a shop than the same leak on a policy page.

Everything else is still there. The next lane holds leaks that missed the five, later holds the housekeeping, and the full list of every finding is one click away on the days you want it.

A tool that ranks things owes you the list of things it refuses to rank. We do not put crawl budget in your week, because crawling is necessary for a page to appear in search but it is not a ranking signal, and for the overwhelming majority of sites the topic does not apply at all.

We do not treat a page you deliberately keep out of search as a defect. A directory with thousands of noindex filter pages made a choice, and the card offers to record that choice rather than nag about it. You can snooze a card with a date or ignore it for good, and the next reading honours both, which is the part that lets a weekly list survive contact with a real site.

Where the catalogue genuinely cannot tell, we say so rather than invent a number. That is the same instinct behind Screaming Frog's disclaimer. The difference is only what we each do next with it.