In dependency order, not severity order. A page Google will not index cannot be rescued by a faster load, a sharper title or a cleaner form. Each layer is a precondition for the one above it, and work on an upper layer returns nothing while a lower one is broken.
Why does the order matter more than the list?
Because audit output is sorted by severity score, and severity is not dependency. A tool will flag a missing meta description on a page that Google has excluded from the index, and rank it next to the exclusion itself as though they were parallel tasks. They are not. Five hours spent rewriting the call to action on an unindexed page returns exactly zero, and the ticket still closes green. The ordering below is the one the mechanism imposes: crawl, index, canonical, relevance, presentation, offer.
Which layer is broken? Read it off the funnel.
The funnel in Google Search Console identifies the layer without any guessing, because each layer fails with its own signature.
| What the reports show | Broken layer | Where the fix lives |
|---|---|---|
| URL missing or excluded in the Page indexing report | Crawl and index | Robots rules, noindex, server errors, canonical targets |
| Indexed, zero impressions | Relevance | Nobody searches this, or your own site answers it better elsewhere |
| Impressions, almost no clicks | Presentation | Title, description, result shape, competition on the page |
| Clicks, no conversions | Offer | The page, the form, the price, the proof |
| All of the above working, page slow or unstable | Experience | Core Web Vitals and rendering |
Nothing above indexing is worth touching until indexing is fixed.
URL Inspection gives the verdict for one URL; the Page indexing report gives the pattern across the site. The exclusion reasons are not interchangeable. "Crawled, currently not indexed" means Google read the page and declined it, which is a content and demand problem. "Discovered, currently not indexed" means Google has not got to it, which is usually a server speed or internal linking problem. "Duplicate, Google chose a different canonical" means your canonical markup lost an argument it was making. Three different fixes, one symptom, and only the first of them is editorial.
Severity ranks problems by how bad they look. Dependency ranks them by what they block. Only the second ordering tells you what to do on Monday morning.
Zero impressions is a relevance problem, not a traffic problem.
If a page has been indexed for a month and still shows no impressions, the page is written about something nobody searches for, or a stronger page of yours already owns the query. Both are editorial calls, not technical ones. Filter the Performance report to the URL and look at the query list: an empty list is the first case, and a list of phrases that your other pages also rank for is the second. Two pages competing for one query split the topical authority you were trying to build, and the fix is to merge, retarget or retire one of them rather than to add more traffic tactics on top.
Impressions without clicks is a presentation problem, but only sometimes.
Average position decides which. A page sitting near the top of page one whose click-through rate is well below the other rows in your own report has a title or snippet problem you can fix this afternoon. A page averaging position thirty has a ranking problem wearing the same symptom, and rewriting the title will not move it. Check what Google actually prints, too, because it rewrites titles freely, and structured data changes the shape of the result and therefore how much of the page's attention it takes.
Clicks without conversions is not a search problem at all.
Conversion rate sits downstream of everything organic search does. If the query matched, the page loaded and the visitor arrived, search did its job and handed the problem on. What remains is the offer, the proof and the form. This is worth saying plainly inside a company, because the ticket usually stays with the search team by default, and the search team has no lever left to pull.
Where do Core Web Vitals belong in the order?
Wherever their symptom appears, which is usually lower than people place them. If a page renders so slowly or so incompletely that Googlebot gives up, the problem belongs at the indexing layer and goes first. If a layout shifts under a reader's thumb on a page that already ranks and gets clicks, it belongs at the offer layer and competes with the form and the price. Experience work in the middle, on a page with impressions and decent clicks, is real but rarely the binding constraint.
What if two layers are broken at once?
Fix the lower one, then re-measure before touching the upper one, because the lower fix changes the evidence you were about to act on. An indexing fix changes impressions, and impressions are the denominator of the click-through diagnosis you had queued next. Ship both in the same week and the result is unattributable: you will know something improved and not which change did it, which guarantees you repeat the expensive half next quarter.
What should you open first tomorrow?
Open the Page indexing report before the keyword tool. If the pages you care about are indexed and canonical, close it and go straight to the Performance report filtered to those URLs. Whatever that view shows decides the layer, and the layer decides the owner.
That is the real output of triage: not a list, but a name. This quarter's work belongs to an engineer, an editor or whoever controls the offer, and it belongs to exactly one of them at a time. Pick which before anyone writes a ticket.