A larger indexable set does not guarantee a larger ranking set. In this case, the top of the funnel improved while the bottom weakened: crawler reach rose, absolute indexable pages rose sharply, yet ranked pages declined. That is why SEO performance is easier to read as a funnel than as one blended visibility number. Google’s documentation separates crawling from indexing because they are related, but not interchangeable, processes (crawling and indexing systems).
The funnel improved at the top and weakened at the bottom
The figures below show association, not controlled causation. This is a two-snapshot comparison, so it is useful for direction and magnitude, not for proving why the change happened.
Funnel stage
Earlier snapshot
Later snapshot
Absolute change
Percentage-point change
Pages observed
1.7M
2.3M
+600K
—
Crawled share
8%
12%
—
+4 pp
Indexable pages
1.5M
2.0M
+500K
indexable rate 88% to 87%, -1 pp
Ranked pages
402K
383K
-19K
ranking coverage 27% to 19%, -8 pp
The arithmetic is the point. Indexable pages increased from 1.5M to 2.0M, a gain of 500K pages, or about 33%. Crawler reach improved from 8% to 12%, a four-point gain. If you stopped there, you could tell a success story.
But the bottom of the funnel moved the other way. Ranked pages fell from 402K to 383K, down 19K, while ranking coverage dropped from 27% to 19%, an eight-point decline.
That is the decision signal. When more URLs can be reached and more URLs are eligible for admission, but fewer URLs rank, the first reading should not be “crawlability improved, so SEO improved.” The bottleneck is not proven to be crawlability or indexability. It may still involve them at the margin, but the clearer reading is that downstream ranking coverage deteriorated faster than upstream eligibility improved. This is also why “missing from search” needs stage-specific diagnosis rather than one aggregate explanation (why a page may be missing).
One longer-window detail points the same way without overclaiming: non-indexable listing pages receiving crawl activity fell 24%. Some waste appears to have been reduced, yet the ranking set still shrank. So the practical lesson is simple: treat each gate separately, and do not infer ranking progress from better crawl or index numbers alone.
Crawlability, indexation and ranking are different — Convertos.ai original workflow poster.Crawlability, indexation and ranking are different — Convertos.ai original explainer.
This narrated explainer reduces the article to three checks: Fetch, Index, Compete. It does not add claims beyond the text. Captions and a transcript are included for accessibility.
Video transcriptCrawlability, indexation, and ranking are three separate diagnostic layers. First ask whether a search engine can fetch the page. Check status codes, robots rules, links, rendering, and logs. If fetching works, ask whether the engine kept the page in its index. Check canonicals, duplicates, soft errors, and page value. Only after both layers pass should you investigate relevance, authority, intent, and competition. Fixing titles cannot solve a blocked crawl. Resubmitting cannot make a weak duplicate rank. Diagnose in order.Top-of-funnel technical gains did not prevent ranked-page coverage from falling from 27% to 19%.
Crawlability answers only whether Google can reach the URL
Crawlability is the reachability gate. It answers a narrow question: can Googlebot discover the URL, request it, and receive a usable response? That depends mostly on internal links, XML sitemaps, robots rules, HTTP status, and network reliability, not on whether the page deserves to be indexed or can rank. Google’s documentation keeps crawling separate from indexing for exactly this reason: a URL can be fetched successfully and still never enter the index or search results (crawling and indexing systems, how crawling is affected by status, DNS, and network errors).
The case numbers show why the distinction matters. In one snapshot, 1.7M pages were observed and 8% were crawled. In the later snapshot, 2.3M pages were observed and 12% were crawled. That is a four-point gain in crawler reach, but the denominator also expanded sharply. In absolute terms, crawled pages rose from about 136K to about 276K.
That means reach improved. It does not mean those URLs were indexable, useful, canonical, or competitive. The sample is limited to two snapshots from the same case, so the figures show association, not controlled causation.
A practical read of crawlability combines logs with URL state checks. Server logs tell you whether Googlebot actually requested the URL. Status codes tell you whether the fetch succeeded. Robots rules tell you whether access was blocked. Internal links tell you whether discovery is plausible at scale. Crawl frequency then becomes a prioritization signal: URLs that are linked and repeatedly fetched are reachable; URLs that exist in exports but never appear in logs usually have a discovery problem before they have an indexing problem.
Crawlability check
What to verify
Decision use
Log presence
Googlebot requested the URL
No log hit usually means discovery or access failed
HTTP status
200 for target URL, not persistent 3xx/4xx/5xx
Fix transport before judging indexation
Robots
Not blocked by robots.txt or noindex at fetch time
Remove accidental blocks before deeper analysis
Internal links
URL is linked from crawlable pages
Add or repair links if logs are sparse
Crawl frequency
Repeated fetches vs one-off fetches
Repeated crawling means reachable, not necessarily index-worthy
One longer-window signal in the same case sharpens the point: non-indexable listing pages receiving crawl activity fell 24%. That is useful because it suggests crawl attention shifted away from some low-value URLs. But it is still a crawlability observation first. It says Google reached fewer non-indexable listings, not that ranking quality improved elsewhere.
Indexability is an admission decision, not a crawl count
A URL becomes indexable when nothing important tells Google to keep it out, and when the page is eligible to be consolidated as the version worth storing. That is a different question from whether Googlebot merely reached it. Crawl data says “seen.” Indexability says “admitted.” A page can be fetched and still be excluded because of noindex, canonical selection, duplication, soft errors, or other signals that make another URL the preferred representative in the index (crawling and indexing systems, why a page may be missing).
In the sample here, the later snapshot showed more pages observed and more crawler reach, but indexability needs its own check. Earlier, 1.7M pages were observed, 8% were crawled, and 1.5M were classified as indexable, or 88%. Later, 2.3M were observed, 12% were crawled, and 2.0M were indexable, or 87%. This is a snapshot comparison, not a controlled test, so it shows association only.
Still, the arithmetic matters. Absolute indexable pages rose by about 500K, roughly 33%, even while the indexable rate slipped by one point because the denominator expanded faster than admissions did.
Snapshot
Observed pages
Crawled rate
Indexable pages
Indexable rate
Earlier
1.7M
8%
1.5M
88%
Later
2.3M
12%
2.0M
87%
That pattern can fit technical cleanup on exclusion signals. If noindex was removed from pages meant to rank, canonicals were corrected so valid URLs no longer pointed away, hreflang clusters stopped nominating the wrong alternates, or sitemaps stopped over-submitting non-canonical and blocked URLs, more pages could become admissible without any guarantee that they would later rank. The longer-window clue supports that reading: non-indexable listing pages receiving crawl activity fell 24%. Less crawl appears to have been spent on pages that were never eligible in the first place.
The decision rule is straightforward: treat indexability as passed only when exclusion evidence improves on its own terms. Do not infer it from crawl growth. Check whether the pages intended to rank are free of noindex, self-canonical or canonically selected, correctly represented in hreflang sets, and included in clean sitemaps. Then verify indexed evidence separately at the URL-pattern level. If crawl rose four points and indexable pages rose 33%, that is progress at the admission gate, not proof that ranking coverage should rise with it.
Ranking coverage can fall after the first two gates improve
This is where teams often misread the funnel. In this sample, the upstream gates improved across two snapshots, yet the ranking gate weakened. Observed pages rose from 1.7M to 2.3M. Crawl reach moved from 8% to 12%. Indexable pages increased from 1.5M to 2.0M, roughly a 33% gain in absolute terms, while the indexable rate stayed nearly flat at 88% versus 87%. But ranked pages fell from 402K to 383K, and ranking coverage dropped from 27% to 19%. The period and sample are limited to these snapshots, so the figures show a pattern, not controlled causation.
A simple reading is that more URLs became eligible, but a smaller share earned visibility. Ranking is not awarded because a page was crawled or admitted to the index. It depends on whether the page is competitive for a query cohort that actually exists, whether Google sees enough distinct value on the page, and whether the site routes enough internal equity to that URL to justify surfacing it. That is why “indexed” never guarantees rankings (crawling and indexing systems, why a page may be missing).
Snapshot
Observed pages
Indexable pages
Ranked pages
Ranking coverage of indexable
Earlier
1.7M
1.5M
402K
27%
Later
2.3M
2.0M
383K
19%
That table isolates the downstream loss. The site added about 500K indexable pages, but ranked pages declined by 19K.
That kind of deterioration can point to a short list of practical issues: quality dilution, demand mismatch, internal equity spread, or stronger competition. New or expanded URL sets may be technically indexable but too thin, duplicative, or weakly differentiated to win queries. Pages can also be valid and useful yet target little or no search demand. And when the indexable set grows faster than internal linking support, more pages compete for the same authority pool.
One clue supports that reading. Non-indexable listing pages receiving crawl activity fell 24% over a longer comparison window. Some crawl waste was reduced, which is good, but it still did not translate into broader ranking coverage. So if crawlability and indexability improved while ranking coverage fell, stop blaming discovery first. Audit query demand, page distinctiveness, internal linking concentration, and SERP competition before making another crawl-budget argument.
Diagnose the first failed gate and stop there
When more pages become crawlable and indexable but fewer pages rank, the right diagnosis is not “SEO got worse” in the abstract. It is “which gate failed first?” The sequence matters because each stage depends on the one before it. If a URL cannot be reached, ranking is irrelevant. If it is reached but not admitted to the index, ranking is still irrelevant. Only after those two checks pass does ranking coverage become the live problem. Google’s documentation separates crawling from indexing for exactly this reason: they are related systems, not the same outcome (crawling and indexing systems, why a page may be missing).
Use a simple stop-at-first-failure decision tree. Start with the affected URL set, not the whole site. Ask first: can Googlebot reach these URLs reliably? If no, collect status, DNS, timeout, robots, and discovery evidence and stop there. If yes, ask next: are they indexable and actually indexed? If no, collect canonical, noindex, duplicate-cluster, and “crawled or discovered, currently not indexed” evidence and stop there. Only if both answers are yes should you investigate ranking loss with query, intent, internal linking, and page quality evidence.
That discipline prevents a common reporting error: explaining rank decline with crawl metrics when the pages in question were never admitted to the index.
A worked example from the available snapshots shows why. The sample is limited to two observed periods and shows association, not controlled causation.
Gate check
Earlier snapshot
Later snapshot
What it means
Pages observed
1.7M
2.3M
Larger known URL universe
Crawled share
8%
12%
Reach improved by 4 percentage points
Indexable pages
1.5M
2.0M
Absolute indexable count rose about 33%
Indexable share
88%
87%
Admission rate was roughly flat
Ranked pages
402K
383K
19K fewer ranked pages
Ranking coverage
27%
19%
Bottom-of-funnel weakness
The stop point here is the ranking gate, not crawlability and not indexability. Crawler reach improved, and absolute indexable pages also increased materially. Even the longer-window signal supports that upstream cleanup happened: non-indexable listing pages receiving crawl activity fell 24%. Useful, yes, but it does not explain why ranked pages dropped from 402K to 383K.
Once the first failed gate is ranking, the evidence requirement changes. You need page groups that are both crawlable and indexable, then compare which of those lost rankings and what changed in their query fit, internal prominence, or competitive position. Without analytics, the clicked stage is unavailable, so stop before making traffic or business claims.
The missing clicked stage changes what can be claimed
The funnel in this case stops at ranked pages because analytics was not connected. That changes what can be said. We can say crawler reach improved in the observed snapshots, from 8% of 1.7M observed pages to 12% of 2.3M. We can say absolute indexable pages rose from 1.5M to 2.0M. We can also say ranked pages fell from 402K to 383K, so ranking coverage dropped from 27% to 19% within the compared snapshots.
What we cannot say is whether search clicks, sessions, leads, orders, or revenue rose or fell as a result. Ranking is a visibility state, not a traffic outcome.
That missing stage weakens every business claim. A loss of 19K ranked pages may be harmless if those URLs sat low in results and never earned demand. A smaller set of newly ranked pages could still outperform them if they entered higher-intent queries. The reverse is also true. More indexable pages can coexist with weaker traffic if the added URLs target thin demand, duplicate intent, or low-value query classes. Google’s documentation separates crawling, indexing, and serving, which is why downstream performance cannot be inferred from upstream access alone (crawling and indexing systems).
Stage available
What the data supports
What it does not support
Crawled
Google reached more of the observed URL set
That users saw or visited those pages
Indexable
More URLs were eligible for inclusion
That Google chose them for valuable queries
Ranked
Fewer URLs appeared in rankings in the later snapshot
That clicks or revenue necessarily declined
Clicked
Not available in this case
Any claim about traffic, conversion, or ROI
So the concrete decision is this: treat the current bottleneck as ranking coverage, not crawlability or indexability, and do not make traffic or ROI claims until the clicked stage is available.
Annotate the release and its next check in the SEO audit workflow. The Convertos SEO guides is useful after the evidence points to a page action, not before.
Disclosure
The case data comes from a private 2026 operating review of a large cross-border marketplace. It is reproduced with permission after company, vendor, domain, system, and personnel identifiers were removed. The figures show association, not controlled causation.
