Google Search Console Performance Comparison

Organic traffic is down 18%. That sentence starts every difficult SEO conversation and contains no information about what happened.

An 18% decline could be one high-traffic page losing its ranking, a hundred pages each slipping two positions, a seasonal pattern that happens every year, a SERP feature absorbing clicks at unchanged positions, or a tracking change that has nothing to do with search at all. These require completely different responses, and the aggregate figure cannot distinguish between them.

Using Google Search Console's own comparison export at the query or page level replaces the symptom with a diagnosis. Usually the explanation is concentrated in a surprisingly small number of rows.

Most bad conclusions come from bad comparisons rather than bad analysis. Four rules prevent nearly all of them.

• Equal period lengths. Twenty-eight days against twenty-eight days, not a month against six weeks.
• Day-of-week alignment. Search behaviour is weekly. Comparing periods that contain different numbers of weekends introduces a difference that is not real.
• Consistent filters. Same country, same device, same search type on both exports. Mixing them produces movement that exists only in the filter.
• Mind the data lag. Search Console data is incomplete for the most recent two to three days. Including them will always show a decline.

You lost rankings. This is the most straightforward diagnosis. Now determine scope: a handful of pages points at page-level causes — content changes, lost links, technical problems on those URLs. Broad decline across many pages points at something sitewide, such as an algorithm update or a technical change affecting the whole site.

You are still appearing as often but being clicked less. Almost always the results page changed around you: an AI Overview appeared, a competitor gained rich results, an extra ad block pushed organic down. Your position number may be identical while your actual visibility is lower.

Usually new low-position visibility. You are appearing for more queries, but at positions where clicks were never likely. Not a problem, and often an early sign that new content is starting to gain traction.

Counterintuitive but common. You gained visibility across many low-position queries, which drags average position down while adding traffic. This is why average position is a poor headline metric.

Both views answer different questions and you need both.

Query comparison tells you what changed in demand and in your coverage of it. New queries appearing, old queries disappearing, and shifts in which terms drive traffic.

Page comparison tells you where the change landed. A single page losing half its clicks is a page problem; fifty pages each losing a tenth is a sitewide one.

Run the query view first to understand the shape, then the page view to locate it. Doing it in the other order tends to produce a long list of small movements without a story.

• Algorithm update — broad decline across many pages starting on a specific date, positions down, impressions down.
• Technical problem — sharp drop, often on a subset of URLs, sometimes with a coverage error appearing in Search Console at the same time.
• Lost featured snippet — one query, large click loss, position roughly unchanged.
• Seasonality — gradual, matches the same period last year, spread evenly across queries.
• Competitor movement — specific queries, position down, everything else stable.
• Cannibalization — position unstable and the ranking URL for a query has changed between periods.
• Migration fallout — decline beginning on the launch date, often with new 404s appearing.

The output of this analysis should be three lists, not a spreadsheet.

1. What changed — the specific queries and pages responsible for most of the movement, with numbers.
2. Why, as far as the data shows — the pattern each one matches, stated with appropriate confidence rather than certainty.
3. What to do — the action for each, ordered by the size of the opportunity.

Being honest about uncertainty matters here. "Position held while clicks fell, which is consistent with a SERP layout change" is more credible and more useful than an unqualified claim about an algorithm update.

Beyond the metric movements, a period comparison sorts every query into three buckets, and each one tells you something different.

Terms receiving impressions in the current period but not the previous one. These show what is starting to work. Recently published content usually appears here first, at low positions with few clicks, weeks before it produces meaningful traffic. It is the earliest available signal that a content investment is landing.

Terms that had impressions before and have none now. Worth checking individually rather than in aggregate. A high-value query disappearing entirely usually means a page was removed, deindexed, or dropped out of the top hundred — all of which are specific, findable problems.

The largest bucket and where most of the traffic change lives. A query present in both periods with a meaningful click difference is the clearest signal available, because nothing about its existence changed — only its performance.

A migration makes period comparison harder because URLs change, so a page-level comparison shows every old URL as lost and every new one as new. The fix is to compare at query level instead, where the terms remain constant regardless of which URL serves them.

Run the query comparison, identify queries that lost clicks, then check which URL now ranks for each. Redirect problems show up as queries where the ranking URL is a redirect target that does not match the intent, or where nothing ranks at all.