How to read the Search Console Performance report

The Performance report in Google Search Console is the only place where Google tells you, in its own numbers, how often your pages appeared in search results and how often people clicked. It is also easy to misread. Four metrics, two main dimensions, a two-day delay and a privacy filter combine into charts that look simple and hide a lot. This guide walks through what each number really counts and how to use the report to diagnose real problems.

The four metrics, precisely

Metric	What it counts	Common misreading
Clicks	Clicks on a Google search result that took the user to your site. Clicks inside Google features that do not open your site are not counted.	Expecting it to match analytics sessions. It will not: analytics misses visitors who block scripts and counts visits from other sources.
Impressions	The number of times a link to your site appeared in results the user could have seen. A result on page 2 counts only if the user opened page 2.	Treating it as "people who saw the link". An impression is logged even if the result sat below the fold and was never scrolled to.
CTR	Clicks divided by impressions.	Comparing CTR across queries with very different positions. A 2% CTR at position 9 can be healthy; at position 1 it is a problem.
Average position	The topmost position your site held in each impression, averaged across impressions.	Reading it as "my ranking". It is an average over many searches, devices and locations. See why average position misleads.

Query view versus page view

The report can group data by query or by page (and also by country, device, search appearance and date). The two views do not add up to the same totals, and that surprises almost everyone the first time.

• By page, each URL that appeared gets its own impressions. If two of your pages showed for the same search, the page view counts two impressions.
• By property (the chart totals at the top), that same search counts as one impression for the site, at the position of the best-placed URL.
• By query, some queries are hidden. Google omits rare queries to protect user privacy (often called "anonymized queries"). Their clicks and impressions are still included in the chart totals, but not in the query table. On sites with a long tail of unusual searches, the query table can account for well under the full total.

Practical consequence: never add up the rows of the query table and compare them with the chart. Use the chart totals for trends and the tables for "which queries and pages are driving the trend".

The data delay and the last few days

Search Console data is not real time. The latest complete day is usually about two days behind today, and the most recent days can still be revised upward as Google finishes processing. If you compare "last 7 days" on a Monday morning, the final day or two may look like a drop that is not real. When a trend matters, end the comparison window three days back, or wait and look again.

DeskSEO pulls the same numbers through the Search Console API and re-fetches the last five days each morning for exactly this reason, so a provisional low day gets corrected rather than frozen.

Comparing periods without fooling yourself

The Compare tab in the date filter is the fastest way to answer "what changed". A few rules keep it honest:

• Compare equal lengths and the same weekdays. 28 days against the previous 28 days keeps four Mondays on each side. Comparing a week containing a public holiday with a normal week will show a drop that is just the calendar.
• Use year-over-year for seasonal businesses. A garden supply shop will always lose impressions from October to December. Compare against the same weeks last year.
• Sort the comparison table by the difference column, not by current clicks. The rows with the largest negative difference explain most of a drop.

Filters that answer real questions

• Query contains your brand name versus does not contain: separates people who already know you from people who found you. Brand search usually moves with marketing and press; non-brand search moves with SEO work.
• Page equals one URL, then switch to the query tab: shows every query that page appears for. Useful before rewriting a title, because it shows what the page is already relevant to.
• Device: mobile versus desktop: a drop on one device only points at a layout, speed or interstitial problem on that device, not a content problem.
• Country: if you serve several markets, a drop in one country after adding hreflang or changing a domain is far easier to spot with this filter than in the totals.
• Search appearance: shows rich result types such as review snippets or videos. If one of these disappears, structured data broke or stopped qualifying.

Regular expression filters (the "Custom (regex)" option) are worth learning. For example, ^(how|what|why|can|should)\b on the query dimension isolates question-style queries, which are often the ones AI answers and featured snippets compete for.

Diagnosis patterns

Impressions up, clicks flat

The site is showing for more searches, but the new impressions are low on the page or on searches where people do not click. Check the average position of the new queries: if they sit at 20–60, this is early visibility, not a problem. If impressions grew on queries where you already ranked well, look at the results page itself: an AI Overview, a video carousel or more ads above you can absorb clicks without your position changing.

Position improved, clicks went down

Usually a mix effect. Losing many impressions at poor positions raises the average without anything ranking better. Check impressions: if they fell, the "improvement" is an artifact. If impressions held and clicks still fell, compare the title and snippet Google shows now with what it showed before; Google sometimes rewrites titles, and a rewrite can be less appealing.

Everything dropped on one date

A sudden step down across all pages on a single day is rarely a ranking change. Check, in this order: whether the site was unreachable or returned errors (the Pages and Crawl stats reports), whether a deploy added noindex or changed robots.txt, and whether the property lost data because the site moved host or protocol. The free SEO checker catches the first two in a minute.

One page lost most of its clicks

Filter by that page and compare periods on the query tab. If the same queries are still there with a worse position, a competitor or a SERP feature moved above you. If the queries now point to another of your URLs, you have two pages competing; consolidate them and set a clear canonical (see canonical tags).

High impressions, CTR far below neighbours

Sort pages by impressions, then look for rows with a position of 1–5 and a CTR far below what nearby rows achieve. Those pages are seen but not chosen. Rewrite the title and description to match the intent of the top queries (see titles and meta descriptions).

Limits worth knowing

• The interface shows at most 1,000 rows per table. The API returns far more, which is why tools that read the API can show deeper query lists.
• Data is kept for 16 months. If you need longer history, it has to be exported and stored somewhere else before it ages out.
• Discover and Google News have their own reports and are not part of the web search numbers described here.

Getting the data out of the interface

The interface is fine for a quick look but tedious for watching a dozen sites. DeskSEO reads the same Performance data through a service account that you add as a Restricted user, which can read reports but cannot change settings or submit anything. The step-by-step setup is on the Search Console connection page; after that, daily clicks, impressions and average position per query are stored and kept for the full 16 months.