What are the four Performance metrics?

Clicks, impressions, click-through rate, and average position. They are simple individually and misleading in combination, because each one moves for reasons that have nothing to do with the other three. Most reporting mistakes come from reading one and inferring the others.

My position improved but traffic did not. Why?

Because average position is an average across every query and every impression, so it can improve while the queries that actually convert stay where they were. Position four on a term nobody searches and position eight on your money term average to something that looks like progress.

Click-through rate went up. Does that mean the SEO is working?

Not on its own. Click-through rate rises when impressions fall on queries you were never going to win, which is arithmetic rather than improvement. Read it alongside clicks: rate up with clicks flat usually means you lost the impressions that were dragging the average down.

Impressions are up sharply and clicks are flat. What happened?

Usually you started appearing for a broad term at a position nobody clicks, or a feature above the results absorbed the click. Both look like growth in the chart and produce nothing. Segment by query to see which impressions arrived, and whether any of them were ever buyers.

We ranked first yesterday and seventh today. Is Google broken?

No. Rankings are computed per query, per location, per device, so the number you see is an average over a moving population of searches. Day-to-day movement of several positions on an average is normal and reading it as a change in your standing is the most common self-inflicted panic.

We should rank for a term and do not. Is that a Google error?

Almost never. It usually means the page does not answer that query as directly as the pages that do, or you have no page targeting it at all, or the intent behind the term is different from what you assumed. The results page tells you which, and it takes two minutes to look.

How should I compare two time periods?

Same length, same days of the week, and mind the seasonal shape of your market. Comparing a four-week period against the previous four weeks is honest; comparing this month against last month when one has an extra weekend and a holiday is not, and the difference is often larger than the change you are measuring.

Why separate branded and non-branded queries?

Because branded search measures demand you already created and non-branded measures search working. Blended together, a marketing campaign that raised brand awareness looks like SEO success, and genuine SEO progress can be invisible under a branded decline.

How do I diagnose a click drop?

Find the date, then check in order: did indexing change, did specific queries lose clicks, did impressions fall or only clicks, and what does the results page look like now. That sequence separates a technical problem from a competitor from a change in how the results are presented.

Should I look at country and device breakdowns?

Yes, because both hide problems in the average. A site performing well overall while failing on mobile is common, and traffic from countries you do not serve inflates impressions and depresses click-through rate, making your real performance look worse than it is.

How do I tie this data to revenue?

By tracking which queries lead to the pages that produce enquiries, and instrumenting those enquiries properly. Search Console stops at the click. Everything after it has to come from your analytics and your own records, and the join is what turns a ranking report into a business case.

When should I export the data?

When you need more than the interface shows, meaning more rows, a longer window, or a join with another dataset. The interface truncates and samples, so any analysis that depends on the long tail of queries needs the export rather than the screen.

Search Console SEO How-to

Google Search Console Performance Report: Reading the Data Without Lying to Yourself

Most teams misread the Performance report the same five ways. Here’s the honest read.

John Cravey with AIFounderApr 25, 20265 min readUpdated Jul 6, 2026

The GSC Performance report is the most useful data in SMB SEO. It’s also the most misread. Most agencies present it to clients with a “see, traffic is up” headline that doesn’t hold up under scrutiny. Here’s how to read it like a professional, with the five misreads we see most often.

Free estimate · 2 minutes

Your Search Console data already shows the gap.

The estimate doubles as intake for an audit: we read your data, show where the clicks are being lost, and price the fix.

#The four metrics

Clicks: actual visits from Google search.
Impressions: times one of your URLs appeared in Google’s results, regardless of click.
CTR: clicks ÷ impressions, expressed as a percentage.
Average position: average rank of your URLs across the queries they appeared for.

#Misread #1: ‘position improved, traffic should follow’

Average position is an average across every query your site appears for. If your site started ranking for 50 new low-position queries (positions 30-90), the average position metric gets worse even though you’re ranking for more terms. Same in reverse: dropping low-position queries makes the metric look better but means less surface area. Treat average position as a directional signal for specific queries, not a portfolio-level number.

#Misread #2: ‘CTR went up, the SEO is working’

CTR varies by position. Position 1 has a 25-30% CTR. Position 10 has a 2-3% CTR. If your average position drops from 8 to 5, your CTR will go up regardless of whether anything else changed. Always look at CTR per query at a specific position, not aggregated CTR.

#Misread #3: ‘impressions are way up but clicks are flat’

Three possible causes: (1) you’re ranking for new informational queries that get AI Overviews and zero-click; (2) your title/description rewrites aren’t compelling at the position you’re ranking; (3) you’re ranking deeper for queries you’ve always been near the top of. Each cause needs a different response. Look at the specific queries with the biggest impression growth and see which bucket they fall into.

#Misread #4: ‘we ranked #1 yesterday but #7 today, Google is broken’

Position data in GSC is an aggregate across the day for users in different locations. A query that ranked #1 in San Francisco and #15 in Houston averages out to something in between. Individual user position varies by intent signals, location, search history. Don’t obsess over single-day position fluctuations.

#Misread #5: ‘we should rank for X — we don’t — therefore Google is wrong’

If GSC shows you don’t rank for a query you think you should, look at the SERP. Who’s ranking #1-10? Are they bigger brands? More authoritative? Better content? Older domains? Local entities with stronger location signals? Whatever you’re looking at, that’s what Google is rewarding. The conclusion is usually “my content needs to be better than theirs,” not “Google is wrong.”

#The right way to compare two time periods

GSC’s built-in compare feature lets you put last 28 days next to previous 28 days. Watch for:

Queries that newly appeared in the current period — what new content or links explain this?
Queries that dropped out — what happened? Algorithm update? Competitor invested? Page got deindexed?
Pages with the biggest click change in either direction — investigate page-by-page.
CTR changes on queries you’ve ranked stably for — usually a title/description rewrite or a SERP feature change.

#Reading branded vs non-branded separately

Filter by query containing your brand name → branded traffic. Filter by query NOT containing → non-branded. Always look at both. Branded traffic up could mean a PR mention or paid campaigns; non-branded up means SEO is bringing in new buyers. The two have different action implications.

#Looking at country and device

Filter by country and device for the same queries. A query ranking #3 on desktop USA might be ranking #8 on mobile, where most of your audience is. Filter by mobile + USA to see what most users are actually experiencing.

#Diagnosing a click drop

Filter by date range that includes the drop. Note the exact day clicks dropped.
Check the industry trackers for a Google update on that date.
Compare top queries before and after the drop. Did specific queries lose ranking? Or did all queries lose roughly equal ranking?
If specific queries: those topics got hit. Look at the SERP for those queries today — who took your spot?
If all queries: site-wide signal — usually a Core Update, a manual action, or a technical issue.

#What to do with the data

Every weekly review should produce one of: (1) a content edit (rewriting a title or meta description for a high-impression-low-CTR page), (2) a new content idea (a query showing up in impressions where you don’t have a perfect-fit page), (3) a technical fix (a CWV regression, an indexing issue), or (4) nothing — sometimes the data just shows everything’s fine.

#Exporting data for deeper analysis

GSC’s built-in UI hides queries beyond the top 1000 per filter. For deeper analysis, use the GSC API or the GSC integration in GA4 to pull the full dataset. Most SMB-tier engagements don’t need this; once you’re analyzing tens of thousands of queries, the API is worth setting up.

#Tying GSC to revenue

GSC shows visits, not revenue. Tie GSC to your GA4 conversion data to see which organic landing pages actually convert. A page with 5,000 clicks and 0 conversions is worth less than a page with 200 clicks and 30 conversions. The conversion-weighted view is the right way to prioritize SEO investment.

#How this lands across FH client work

We pull a GSC weekly read for every FH SEO engagement, package it into a one-page Monday-morning brief, and use the brief to drive that week’s work. The brief takes 15 minutes to produce. The week-over-week discipline is what makes case studies like BHR compound. If you’re running SEO without a weekly GSC read, book a consultation — it’s the single highest-ROI process change in SMB SEO.

Answers

Frequently asked questions

What are the four Performance metrics?: Clicks, impressions, click-through rate, and average position. They are simple individually and misleading in combination, because each one moves for reasons that have nothing to do with the other three. Most reporting mistakes come from reading one and inferring the others.
My position improved but traffic did not. Why?: Because average position is an average across every query and every impression, so it can improve while the queries that actually convert stay where they were. Position four on a term nobody searches and position eight on your money term average to something that looks like progress.
Click-through rate went up. Does that mean the SEO is working?: Not on its own. Click-through rate rises when impressions fall on queries you were never going to win, which is arithmetic rather than improvement. Read it alongside clicks: rate up with clicks flat usually means you lost the impressions that were dragging the average down.
Impressions are up sharply and clicks are flat. What happened?: Usually you started appearing for a broad term at a position nobody clicks, or a feature above the results absorbed the click. Both look like growth in the chart and produce nothing. Segment by query to see which impressions arrived, and whether any of them were ever buyers.
We ranked first yesterday and seventh today. Is Google broken?: No. Rankings are computed per query, per location, per device, so the number you see is an average over a moving population of searches. Day-to-day movement of several positions on an average is normal and reading it as a change in your standing is the most common self-inflicted panic.
We should rank for a term and do not. Is that a Google error?: Almost never. It usually means the page does not answer that query as directly as the pages that do, or you have no page targeting it at all, or the intent behind the term is different from what you assumed. The results page tells you which, and it takes two minutes to look.
How should I compare two time periods?: Same length, same days of the week, and mind the seasonal shape of your market. Comparing a four-week period against the previous four weeks is honest; comparing this month against last month when one has an extra weekend and a holiday is not, and the difference is often larger than the change you are measuring.
Why separate branded and non-branded queries?: Because branded search measures demand you already created and non-branded measures search working. Blended together, a marketing campaign that raised brand awareness looks like SEO success, and genuine SEO progress can be invisible under a branded decline.
How do I diagnose a click drop?: Find the date, then check in order: did indexing change, did specific queries lose clicks, did impressions fall or only clicks, and what does the results page look like now. That sequence separates a technical problem from a competitor from a change in how the results are presented.
Should I look at country and device breakdowns?: Yes, because both hide problems in the average. A site performing well overall while failing on mobile is common, and traffic from countries you do not serve inflates impressions and depresses click-through rate, making your real performance look worse than it is.
How do I tie this data to revenue?: By tracking which queries lead to the pages that produce enquiries, and instrumenting those enquiries properly. Search Console stops at the click. Everything after it has to come from your analytics and your own records, and the join is what turns a ranking report into a business case.
When should I export the data?: When you need more than the interface shows, meaning more rows, a longer window, or a join with another dataset. The interface truncates and samples, so any analysis that depends on the long tail of queries needs the export rather than the screen.

Question we did not answer? Ask us directly and we will answer it here.

Tagged#Search Console #SEO #How-to

Written by

John Cravey

Founder

Founder of Frontend Horizon. Writes most of the long-form work on the FH blog.