What Is the Performance Report in Google Search Console?
The most used and most misread part of the tool. The four headline numbers on their own tell you almost nothing about your business. Everything worth knowing is one filter away. This guide is about which filters those are.
The Totals Tell You Nothing
Four big numbers sit at the top of this report and they are the least useful thing in it. Unsegmented, they mix your customers with people in other countries, your brand name with the work you sell, plus mobile with desktop. Any conclusion drawn from them is drawn from a blur.
Why they get quoted anyway. They are the first thing you see.
They are large, they move and they fit in a monthly summary. That combination makes them the most reported and least informative figures in search.
What is wrong with them. Everything is averaged together.
A total that includes traffic from the other side of the world, searches for your own business name and two devices behaving completely differently is a number describing several things at once.
What the report is actually for. Splitting.
Filtered properly, the same data answers most of the questions a business has about search. That is the whole argument of this page and the reason the sheet calls this report the most misread part of the tool.
Where to start. Two filters, always.
One country and a like for like comparison period. Apply those two before you read a single figure and everything else becomes considerably more trustworthy.
What the unfiltered view is good for. One thing.
Spotting a collapse. If every figure has halved since last month, that is worth seeing immediately. Beyond that single use, the totals are a starting point rather than a finding.
What is not covered here. What the metrics mean.
Our material on measuring performance owns the interpretation of impressions, clicks and average position, including why that last one is the most misread figure of the four. This page owns the report.
The Four Numbers
Clicks, impressions, click through rate and average position. Here is what each covers, kept short deliberately, because reading them together is a subject of its own.
Impressions. You appeared.
Your site showed up in results for a search. Nobody necessarily saw it properly and nobody necessarily wanted it. It is a measure of visibility rather than of interest.
Clicks. Somebody chose you.
The only one of the four that represents a person deciding to visit. If you look at one number, look at this one, then look at what it did over time.
Click through rate. The relationship between the two.
How often appearing turned into a visit. Useful as a comparison against your own pages and meaningless as an absolute, which is why we publish no benchmark figure for it anywhere.
Average position. The most misread of the four.
An average across every appearance, which means it can move for reasons that have nothing to do with your ranking anywhere. Our measuring performance material works through why treating it as your position is a mistake.
How to use them together. As a sequence.
Impressions tell you about visibility, clicks about appeal and the relationship between them about where the problem sits. Blocks six to eight are about finding out which.
Comparing Date Ranges
A number without a comparison is not information. Two thousand clicks is neither good nor bad until you know what it was before. Comparison is therefore the first thing to set up and the thing most frequently done wrongly.
The rule. Like for like.
Equal length periods, comparable in character. Adjacent calendar months break both halves of that rule, which is why so many reported declines are arithmetic rather than performance.
Month lengths. The commonest false alarm.
February against January is three days shorter before anything real happens. That alone can produce a fall of around a tenth, which somebody will then explain with a theory about an algorithm update.
Seasonality. The second one.
A roofer in November and a roofer in June are different businesses. Comparing consecutive months across a seasonal boundary produces a number describing the weather.
What the useful comparison usually is. The same period last year.
It handles both problems at once. The catch is that the tool holds a limited history, so this only works if somebody has been exporting, which our how to use guide covers.
One more discipline. Note what changed.
A comparison showing a difference is the start of a question rather than an answer. What was published, redesigned or removed in that period is what turns it into a finding.
Filtering By Country
A UK business looking at worldwide figures is reading a blend of its actual market and a quantity of noise. Filtering to one country is the single cheapest improvement in the accuracy of everything else in the report.
Where the foreign impressions come from. Several places.
Automated traffic. People searching in other markets for terms your pages happen to match. Content of yours being surfaced somewhere it cannot possibly convert.
Why it distorts so badly. It inflates impressions.
Impressions that will never produce a customer still count, which drags your click behaviour down and makes your average position describe results you do not care about.
What filtering reveals. A smaller, better picture.
Fewer impressions, a healthier relationship between appearing and being clicked, plus a position figure that describes the market you actually sell into.
What to do about the foreign volume. Usually nothing.
It is not a problem to solve and it is not an opportunity either. Businesses occasionally get excited about international demand that is a handful of mismatched searches.
When it is worth a look. A genuine pattern.
Consistent, clicking traffic from one other country over months is different from scattered impressions. That is worth investigating. Our technical material covers what serving another market properly involves.
Filtering By Device
The same page frequently behaves completely differently on a phone and on a computer. A business looking only at the combined figure will never know. This is one of the quicker checks in the report and one of the more revealing.
What differs. Everything, sometimes.
Position, click behaviour and even which searches you appear for. The results themselves are laid out differently, so the same position means something different on each.
Why the total hides it. Averaging.
A page performing well on desktop and badly on mobile produces a middling combined figure that looks unremarkable and describes neither reality.
What a mobile problem points at. Usually the page.
Slow loading, awkward layout or content buried below advertising and menus. Our technical material owns that, including the point that the mobile version is the one being assessed.
What a desktop skew points at. The subject.
Research and comparison happen at desks. Urgent local jobs happen on phones. A split that leans one way tells you something about how people buy from you.
What to do with the finding. Prioritise properly.
If most of your enquiries come from phones, a mobile problem on your key pages is the most valuable thing on this page to fix. That is worth knowing before anybody redesigns anything.
Branded Against Non-Branded
Searches for your business name and searches for what you sell are two completely different things sitting in one report. Mixing them hides everything that matters. Separating them takes about two minutes.
What a branded search is. Somebody who already knew.
They typed your name. They heard about you from a van, a neighbour, a previous job or an advert. Search delivered them rather than winning them.
What a non-branded search is. A new customer.
They typed the trade and perhaps a place. They had no idea you existed. Every one of these is work your website earned.
Why mixing them misleads. Brand traffic flatters.
Name searches convert brilliantly and click at high rates, so a report full of them looks excellent while telling you nothing about whether you are reaching anybody new.
How to separate them. Filter the query list.
Exclude anything containing your business name to see the non-branded picture, then include only those terms to see the branded one. Two views, two very different stories.
What a branded heavy report means. Reputation without reach.
The business is known to the people who already know it. That is worth having and it is not growth. It is also an extremely common picture for established trades with old websites.
What to watch. Non-branded growth.
That is the number telling you whether the work is producing anything. Branded growth is worth having and it usually reflects offline activity rather than search work.
Top Performing Pages
Switching from search terms to pages groups everything by address instead. That answers a different question, being which parts of your site are doing the work. The answer is regularly not what the business expected.
What you usually find. A surprise in the top five.
An old blog post, a technical explainer or a page nobody has touched in three years, sitting above the service pages the business spends its money promoting.
Why that matters. It shows what you are valued for.
The pages people arrive on describe what the market thinks you are good at, which sometimes differs from what you sell hardest. That is strategic information rather than a curiosity.
What to do with a strong page. Use it.
Make sure it leads somewhere commercial, keep it current and look at what else could be built around the subject. A page already earning attention is the cheapest foundation available.
What to do with the weak ones. Ask why they exist.
Pages with almost no impressions are either about something nobody searches for or are not competing at all. Both are worth knowing and the responses are different.
Where the follow up sits. Two guides.
Pages with high impressions and low clicks and pages losing rankings are the two page level findings worth acting on first.
Queries And Pages Together
Filtering to one page and then looking at the searches it appears for is the step that produces most of the content decisions we make for clients. It is also the step almost nobody takes, because both halves are usually read separately.
What it shows. What a page is actually about.
Not what you intended it to be about. The searches it appears for describe how it is understood, which is frequently narrower or stranger than the brief.
The commonest finding. Appearing for something better.
A page turning up for a question you barely addressed, weakly but consistently, is a page telling you there is demand you have not served properly. That is the cheapest content research available.
The second finding. Competing with yourself.
Two pages appearing for the same searches means you have split one subject across both. Usually the answer is to merge them rather than to optimise both.
The third. Mismatched intent.
Plenty of impressions for searches your page does not actually answer, which explains poor click behaviour better than any title rewrite would.
What to do next. One change, then wait.
Using the data to improve content covers turning these findings into work, including why changing several things at once makes the result unreadable.
The Data Has Limits
This is the best search data available about your own site and it is not an audited ledger. Knowing where it is soft prevents arguments and stops decisions being made on precision the report does not have.
Not every query is shown. By design.
Rare searches and anything that might identify an individual are withheld, so the query list will never account for all your impressions. The gap is normal rather than a fault.
Some figures are rounded. Treat them as directional.
Which is why the numbers in a filtered view may not sum neatly to the total above them. Chasing that reconciliation is a waste of an afternoon.
It will not match your analytics. Ever.
Different things counted at different points, with consent and blocking affecting one far more than the other. The comparison guide covers why they are not supposed to agree.
The history is limited. Export it.
Older data falls away, so long term comparison depends on your own record. That is the one habit in this cluster with a deadline attached.
What it still cannot tell you. Whether it was worth anything.
Clicks are not customers and this report knows nothing about revenue. Everything sits on the Google Search Console guide. Our measuring performance material owns the outcome question.
Most reports
quote the four
numbers that
matter least.
Unsegmented totals mix your market with other countries, your own name with the work you sell, plus two devices behaving nothing alike. Filtered properly, the same data tells you whether you are reaching anybody new, which is the only growth question worth asking.
How we read it:
If your clicks fell because February is shorter than January, we will tell you that rather than write you a recovery plan.
Every guide.
One tool.
Getting set up and verified, using the reports properly, the performance and indexing data, finding what is broken, finding what is being passed over and turning any of it into work worth doing.
