---
title: Which Search Console Numbers Actually Matter
description: How to read the Search Console performance report without chasing numbers that mean nothing, and what to fix once you spot a real problem.
date: 2026-09-23
updated: 2026-09-23
tags: [seo wednesday, search console, measurement]
url: "https://drivenhost.com/blog/search-console-numbers-that-matter"
author: DrivenHost
---

Most site owners open Search Console, glance at the big blue graph, and close the tab. That is a reasonable reaction. The default view is a line moving up or down with no explanation attached to it. Underneath that line are four numbers, and once you know which one lies to you, the report stops being a wall chart and turns into a short list of jobs for the week. This is how we read it when a customer asks us why their traffic moved.

## Clicks first, everything else second

Clicks are people who saw your site in Google's results and came to you. That is the number tied to enquiries and orders, so it is the one to open with. Set the date range to the last three months, turn on the comparison with the previous period, and look at the direction of travel rather than any single day.

Impressions count the times a page of yours appeared in results at all. CTR is clicks divided by impressions. Average position is where your listing sat, averaged out. Four numbers, three of which are only useful once you have filtered them down to something specific.

## Average position is the number that fools people

Sitewide average position is an average across every query and every page you appear for. A single obscure phrase where you rank second can pull it one way; twenty new phrases where you rank fortieth can pull it the other, while your actual traffic does not change at all. We have watched people celebrate a move from 18.4 to 17.6 that meant nothing whatsoever.

Position is worth reading in one situation only: filtered to a single page, or a single query, over a decent stretch of time. "Our pricing page went from 14 to 8 for this phrase" is a real event. The site-wide figure is noise with a decimal point on it.

## Impressions with no clicks usually means the snippet is wrong

Here is the filter that earns its keep. Sort by impressions, then look for rows with plenty of impressions, a position inside the top ten, and a click count near zero. Google is showing you to people and they are picking somebody else.

Nine times out of ten that is a title and description problem rather than a ranking problem. The title does not match the wording of the query, or it leads with your company name instead of what the page does, or the description was auto-generated years ago and reads like filler. We wrote about fixing those in [titles and descriptions worth rewriting](/blog/titles-descriptions-fix), and it remains the cheapest work in SEO. You are editing two lines of text on a page that already ranks.

## Find the pages sitting just off the first page

Filter position to roughly 8 through 20 and you get a list of pages Google already considers reasonable answers. Nobody sees them, because almost nobody scrolls. These are better prospects than anything you could write from scratch this month.

Pick two or three, open each one, and ask whether it actually answers the query it is showing up for. Often it half-answers, buried under an introduction. Tighten the opening, add the detail that is missing, give it a heading that matches how people ask the question, then point a link at it from a page that already does well. Our notes on [internal linking for small sites](/blog/internal-linking-small-sites) cover that last part.

## Check the boring explanations before you rewrite anything

Search Console runs a couple of days behind, so a cliff at the right-hand edge of the graph is usually the data catching up rather than a disaster. Check which property you are looking at too, because a URL-prefix property for the www version will not show the non-www one. A domain property covers the lot.

Then check the server side, since we see this often enough to mention it. Downtime during a crawl, a redirect chain left over from a migration, a certificate that lapsed for a weekend, a database query that pushed response times into several seconds: all of them show up as a traffic dip that looks like a ranking problem. Our [status page](/status) records platform-wide events, and your access logs will tell you the rest.

Work through it in that order and most "we lost our rankings" panics turn out to be a snippet nobody clicks, a slow page, or a property showing half the picture. If you want a second pair of eyes on the hosting side of it, send us the page and the dates and [our support team](/support) will check the logs with you.