How to read Search Console without fooling yourself
The dashboard shows four numbers and none of them tells you what to do next.
Separate brand from non-brand before anything else
Brand queries click several times higher than everything else. Leave them in and every rate you compute is a blend of two populations that behave nothing alike. On this site brand was 42.3% of clicks, enough to distort every conclusion.
Use the widest window available, not 28 days
Results settle on a one-to-four month delay, so a 28-day window is a report on work you have already forgotten doing. Sixteen months tells you about your strategy. A month tells you about noise.
Read clicks, use impressions only to diagnose
Impressions are the biggest and most flattering number available and the least connected to outcomes. Their one good use is separating "not seen" from "seen and not chosen".
Compare pages to your own curve, not to benchmarks
Expected CTR at each position, computed from your own data, then each page measured against it. That comparison finds snippet problems in minutes and it is the single highest-value thing the export supports.
The four numbers, and which two lie
Clicks and impressions are counts and they are honest. Click-through rate and average position are derived, and both are averages over wildly heterogeneous populations, which makes them easy to misread and easy to move without anything real changing. Make decisions on the counts. Use the derived pair to generate hypotheses, never to conclude.
The report that changed decisions here
Queries, filtered to positions 5 to 20, sorted by impressions. That view is a list of things you nearly rank for, ordered by how much they would be worth. It is where the cheapest wins on any site live, and it is one filter away from the default view that almost nobody sets.
Measured, not asserted
Every number on this site came from that export: 42,320 clicks, 1,228,913 impressions, 42.3% brand share, across 224 days.
Free tool for this: Search Console Analyzer. No account, nothing uploaded.
Where this goes deeper
Every number on this page comes from one complete dataset: one product taken from zero to 100K+ users on search alone, with nothing spent on advertising. The full argument is Chapter 18 and Chapter 20 of the book. Five chapters are free to read.