Benchmarking is the boring step that decides whether any of the later work can be evaluated. It means writing down where things stand before you change anything: which queries the site appears for, where the pages you care about sit, and what the traffic to them is. Skip it and every subsequent claim of improvement is an argument, because the comparison has no starting point that anybody wrote down at the time.
Baseline the queries, not just the traffic
Total sessions move for a dozen reasons that have nothing to do with your work: seasonality, a campaign, a link from somewhere, a change in the results themselves. A baseline held at query and page level survives all of that, because you can ask a specific question later. Did the page we rewrote for this query move, and did its impressions follow? That is answerable; whether the site is doing better is not.
Write down the conditions, not just the numbers
A benchmark is only comparable if the next reading is taken the same way: same source, same device, same location, same period length. Those four lines beside the numbers are what make the comparison honest three months later, and they are exactly what nobody records because at the time they seem too obvious to write down.
Choose a period long enough to mean something
A month is the usual unit and a fortnight is rarely enough, because normal variation swamps the signal. For seasonal subjects, comparing to the same period last year is the only comparison that means anything, which is a good argument for keeping the record for longer than the tool you are using keeps it. Search Console's Performance report opens on the past three months by default, and a baseline you wrote down outlives whatever window a dashboard happens to show you.
What benchmarking cannot do
It cannot isolate your work from everything else that happened. Search results change, competitors publish, and the ranking systems change too, so a page that held its position through a turbulent quarter may represent a better outcome than one that rose in a quiet one. Benchmarks tell you what happened; attribution is a judgement you should make out loud rather than pretend the numbers made for you.
Questions people ask about seo benchmarking
What should an SEO benchmark include?
Query-level impressions, clicks and average position for the pages you care about, the date and period, and the source and device the readings came from.
How long before I compare?
A month at minimum, three for anything you rewrote. Shorter comparisons mostly measure normal variation.
Do I need a benchmarking tool?
Not necessarily. Search Console holds the data for your own pages; what most people lack is somewhere to keep the reading with its conditions so the next one is comparable.
How does Longtailo help?
It keeps each reading against the keyword, the page, the source and the date, which is the part of benchmarking that gets lost.