How to Benchmark SEO Performance Accurately
A benchmark is what turns a number into information. It takes an afternoon, it cannot be taken afterwards and almost nobody does it, because it happens before there is anything visible to show for the money.
A Number Alone Means Nothing
Visits last month is a fact. It becomes information only when set beside something else. Without that something else a report is a list of true statements that answer nothing. Every figure in this series requires a comparison to mean anything at all.
What a comparison supplies. Context.
Whether a figure is good, bad or unremarkable is not a property of the figure. It emerges entirely from what it is set against, which is why choosing that thing matters more than the number itself.
The three things available. Each different.
Your own past. What you expected. Where you began. Those answer three separate questions and a report should be explicit about which one it is using.
Which one this page is about. Where you began.
The starting point. The other two are covered elsewhere in this series. This one is the only one that expires if you do not capture it.
Why it gets skipped. Nothing to show.
Recording where things stand produces no visible improvement. In the first weeks of an engagement, when everybody wants progress, an afternoon spent writing things down feels like the wrong use of the time.
What skipping it costs. Every later argument.
Without a baseline, nobody can establish what changed. Every subsequent conversation about performance becomes a matter of opinion, which suits nobody.
Record It Before You Start
A benchmark taken at the beginning of an engagement is the only way anybody can later establish what changed. It should be done before anything is altered, with the effort involved amounting to an afternoon rather than a project.
What before means. Before any change.
Not before the results arrive. Before a single page is edited, because the moment something is altered the record describes a site that no longer exists.
Who it protects. Both sides.
The business gets evidence of what it bought. The supplier gets evidence of what they inherited, which matters when a site arrives in poor condition.
What it prevents. The retrospective argument.
Six months in, somebody asks whether things have improved. With a baseline that is a factual question. Without one it is two people remembering differently.
Why an afternoon is enough. It is a record.
Not an analysis and not a strategy. A written note of where things stand, which is why the effort is so small relative to what it protects.
Where it should live. Somewhere permanent.
With the business rather than only with the supplier, because suppliers change and the record needs to outlive the arrangement.
What to ask a supplier. One question.
Ask what they recorded before starting. A firm that recorded nothing has removed everybody's ability to judge the work, including their own.
You Cannot Take It Retrospectively
Some of what should be recorded is not fully available later. That is the difference between benchmarking and every other part of an engagement, because everything else can be caught up on and this cannot.
What expires. Search data has a horizon.
Search reporting holds a limited history. Reach a certain distance back and the detail is simply not there, so a baseline taken late is missing the period that mattered most.
What else goes. The state of the site.
What the pages said, how they were structured and what was wrong with them. Once altered, nobody can reconstruct what was there before.
The third thing. What you appeared for.
The set of searches showing your pages at the outset. Recoverable only within the reporting window and gone entirely beyond it.
Why this argument sounds familiar. It is the same one.
Our site migration material makes an identical case about recording a site before it is rebuilt. Both are the same principle applied at different moments. Both are irreversible.
What a late baseline is worth. Something, not everything.
Better than nothing and permanently compromised. If you are already six months in, record now rather than concluding it is too late.
The consequence for judging results. Unanswerable questions.
Without a baseline, the question of whether the work has been worth it has no factual answer. It only has two accounts, neither of which can be checked.
What A Benchmark Should Contain
Four things, arranged around the three layers this series works through. None requires specialist skill and all of them are written down rather than calculated.
Visibility. What you appear for.
The searches showing your pages, how often and roughly where you sat. Exported and kept, since that export is the part most likely to become unavailable.
Traffic. By page and by type.
How many people arrived, which pages received them and whether they came from search, from your own name or from somewhere else entirely. A single total is not a baseline.
Outcomes. Where they are tracked.
Enquiries, calls, bookings or sales, with whatever attribution exists. Frequently the weakest part of the record. Recording that weakness is itself useful.
The technical state. What condition it is in.
What works, what does not and what is obviously wrong. That establishes what the supplier inherited rather than what they created.
The fifth thing worth adding. The seasonal shape.
When the business is busy and quiet, which our seasonality guide covers. Recorded once, it makes every later report readable.
What ties them together. The three layers.
Visibility, engagement and outcome, which our measurement guide sets out. A baseline covering only the first is the commonest version and the least useful.
Benchmark Against Yourself First
Your own previous performance is the only genuinely like for like measure you have. Everything about the business is held constant except the passage of time, which is exactly the condition a comparison requires and never gets anywhere else.
Why it is uniquely fair. Same everything.
Same trade, same area, same customers, same size, same prices. Nothing about the comparison needs adjusting because nothing about the business differs.
What it lets you ask. A clean question.
Are we better off than we were. No other comparison answers that without introducing something that has nothing to do with you.
The complication. The season.
Comparing against yourself still requires comparing against the equivalent period. Against last month you are comparing two different points in the year.
How far back to look. A meaningful window.
Long enough that ordinary variation averages out. Our trends guide covers choosing one.
What it cannot tell you. Whether you are winning.
Improving against yourself while the whole market improves faster is possible. That is the legitimate case for looking outward, carefully, which is blocks six and seven.
The order to work in. Yourself, then outward.
Establish your own trajectory first. External comparison is a supplement to that rather than a substitute for it.
Industry Averages Are Nearly Useless
Published averages combine businesses with almost nothing in common and present the result as though it described yours. For a local trade the figure is not merely imprecise, it is describing a different kind of business entirely.
What goes into one. Everything.
National operations and single vans, online sellers and appointment based services, businesses in busy cities and quiet counties. Averaged together into one figure.
Why that ruins it. The spread is enormous.
When the underlying values differ by orders of magnitude, the average describes nobody. It sits between two entirely different realities and matches neither.
The second problem. The sample.
These figures are frequently drawn from whoever uses a particular product or answered a particular survey. That is not a representative selection of businesses in your trade.
The third. Definitions differ.
Two businesses measuring the same named thing are frequently measuring differently, so the average combines figures that were never comparable in the first place.
Why we publish none. A standing rule.
Not one appears anywhere on this site, including as an illustration. Any figure we printed would be quoted back as though it described a real business.
What to do when handed one. Ask two questions.
Which businesses were included. How was the thing defined. Neither usually has a good answer.
The narrow exception. Direction.
Whether a whole market is rising or falling can be worth knowing. That is a different claim from a target you should be hitting.
Competitor Comparison, Carefully
Watching competitors is genuinely useful and one half of it is not. Which searches they appear for is observable. What their traffic amounts to is inferred. The difference between those two matters enormously.
What is observable. Where they appear.
Search a term and see who is there. That is direct evidence, it costs nothing and it tells you who you are actually competing with.
What that is good for. Finding subjects.
Areas they cover and you do not. That is a legitimate use and it produces a list of things to consider rather than a verdict.
What is inferred. Their traffic.
Any figure claiming to show how many visitors another site receives is a projection produced by a third party from what it can observe. Nobody outside that business knows their real numbers.
Why that distinction matters. It gets presented as fact.
Competitor figures appear in proposals as though somebody measured them. They are estimates, they carry substantial error and they should never be shown to you as data.
The question to ask. Where did this come from.
If a report shows a competitor's traffic, ask how it was arrived at. The answer establishes immediately whether it is evidence or inference.
The deeper problem with it. Wrong target.
Matching a competitor's traffic is not a business objective. Getting more of the right enquiries is. The two are frequently unrelated.
Re-Benchmark After A Structural Change
A rebuild, a migration or a major content change makes the old baseline incomparable, because the thing being measured is no longer the same thing. A new one should be taken deliberately rather than discovered as a gap later.
What counts as structural. Four things.
A rebuild, a change of address, a substantial reorganisation of the pages or a change in what the business actually sells.
Why the old record stops working. Different site.
Comparing after against before tells you about the rebuild rather than about the work. Those are separate questions and merging them answers neither.
When to take the new one. Both sides.
Immediately before the change and again once things have settled. Our migration material makes the same argument about recording before a rebuild.
What to keep from the old one. All of it.
The previous baseline remains the record of an earlier era. Superseded rather than deleted, since long term direction still needs it.
The commercial change nobody records. A new service.
Adding or dropping a service changes what the site is for. Worth noting alongside the technical changes, because it explains movement that otherwise looks inexplicable.
How often otherwise. Only when something changes.
A baseline is not an annual exercise. It is taken at the start and refreshed when the thing being measured stops being the same thing.
Where everything sits. The hub.
All nineteen guides are on the measuring performance guide.
You cannot go
back and take
it later.
Search reporting holds a limited history, so reach far enough back and the detail is simply not there. What the pages said before they were edited cannot be reconstructed either. That makes a baseline the one part of an engagement that genuinely expires. An engagement without one can never fully establish what it achieved. Every conversation about results becomes two accounts, neither checkable.
What we record before touching anything:
Ask any supplier what they recorded before they started. A firm that recorded nothing has removed everybody's ability to judge the work, including their own.
Every guide.
One question.
What each number actually means, how to read the data without being misled, what belongs in a report and what does not, then how to connect any of it to the enquiries the business is actually paying for.
