How to report SEO without lying to anyone
Including yourself. Most SEO reporting is optimised for looking busy rather than for being checkable.
60 pieces on organic search, paid search and measurement. Each one is about a decision we have watched people get wrong — why the obvious answer is wrong, and what the number actually responds to.
Rankings, crawling, and the organic half of the picture.
Including yourself. Most SEO reporting is optimised for looking busy rather than for being checkable.
It is four lines long and it can remove a site from search without producing a single error.
Nothing, for a while. Here is what that period actually looks like and what is worth doing in it.
A page that already ranks at 12 is a shorter path to traffic than a page that does not exist.
A tie-breaker in search, and a direct cause of abandonment. Optimise for the second reason.
It is not a ranking factor. It is an eligibility requirement for features that change how your result looks.
There is something worse: dilution, and Google quietly picking a version you did not choose.
Proximity, prominence and relevance — and only two of them respond to anything you do.
Most migration damage comes from four things, and all four are checkable before launch.
No outreach, no approval, no budget. The pages you already own decide which of your pages matter.
Impressions hold, clicks fall, and average position looks fine throughout. What to actually do about it.
It is a statement about which pages matter. Filling it with redirects and 404s makes the statement worthless.
A canonical is a suggestion, not an instruction. Most canonical problems come from expecting obedience.
Truncation is the visible symptom. Putting the wrong words first is the actual mistake.
It is real, it matters at scale, and it is invoked constantly by sites nowhere near the threshold.
One line, invisible in the browser, and it removes a site from Google entirely. Here is how it survives review.
The most common shape in Search Console, and it is almost never a ranking problem.
Silently, irreversibly, and no export fixes it after the fact. The only useful time to act is before.
It moves for reasons that have nothing to do with whether anyone found you, and stays still through changes that matter.
Paid search: structure, bidding, budget and waste.
Not the metrics. Three questions, in order, and most reports answer none of them.
Letting a rule change bids in an automated account produces two systems fighting. Letting it tell you things is invaluable.
Smart bidding adapts, slowly. The tools for telling it what is coming exist and are barely used.
It is legal, it is expensive, and it works in a narrower set of circumstances than people assume.
There are no keywords in Shopping. There is a product title, and it does the same job.
It targets people who were already going to convert, then takes credit when they do.
You cannot A/B test ads properly any more. Here is what still works.
Both are usually applied from intuition. Both have data that contradicts the intuition.
It optimises for cheap leads, which is a different objective from profitable ones.
Fewer, larger campaigns beat many small ones, for a reason that is about data rather than tidiness.
They are not measuring the same thing, and reconciling them exactly is a waste of an afternoon.
The honest answer is: sometimes, and you can find out for your own account in about three weeks.
The list you built at launch stops working within a season. Here is a routine that keeps it honest.
Chasing the number is a waste. Reading its three components tells you exactly what is wrong.
It works. It also removes most of the evidence you would use to know why.
A capped campaign is not just spending less. It is spending differently, and worse.
Most paid search underperformance is a landing page problem wearing an ad account costume.
Keywords are what you asked for. Search terms are what you bought. They are not the same list.
Every optimisation downstream of a broken conversion action makes things worse with more confidence.
It is not reckless and it is not magic. It is a bet that your conversion data is good enough to steer it.
GA4, Tag Manager, and getting the measurement right first.
Pull reporting fails because it depends on somebody remembering. Push the exceptions instead.
Nine checks, in order. Every one of them has silently invalidated somebody's reporting.
Four tools, in order, and most people stop at the first one, which is the least reliable.
Rename one and you split its history in two, permanently. Decide once, in writing.
Blended data, mismatched date ranges and default sampling produce confident, incorrect dashboards.
An approximate value beats no value by an enormous margin, and you can derive one in an afternoon.
Defining an internal traffic filter does nothing. Activating it is a separate step people miss.
Configuring the domains is the easy part. The link that drops the parameter is the part that breaks it.
The standard reports are deliberately shallow. Everything people miss about Universal Analytics is in the other tab.
It improves data quality and page speed. It does not make consent unnecessary, and it is not free.
The test for every chart: what would I do differently if this number were different?
It tells you the number of successes and nothing about the failures, which is where all the improvement is.
Case sensitivity, inconsistent naming, and one missing parameter are enough to fragment a channel into unreadable pieces.
GA4 defines it as the inverse of engagement rate, which makes it a different number with the same name.
The setting is two months out of the box, most people never change it, and it cannot be applied retrospectively.
A checkout that routes through a third party can silently attribute every sale to that third party.
A container is a delivery mechanism. Installing one and stopping is the most common half-finished job in measurement.
If most visitors decline, most of your data is modelled. That is fine, as long as you know which parts.
Standardise event names across every site you run, or give up on comparing them.
It is more common than anyone admits, it is nearly invisible, and it quietly makes every report arguable.
The free audit runs on public data in about ten seconds. No account, no card.
Audit my siteStart telling clients what to change — and prove it moved.
Get started