How to Measure Whether Your Ecommerce SEO Is Actually Working
Reporting is not fixing. In ecommerce there is one split that matters more than every other number combined. A report that does not make it can show healthy growth while the work has achieved nothing at all.
What Actually Needs Measuring
Revenue from non-brand organic traffic. Money, from people who found you in search without searching for your name. Everything else on this page exists to support that one figure or to stop it being misread.
Why that specific measure. Because each word in it is doing work.
Revenue, because traffic that does not buy has cost you money to acquire and returned nothing.
Non-brand, because demand for your own name would have grown anyway, which is block two.
Organic, because otherwise you are measuring what advertising did.
Why everything else is secondary. Not worthless, secondary.
Rankings, impressions, sessions and indexed pages are all useful. Every one of them can improve while this figure stays flat, which makes them diagnostic rather than conclusive. They tell you whether something is happening. This tells you whether it was worth paying for.
What this measure will not do. Move early.
It is the last thing to change, which is why block seven exists. A report in month two showing nothing here is not evidence of failure. A supplier who cannot explain that is measuring the wrong things.
Brand Against Non-Brand
Searches for your own name and searches for what you sell are different things arriving through the same channel. A report combining them can show growth that had nothing whatever to do with the work.
Why brand demand grows on its own. Almost everything you do feeds it.
Advertising, social, word of mouth, packaging, a returning customer, a van with your name on it. All of that produces people who later search your name, arrive through organic search, then appear in the same report as the work you commissioned.
What that means for a combined report. It flatters everybody.
A shop increasing its advertising will see organic revenue rise, because advertising creates brand searches. A supplier reporting the combined figure has produced a chart showing their work succeeding, using demand somebody else generated. Frequently nobody involved intends this.
How to separate them. Simply. It takes an afternoon once.
Build a list of everything containing your shop name and its common misspellings. Everything on that list is brand. Everything else is non-brand. Then report the two separately, permanently.
What the split reveals. Three situations, each meaning something different.
Non-brand rising means the work is reaching new people, which is the point. Brand rising while non-brand is flat means something is building awareness and it is not this. Both flat means nothing is working, whatever the combined total says.
Why this is also a supplier test. Ask whether they separate them.
A supplier who does not is either unable to demonstrate their own contribution or has chosen not to. Per block eight, it is the first thing to look for in any report you receive.
Revenue Rather Than Traffic
Traffic is the easiest thing to report and the least useful thing to know on a shop, because a shop has an outcome that traffic is only a step towards.
Why traffic misleads on a shop specifically. Visitors are not equal.
A thousand people arriving from a guide about how to choose something are worth a fraction of two hundred arriving on a category page ready to buy. A report showing traffic up says nothing about which of those happened.
How a traffic figure can improve while the shop does worse. Frequently. It looks like success.
Content reaching people earlier in their thinking increases visits and reduces the proportion who buy. Both numbers move, one looks good. Revenue is unchanged.
Why margin beats revenue as the numerator. The refinement worth making.
Revenue treats every sale as equal when they are not. A shop whose organic revenue grew entirely in its thinnest margin category has gained less than the figure suggests, possibly nothing after the cost of fulfilling it.
Why most shops still report revenue. Because margin data is harder to attach.
That is a reasonable compromise, provided somebody looks at which categories the growth came from, which is block five.
What traffic is still good for. Diagnosis, per block seven.
Attribution, And Why Nothing Agrees
Your analytics, your shop's own order data and any advertising account will report different numbers for the same period. All three can be right, so knowing which to trust for which question prevents a great deal of wasted argument.
Why they differ. They are answering different questions.
Analytics reports sessions it observed and attributes each sale to a source using its own rules. The shop reports orders it actually took. An advertising account reports sales it believes it influenced, using rules set to give itself credit where it plausibly can.
The specific reasons the numbers move apart. Four, all ordinary.
Visitors who block tracking are missing from analytics and present in the shop. A sale after several visits from different sources gets assigned differently by each system. Advertising platforms count a sale that happened days after a click. And orders cancelled or refunded may be in one figure and not another.
Which to trust for which question. The useful part.
For how much you actually sold, the shop's own order data, always. It is the only system that knows what money arrived.
For where visitors came from and how they behaved, analytics, since the shop does not record that.
For whether advertising is working, the advertising account, while understanding it is marking its own work.
The rule that follows. Pick one source per question and keep it.
Consistency matters more than accuracy here. A figure measured the same way every month shows change reliably even if it is not the truest number available.
Reporting At Category Level
A site total conceals almost everything interesting. Reporting by category is the single change that turns a report into something you can act on.
What a total hides. Movement in opposite directions.
Organic revenue flat across the shop can mean nothing happened. It can equally mean two categories doubled while three declined, which is a completely different situation requiring completely different action.
What category level shows. Where the work landed.
Categories that have been worked on should move before the rest. If they have not, either the work was wrong or the diagnosis was. That is visible within a quarter rather than at the end of the year.
What decisions it enables. Three, none available from a total.
Which categories to work on next, since one that responded quickly usually has more available. Which to stop working on, since one that has had attention and moved nothing is telling you something. And which are quietly declining while the total looks fine.
The margin connection. Per block three.
Category level reporting is also how a shop reporting revenue rather than margin can still tell whether the growth was worth having, by looking at which categories produced it.
What this costs to set up. An afternoon, once.
Grouping pages into their categories is configuration rather than ongoing work. It is the change we recommend before any other reporting improvement.
Adjusting For Seasonality
Comparing this month to last month in a shop with any seasonality produces conclusions that are entirely about the calendar. This is the most common reporting error in the sector.
The false positive. Growth that would have happened anyway.
Work beginning before a peak produces a chart nobody should be pleased with, because almost all of the rise is the season. Everybody is pleased anyway. An expectation forms that January cannot meet.
The false negative. The same shop afterwards.
Revenue falls against the peak and the report shows a decline. Nothing stopped working. The season ended.
Which does more damage. The false positive, reliably.
A shop that credited the season to the work now believes the work is failing. That conversation is considerably harder than the one that would have been avoided by measuring properly in the first place.
The only reliable comparison. The same period last year.
December against December. That removes the calendar and leaves what actually changed, which our seasonal guide covers from the planning side.
What to do in the first year. When no comparison exists yet.
Report non-brand impressions, which are less distorted by the season, alongside revenue with the seasonal caveat stated plainly rather than buried. Then record everything, so the second year has something to compare against.
Leading Indicators
Revenue is the last thing to move. Four measures move before it. Reporting them is how a client sees work happening without anybody having to overstate what it has produced.
Indexed pages. The earliest of the four. The most fundamental.
How many pages are eligible to be found at all. This moves within weeks of technical work, so if it has not moved after a quarter that is the earliest reliable warning available.
Non-brand impressions. How often you are shown for terms that are not your name.
This rises before positions do, because a page appearing on the second results page is being shown without being clicked. It is the best single early measure there is.
Category level visibility. Per block five, whether the categories worked on are gaining.
The spread of terms. How many different searches the shop appears for.
A shop improving picks up a widening set of terms before it wins any particular one, which is invisible if you are watching a keyword list.
What to report in month two. These four, with revenue stated as unchanged.
That is a report showing genuine progress without claiming a result. Per our guide to timescales it is the shape of the first quarter on almost every shop.
What A Good Report Contains
Five things, none of them a list of rankings.
Non-brand organic revenue, separated. Per blocks one and two.
The same figures at category level. Per block five.
Leading indicators. Per block seven, so early months show something real.
Comparison against the same period last year wherever a year of data exists.
What was done and what it was expected to affect. The part that makes the numbers interpretable.
Four things to distrust. Each of which we see regularly.
Brand and non-brand combined. The most important, per block two. It is not always deliberate. It is always misleading.
Traffic without revenue. Reporting the input while omitting the outcome, on a shop where the outcome is available.
Ranking screenshots. A selected view of a moving picture, chosen after the fact by whoever is being judged on it.
Attribution windows changed between periods. The most difficult to spot.
If the rules for counting a sale change between one report and the next, the comparison is meaningless while both figures remain individually accurate. Ask whether anything about the measurement changed, then treat any change as a reason to restate the earlier period too.
Setting A Baseline
Whatever is not recorded before work begins cannot be compared to afterwards. This takes an hour and it is skipped almost every time.
What to record. Six things, dated.
Non-brand organic revenue for the previous twelve months, monthly. The same at category level. Indexed pages against total pages. Non-brand impressions. The number of distinct terms the shop appears for. And how each of those was measured, which matters as much as the figures.
Why the last one matters most. Per block eight.
A baseline whose method nobody recorded cannot be compared with confidence a year later, because nobody can be sure the two were measured the same way.
Why this gets skipped. Because everybody wants to start.
A shop that has decided to invest wants work beginning. An hour of recording feels like a delay. Twelve months later it is the difference between demonstrating a result and asserting one.
What happens without it. The argument nobody can settle.
A year in, the shop asks whether this worked. Without a baseline the answer is an opinion. Both parties can hold reasonable and opposite ones. That is the situation this hour prevents.
Who should hold it. You, not your supplier.
Per our guide to choosing an agency, a baseline held only by whoever is being measured against it is not a baseline.
One split decides
whether it worked.
Demand for your own name grows because of advertising, word of mouth and returning customers. A report combining brand and non-brand shows a supplier's work succeeding using demand somebody else generated. Ask yours whether they separate them.
What is included every month:
£350 per month, one target area. No setup fee, nothing billed separately.
Twenty-two guides.
One subject.
This guide covers measurement. The rest of the series covers the sequence, structure, category and product pages, technical health, the marketplaces and everything a store owner has to decide.
