A report that only shows keyword positions is close to worthless, and it has been for a while. A report that shows impressions, clicks and position together, with a stated floor under what counts as a change, is worth paying for. The difference is not the tool. It is whether whoever built the report set a rule before they looked at the numbers.
I will tell you the exact floor we use on ours, because if I am going to say most reports are noise I should show you what we hold our own to.
Why a position number stopped meaning one thing
Ranking used to be a single fact. There were ten blue links, your page was one of them, and the number was the number.
Now the same query returns different results depending on where the person is standing, what device they are on, whether they are signed in, and what Google decided to put above the organic results that day. A map pack pushes the first classic result down the page. An AI summary can answer the question outright. On a phone, position 3 can be below the fold in a way it is not on a desktop.
So "you moved from 8 to 5" describes a thing that does not exist in one place. A rank tracker checking from a data center in Virginia is telling you about a search nobody in your service area performed.
That does not make position useless. It makes position useless on its own.
The floor we put on our own reports
Here is our standard, and you are welcome to hold us to it or to hand it to whoever you are working with now.
For a query to be reported as a mover in one of our digests, it has to clear three things:
- At least 5 impressions in both the current window and the one before it. A query cannot "improve" from a single impression to two. There is nothing there to average.
- At least 20 impressions in total across the two windows. Under that, the position number is being calculated from too little data to mean anything.
- Rows that are 95 percent or more desktop, with 20 or more impressions and zero clicks, get thrown out entirely. That pattern is almost always scraping or bot traffic, not a person, and it will happily drag a position average around and make a chart look like progress.
None of that is complicated. It is just a decision made in advance, which is the part that matters, because a number with no floor under it is a headline waiting to be written.
The reason we wrote it down is that it stops us from reporting good news to ourselves. If the rule only gets applied when the results are bad, it is not a rule.
What actually predicts whether the phone rings
This is the part I would push hardest on, and it is where most reports go wrong even when they are honest.
Clicks are predicted by the share of your impressions that sit in positions 1 through 10. Not by how many impressions you have. A site can triple its impressions and get no more work, because the new impressions all landed at position 40 for queries nobody clicks from.
So the number I want on the front page of a report is what percentage of your impressions are in the top ten, and whether that percentage is going up. A site with 800 impressions where 30 percent are top ten is in better shape than one with 4,000 impressions where 4 percent are. We have watched a contractor site in the Northeast add thousands of impressions across a quarter and take exactly the same number of calls, and the impression chart looked like a triumph the whole time.
Impressions are the easiest metric to grow and the easiest to grow uselessly. Which is why they are on every report.
What a monthly report should actually contain
Two halves, and most vendors send one of them.
The first half is what was done. Specific action items: pages published, pages rewritten, technical fixes shipped, citations built, GBP posts made, links earned, with dates. If you cannot tell from the report what anybody did with the month, you are paying a retainer with no scope, and there is nothing to evaluate at renewal.
The second half is what happened, with the floor applied. Clicks and impressions from Search Console, the top-ten share, the queries that moved with enough volume behind them to count, and the ones that fell. Conversions if conversions are tracked, which they often are not, and that is its own problem.
The honest version of that second half will sometimes say nothing moved this month. Any report that finds a win every single month is not measuring, it is marketing. Most months of real SEO work look like nothing much happened, and then a quarter later the line has moved.
The red flags, plainly
- A report with no action items. You cannot renew against it.
- Positions with no impression counts next to them. Ask for impressions on every row.
- A tracker's numbers with no Search Console data behind them. Console is your own data, from Google, for free. There is no reason for it to be missing.
- Nothing ever gets worse. Real data has bad rows in it.
- Rankings for queries you would never want. Being number one for your own company name is not a result.
- No definition anywhere of what counts as a change. This is the one that hides all the others.
What to send your current vendor
If you are holding a report you do not trust, you do not need to fire anybody this week. Send one email and ask for three things: Search Console clicks and impressions for the period, the share of impressions in positions 1 to 10 and whether it moved, and the list of what was actually done, with dates.
A good vendor will send it back the same day, probably relieved that you asked. A vendor who explains why those numbers are misleading is telling you something useful.
We cover what belongs in a monthly scope in what local SEO includes monthly, the vetting questions in how to vet an SEO company before you sign, and why a big site can still be invisible in why your contractor site isn't ranking with 100 pages. If you want us to read your current report and tell you what it is not saying, send it over. That is a free half hour and there is no pitch attached to it.