Every month a report arrives. Impressions up, clicks up, average position improved by one and a half. Somewhere on page two there is a chart trending in a pleasant direction.
And you still have no idea whether the money is working, because none of those numbers are the thing you care about. You care about whether more of the right people called.
The gap between those two is where most of the frustration in this industry lives — and it has gotten wider, because a growing share of findability now happens in places that produce no measurement at all.
The Numbers That Look Like Progress and Are Not
Some metrics are genuinely useful. Others are mostly comfortable, and it is worth knowing which is which.
- Impressions tell you that you were shown. Shown to whom, for what, and whether it was anywhere near a customer, they do not say. Impressions can rise because you started appearing for searches that will never convert.
- Average position is an average across every query and every location, which means it can improve while your position for the searches that actually matter gets worse. It is one of the most-quoted and least-useful numbers in a standard report.
- Total traffic is only meaningful once you know which pages. Traffic to a blog post from three states away is not the same asset as traffic to a service page from your county, and one line labelled Sessions hides that completely.
- Keyword rankings from a rank tracker are checked from one place. Your customers are not in one place. This is the same problem that makes owners think they rank better than they do.
A number that goes up is not the same as a number that matters.
What to Watch Instead
These are harder to produce and worth more than everything above.
- Calls, form fills, and bookings — counted and attributed. The actual outcome. If you cannot see this, nothing else on the list matters much, and getting it set up is the first job.
- Direction requests and profile calls. Your Google Business Profile reports these directly, and they are among the cleanest local intent signals available. Somebody asking for driving directions is close to a customer.
- Position across your service area, not from your desk. A grid view of where you place from many points is a genuinely different instrument than a single-point rank check, and for local businesses it is the honest one.
- Conversion rate by page. Traffic that does not convert is a content problem or an intent mismatch, and you cannot see either from a traffic total.
- Whether the phone conversations changed. Not a metric, and often the earliest real signal. When people start calling already knowing what you do, having read something, asking about a specific service, something upstream is working.
The Part Nobody Can Measure Yet
Now the honest problem. A growing share of how people find local businesses produces no analytics at all.
When an AI assistant recommends you, there is often no click, no referrer, and no session. Somebody asked where to go, got three names, and called one. If it was you, it arrives in your business as a phone call from a person who cannot really explain how they heard about you.
Be deeply skeptical of anyone selling you a precise number for this. The attribution genuinely is not there yet, and a confident dashboard metric for AI visibility is usually a proxy dressed up as a measurement.
What you can do is cruder and more truthful. Ask the assistants directly, on a schedule — monthly, the same prompts each time, phrased the way a customer would phrase them, from a signed-out session. Write down who gets named. That is a real longitudinal record, and watching your own name appear over a few months is more informative than any score.
And ask new customers how they found you. Not in a survey. In conversation, at intake, and write the answers down somewhere. It is unfashionable and it is frequently the only attribution you will get for the channels that matter most.
How Long Before Any of It Moves
Different work reports back on different schedules, and mixing them up is the single biggest cause of people quitting something that was about to work.
Profile corrections and consistency fixes register fastest, often within weeks, because you are removing a reason to doubt you rather than building something new. Review velocity shows up over a couple of months as the pattern establishes. Content and page structure are slower still, generally a few months before pages settle. Competitive displacement in a hard market is measured in quarters.
We laid out realistic timelines in how long SEO takes. The short version is that the fast wins are real and the durable ones are slow, and any report that shows everything moving at once should raise your eyebrows rather than your confidence.
What a Useful Report Looks Like
If you are paying somebody, this is a fair standard to hold them to.
It should say what was done, not just what happened. It should show outcomes — calls, forms, bookings — before it shows traffic. It should state what did not work, because everything works in a report that never mentions a failure. It should compare against the same month last year rather than last month, since almost every local business is seasonal. And it should end with what happens next and why, in language you would use out loud.
If a report cannot survive the question so what should we do differently, it is decoration.
Frequently Asked Questions
What is the single most important number to watch?
Calls and form fills from people in your service area, attributed to the source that produced them. Everything else on a standard report is a leading indicator of that, and leading indicators are only useful when the actual outcome is also being counted.
My rankings improved but the phone did not ring. What went wrong?
Usually one of three things: you improved for searches that do not convert, you improved in places your customers are not, or the page they landed on did not persuade them. All three are diagnosable, and all three are invisible if rankings are the only thing being measured.
Can anyone actually track whether ChatGPT recommends me?
Not reliably, and claims otherwise deserve scrutiny. There is generally no click and no referrer. Manual testing on a fixed schedule, plus asking customers how they found you, is the honest approach available today.
How long should I give a new SEO engagement before judging it?
Expect early movement on profile and consistency work within weeks, meaningful content results over several months, and competitive displacement over quarters. What you should see immediately is clear reporting on what is being done. Vagueness in month one is the real warning sign, not slow rankings.
Is a monthly report even necessary?
Something is. It does not need to be long. It needs to say what was done, what it produced, what failed, and what is next. A short honest report beats a long automated one, and the automated ones are usually long precisely because they are not saying much.
Get a Baseline You Can Measure Against
You cannot tell whether something is working without knowing where you started. Most businesses begin marketing without a baseline and then argue about results for a year.
Our free competitive analysis gives you one — where you place across your service area, who holds the positions you are missing, and what the AI assistants currently say when somebody asks who to call. Run it before you spend anything, so you have something to compare to later.