What SEO KPIs should you ask your SEO agency for?
Seventeen KPIs. That is what the top result for this search gives you, and it is a good article. But you do not have seventeen decisions to make this month. You have one. Is this money doing anything? I report three numbers. Not because the other fourteen are wrong, but because they never changed what I did on a Monday.
See the three numbers How I workreport
Why do I only report three SEO KPIs?
Let me be upfront about something first. The article sitting at the top of Google for this search is good work. Seventeen KPIs, properly explained, screenshots all the way through. If you run marketing at a company with a team and a dashboard, go and track all seventeen. You will get value out of them.
My clients are not that. They are a winery, a strata business, a builder, a consultant. They carry one number in their head about what this costs them. They carry a quieter one about what it brings back.
For them, a KPI that does not change somebody's behaviour is not a KPI. It is a statistic. Interesting once. Useless on a Monday.
So I cut the report down to three numbers. Then I spend the time I saved on the question underneath them, which is the part nobody sells you. If you want to see how that shapes an actual engagement, it is all on my consultancy services page.
Do you appear for what you actually sell?
Here is what the aggregate numbers hide.
Every search is its own event. One query. One location. One device. One person with one intention.
Google does not hold an opinion about your website. It holds a separate opinion about every sentence anybody might type. Those opinions barely speak to each other.
So when a report tells you your visibility score is 34, it has taken thousands of separate decisions and averaged them into a number that describes none of them. You can lift that number by ranking for searches no customer would ever make. You can dent it by publishing an excellent page that starts life on page four for something valuable. Neither movement tells you a thing.
When someone types the sentence that describes what you sell, in the place you sell it, do you come back?
Answer that ten times and you know more about your SEO than any dashboard will tell you.
AI answers make this sharper, not softer
On an old results page there were ten links. Being sixth still meant existing.
In an AI Overview there is one paragraph and a handful of citations. You are in the answer or you are not in the room. There is no page two of a ChatGPT reply.
Averaged visibility scores were always a rough proxy for presence. Against a single answer they say very little.
Write down the ten sentences a customer would say out loud. Not keywords. Sentences.
- "Roof repairs in Terrigal after storm damage."
- "Bookkeeper who can handle Payday Super for a small team."
- "Cellar door near Pokolbin open on a Monday."
Then check those ten from a logged-out browser, in the place you actually serve. That list is your visibility. The rest is commentary. It takes twenty minutes and it costs nothing.
Why do my rankings look fine when my page is invisible?
Let me use my own site. It is the example I am allowed to be blunt about.
This page exists: roxanepinault.com.au/local-seo-consultancy-services. It is written, it is on brand, it has pricing, proof and a call to action. All of it is there.
It is not indexed. Google does not have it. It cannot appear for anybody. Not once, not ever, not in this state.
And four of the six KPIs below report perfect health.
TODO before publishing: re-check the index status of /local-seo-consultancy-services in Search Console. If it is indexed by then, keep the story and change the tense: "it was not indexed when I wrote this, and here is what fixed it." Then delete this flag.| The KPI | What it reports for this page | Does it flag the problem? |
|---|---|---|
| Average position, site level | Unchanged | No. A page that does not exist cannot drag a number down. Nothing in the report flinches. |
| Impressions | Zero, and the page is missing from the top pages list | No. Absence does not show up in a list of what is present. |
| Organic sessions, site level | Flat, well inside normal variation | No. |
| Domain Rating | Unchanged | No. It was never going to move for one page. |
| Key events from organic | Zero from this page | Yes, as long as the report is broken down by landing page. |
| Non-branded clicks for its target searches | Zero, month after month | Yes. And it names the searches you are missing. |
That is not the tools failing. That is what happens when you measure a site in aggregate. Absence never shows up inside the presence of everything else.
How do I check whether a page is indexed?
- Type the exact page title into Google, inside quotation marks. Does the page come back?
- Open Search Console. Paste the URL into the inspection bar at the top. Read the verdict.
It tells you whether the page is on Google, whether it has been crawled, and usually why not. That is the whole diagnostic. No subscription required.
The fix, in this case, is mostly the link you just read
A page that nothing points to is a page Google has little reason to fetch.
I have now linked that page twice from an article I expect people to read. That is the most ordinary internal linking there is. I will update this section when the page turns up in the index, and I will leave the story in. A working example of the problem is worth more to you than a tidier article.
How many customers do you need for SEO to pay for itself?
This is the part everyone skips. It is also the part that makes every other number mean something.
You need three figures. What you spend on SEO in a year. What one new customer is worth to you, using margin rather than revenue if your margin is thin. How many of those it takes before the spend has paid for itself.
That third one is not an opinion. It is a division.
| One new customer is worth | Customers needed to break even | Which in practice is |
|---|---|---|
| $500 | 12 per year | One a month, every month, before you are level |
| $2,000 | 3 per year | One a quarter |
| $5,000 | 1.2 per year | Roughly one, some time in the year |
| $20,000 | 0.3 per year | One every three years |
Read the top row again. If you sell five hundred dollar items, SEO has to bring you a dozen a year before you are level. That is not a reason to avoid it. It is a reason to say the number out loud in month one, so that month seven is a conversation about progress rather than a conversation about feelings.
Now read the bottom row. If you sell twenty thousand dollar jobs, one client in three years covers the year. The argument is close to over before it starts. The only real question left is whether the work is any good.
I will not promise you a number of customers by a particular month. I would be careful with anyone who does. What I will do is agree the break-even figure with you before we start, and report against it. The money side of that conversation lives on the local SEO pricing page.
How do I track organic conversions in GA4?
This is the lead count. If I could keep one number, this is the one I keep.
Key events from organic search
That is what a working SEO program looks like. Not a traffic spike. A steady rise in the number of people who arrived from a search and then did something.
Mark the right things, or the number lies to you
A key event is only as good as what you decided to mark. Mark what a human being would call interest. A form submitted. A phone number tapped. A booking started. A quote requested. A document downloaded that only a serious buyer wants.
Do not mark scroll depth. Do not mark ninety seconds on a page. Do not mark a click on the menu. Those inflate the number and teach you nothing.
Then segment it to organic search only. A total that blends paid, direct, social and email is a marketing number. It is not an SEO KPI.
"Show me key events from organic search only, monthly, for the last twelve months, broken down by landing page."
"How many of those became customers?"
What should I look at in Ahrefs every month?
This is the maintenance number. It rarely gives you a good news story. It quietly prevents most of the bad ones.
Redirects and crawlability
The health score itself is a composite, so I do not report it as a headline. What I read inside it is redirect chains, internal links pointing at URLs that redirect, pages that changed address in a rebuild and never had their links updated, and old pages now serving a soft 404.
New referring domains, with names
Not the total. The total mostly reflects history. I want the new ones, this month, with names attached and a view on whether any of them are worth anything. A rising chart with no names under it is decoration.
What I recommend locally
For a local service business I usually recommend paying to be in the directories that already rank for the search you want. Those are the listings your customer genuinely sees. It is not romantic. It works.
Redirect housekeeping is dull. It is also where the value of work you already paid for leaks out. A site that has been rebuilt twice is usually carrying a lot of it.
"How many new referring domains did we gain this month, what are they, and what did they cost?"
"How many internal links point at a redirect or a broken URL right now, and what is the plan?"
"Did we lose any referring domains this quarter, and do we know why?"
How do I see non-branded clicks in Search Console?
Branded clicks are not SEO.
When somebody types your business name into Google, they already knew about you. That search was won by a referral, an ad, a conversation, or a van with your logo on the side. Counting those clicks as an SEO result is the most common way a report flatters itself. Usually nobody means to do it.
Non-branded clicks are the useful half. Those people did not know you existed. They described a problem. They landed on you anyway. That is the thing you hired SEO to do.
Set it to sixteen months, which is the maximum, and look at the shape rather than the week.
Twenty-eight days tells you about the weather. Sixteen months tells you about the climate. SEO is a climate business.
"Show me non-branded clicks over sixteen months, and the searches behind the growth."
Which SEO metrics can I safely ignore?
None of these are bad numbers. I look at most of them constantly. They just do not go on the front page, because they do not change what anybody does next week.
| Metric | Why it stays off the front page |
|---|---|
| Keyword rankings, as a headline | Positions are personalised, localised and volatile. A list of green arrows is easy to produce and easy to choose. I check specific searches. I do not present a ranking table as a result. |
| Search visibility scores | A useful sense of direction for a large site. For a business with twenty pages it averages things that are not comparable. |
| Organic sessions | Interesting when it moves for a reason. On its own it answers a question nobody at the business is asking. |
| Bounce rate and time on page | They describe a page, not an outcome. A three second visit to a phone number is a success. |
| Domain Rating, Domain Authority | Vendor scores. Useful for comparing two link prospects. Not a business result. |
| Core Web Vitals | A fix, not a KPI. Get them into the green, then stop reporting them every month. |
| Impressions | The most misread number in Search Console. Impressions rise when Google tries you out for things you do not want. |
Should AI visibility be one of my SEO KPIs?
Track it. Do not put a target on it yet.
Whether ChatGPT, Google's AI Overviews, Gemini or Perplexity name you when someone asks for what you sell is a real thing to know, and it matters more every year. I track it, and I run audits on exactly that question.
But the measurement is young. The answers vary between users and between sessions. I am not willing to put a target next to a number whose movement I cannot yet explain to you. Watch it, log it, talk about it. I would not build a bonus scheme on it this year.
What questions should I ask my SEO agency?
These are not trick questions. A good agency will be pleased you asked. Most of them can be answered while sharing a screen.
| Ask this | A useful answer contains | A thin answer sounds like |
|---|---|---|
| What is our break-even number of new customers? | Your spend, your customer value, a division, and a timeline | "SEO is a long-term investment" |
| Key events from organic, by landing page, twelve months? | A monthly chart, a page list, and an opinion on which pages earn | Total sessions |
| Which ten searches should we be appearing for? | Ten sentences a customer would say, and where we stand on each | A list of two hundred keywords |
| Are all our commercial pages indexed? | A number, a list of the ones that are not, and the reason | "Google is slow at the moment" |
| How many new referring domains this month, and what were they? | Names, cost, and a view on quality | A rising chart |
| Non-branded clicks over sixteen months? | The chart, plus the searches driving the change | Branded and non-branded added together |
| What did you stop doing this quarter, and why? | Something specific that was not working | "We are continuing to optimise" |
The last one is my favourite. An agency that has never stopped doing anything is not making decisions.
What should a monthly SEO report look like?
One page. Three numbers. Two sentences.
Then a sentence on what changed. Then a sentence on what happens next month.
If a report needs a call before it makes sense, it is doing something other than reporting.
Not sure your pages are even in the index?
That is the first thing I check, and it takes minutes. The free audit covers whether Google has your commercial pages, and whether the AI assistants name you when somebody asks for what you sell.
SEO KPIs, the questions I get asked
What makes a good SEO KPI?
One that changes a decision. If the number goes up and nobody does anything differently, and it goes down and nobody does anything differently, it is a statistic. Key events from organic search, new referring domains and non-branded clicks all pass that test. Each one has an obvious next move attached to it.
How long before SEO KPIs start to move?
Site health issues can improve within days of being fixed. Non-branded clicks usually start shifting somewhere between three and six months, and they read properly over twelve to sixteen. Key events follow the clicks.
Anyone giving you a firm date is guessing. Anyone who says "it depends" and stops there is not being much more useful. Ask for the reasoning instead. What is being changed, why that should produce clicks, and roughly when.
Should AI visibility be one of my SEO KPIs in 2026?
Track it, yes. Put a target on it, not yet. Whether an AI assistant names your business is worth knowing and worth working on. The measurement still varies too much between users and sessions for me to be comfortable putting a target beside it.
My agency only reports keyword rankings. Is that a red flag?
Not on its own. Some agencies report rankings because a previous client asked for them and the habit stuck. Ask for key events from organic, broken down by landing page, and see what comes back.
If it arrives, you have a reporting habit to fix. If it cannot be produced at all, you have found something more useful than a red flag. You have found out that nobody has been looking at outcomes.
Do impressions matter at all?
They are useful as a diagnostic and misleading as a KPI. A jump in impressions with no change in clicks usually means Google is testing you for searches that are not yours. That is worth investigating. It is not worth celebrating.
How do I check whether one of my pages is indexed?
Put the exact page title into Google inside quotation marks and see whether the page comes back. Then open Search Console, paste the URL into the inspection tool at the top, and read the verdict. It tells you whether the page is on Google, whether it has been crawled, and usually why not. It takes about thirty seconds.
Can I track these SEO KPIs without paying anyone?
Most of them, yes. Search Console is free. GA4 is free. Between them they answer the questions in this article. Ahrefs and its equivalents cost money and mostly save time. If your budget is tight, spend it on the website itself before you spend it on a retainer.
Roxane Pinault, AI visibility and SEO consultant
I am based on the Central Coast of New South Wales. I have been doing SEO in Australia since 2015. I work with a small portfolio of clients rather than a large one, on purpose. Australia sits in the front row for AI search, which makes it an interesting place to be doing this work right now.
That a particular number will hit a particular figure by a particular month. Search does not work that way, and neither do I.
Agree your break-even number before we start. Report those three KPIs. Tell you plainly when something is not working.
Start with the one that costs nothing
Open Search Console. Filter your own business name out of the queries. Set it to sixteen months. That one chart will tell you more than the last three reports you were sent, and it takes a minute. Where you go after that is up to you.
How I work Ask me a question