Search Volume Explained: Where the Numbers Come From

Full transcript

1,033 words · script draft, under review.

Intro

You look up one keyword in three tools and get three different search volumes. One says thirteen hundred a month. One says twenty-one hundred. One says eight hundred and eighty. Those are made-up numbers for illustration, for a fictional coffee equipment shop, but the situation is one you will meet.

Which one is right? Usually none of them, exactly. But they're not random either. Each number comes from a specific source, is processed in a specific way, and is wrong in a predictable direction.

In this episode we go through where search volume really comes from, what happens to it before you see it, and how to use it without being fooled.

Source one: Google Ads data

Almost every search volume you've ever seen starts in one place: Google's advertising system. Google Ads has a planning tool, the Keyword Planner, that tells advertisers roughly how often a keyword is searched. That data is also available through APIs, and DataForSEO's Google Ads search volume endpoint is the one we use in this series.

This is the closest you get to Google's own count. But remember who it's built for. It's built for advertisers deciding what to bid on, not for SEOs measuring demand. That shapes everything that follows.

For each keyword you get an average monthly volume, plus the twelve individual months behind it. It's cheap to pull: one task covers up to a thousand keywords and costs about nine cents. And it has rules. A keyword with a comma, an ampersand or a slash makes the whole task fail. Very long keywords are rejected. And you're limited to about twelve requests a minute.

What happens to the number: averaging and buckets

First, the headline number is an average over twelve months. A keyword with a big peak in spring and almost nothing in winter gets one calm-looking number. If you build a page in October based on that average, you might be looking at a quiet month for the next half year. That's why I always look at the monthly values, not just the average. We covered seasonality in episode thirteen.

Second, the number is rounded into buckets. Google doesn't report "one thousand one hundred and forty-seven". It snaps to a fixed ladder of values. Ten, twenty, thirty, then fifty, seventy, ninety, and further up, one thousand, thirteen hundred, sixteen hundred, nineteen hundred. The steps grow with the volume.

So two keywords at sixteen hundred might really be fourteen hundred and eighteen hundred. Treat the number as a range, not a measurement. Comparing one thousand and thirteen hundred and calling one "thirty percent better" is fake precision.

And below the bottom of the ladder, small volumes are reported as zero. That's its own episode, the next one.

What happens to the keyword: close variants

The third thing is grouping. Google treats close variants as one keyword in its planning data. Plurals, some misspellings, some reordered words. So when you ask for three variants, you might get the same number three times.

That number is not three separate audiences. It's one group, reported three times. If you add them together, you've tripled the demand on paper. This is one of the most common errors I see in keyword spreadsheets, and it's easy to catch: identical volumes on near-identical keywords means one synonym group. Never sum them.

The grouping isn't perfect, though. Sometimes singular and plural are not grouped, and get very different numbers. That's a problem big enough for its own episode too, episode twenty.

Source two: clickstream-based estimates

Now the other kind of number. Many SEO tools don't just repeat Google's figure. They build their own estimate, often from clickstream data. That's anonymized browsing data from panels of real users, scaled up with a model to the whole population. DataForSEO's Labs endpoints, for example, return their own estimates rather than the raw Google Ads figure.

These estimates have strengths. They can split grouped variants apart, and they can be more precise at low volumes. But they're modelled. The panel is a sample, the scaling is a guess, and the model is refreshed on its own schedule. So a clickstream estimate can be closer to reality for one keyword and further away for the next.

Why tools disagree

So back to our three numbers. Here's why tools disagree, in order of how often it matters.

Different source: Google's planning data versus a modelled estimate. Different time window: last twelve months, versus a different twelve, versus last month only. Different grouping: one tool reports the group, another splits the variants. Different location: whole country versus all countries, or a language setting. And different rounding.

When two numbers disagree, check those five things before deciding one tool is wrong. Most of the time they're answering slightly different questions.

A reality check with your own data

There's one check you can run yourself, for free. Find a keyword where your own page has ranked in the top three for a full month. At that position, almost everyone who searches sees your result. So your monthly impressions in Search Console are close to the real number of searches.

Put that next to the tool's volume. If they're in the same range, good. If the tool says thirteen hundred and you see a hundred and fifty impressions at position one, something's off: the grouping, the location, or the month.

Do this for five or ten keywords and you'll learn how a given source behaves for your market. Is it usually high, usually low, or about right? That's worth more than any general claim about which tool is most accurate.

How I use search volume

Three habits.

One source per dataset. Pick one volume source for a project and stick to it. Mixing Google figures and modelled figures in one ranking is like mixing currencies without converting.

Read the months, not just the average. The average hides the shape of demand.

And use volume as a relative signal. It's good at telling you that one topic is much bigger than another. It's bad at telling you how many visitors you'll get. Your own Search Console impressions are a far better measure of real demand for pages that already exist.

Wrap-up

Search volume is an advertiser's estimate, averaged, rounded and grouped. Use it as a range. The volume sanity checklist is linked in the description. Next episode: why a keyword with zero volume can still have real demand. See you there.