The rank report on your screen is fiction
In September 2025, Google quietly switched off the &num=100 parameter. One tweak, and every rank tracker that used to pull a hundred results in a single request started seeing ten. "We rank #47 for that term" turned, overnight, into a number nobody could cheaply verify anymore.
The tooling panic missed the more useful lesson. That #47 was already fiction. Not because the tool lied, but because the question behind it, "where do we rank for this keyword", quietly assumes a keyword has a rank. It doesn't. It has a spread, and the spread is the whole story.
One query, a dozen answers
Search "best payroll software" and the page you get in Austin is not the page someone gets in London, or Singapore, or São Paulo. Different vendors, different order, sometimes a different intent entirely, because "payroll" drags different tax baggage behind it from one country to the next. Google decides most of that from a country signal, gl, and under that, from your IP address.
And it runs whether you're logged out or not. Incognito doesn't switch it off either. Location personalization is IP-based, and Google has aimed it at signed-out users since 2009. So the "clean" incognito check you trust to reveal the "real" ranking is only ever revealing the ranking for your own office, and nowhere else on earth.
Two searchers, same second, same query, two different cities. Two different pages. Neither one is the true ranking, because there is no true ranking.
A number pretends the map is a dot
A rank report gives you one integer per keyword. That integer flattens a whole map down to a single dot. It tells a client they sit at #3 as if #3 were a fact about the universe, when it's a fact about one IP address on one Tuesday afternoon.
The flattening gets expensive at the worst possible moment: expansion. A founder checks "project management tool" from a desk in Berlin, sees their product sitting mid-page, and green-lights a UK launch on the strength of it. But gl=de and gl=gb are two different countries to Google, with different incumbents and different furniture bolted onto the results page. The Berlin read said nothing about Manchester. The call got made on a dot when the situation needed a map.
What the map actually looks like
To see the spread instead of the dot, you change one variable and freeze the rest. Pin the country with gl, set the language with hl, and for street-level detail encode the exact place into uule. Then move only the location and watch the same keyword slide up and down.
I run this before I'll quote anyone on a market I don't already live in. Take a single query and fire it as gl=us, then gl=gb, then gl=au, then gl=de, every other parameter held identical. Write down the top ten each time. What comes back isn't a ranking. It's a shape: strong in two markets, invisible in a third, knife-fighting some local incumbent you'd never heard of in the fourth. That shape is the thing you're really deciding on, and a single number erases it.
Line two of those pages up side by side, the market you already own against the one you're eyeing, and the vague boardroom question "should we expand" collapses into a concrete one: here is exactly who we'd have to outrank, and here is where. That comparison is worth more than any rank-tracker number, which is the entire reason a geo-comparison view exists in the first place — pick the two markets, fire the same query at both, and read the difference yourself.
The number was never the thing
None of this gets fixed by a better rank tracker. It gets fixed by dropping the wrong question. "Where do we rank" has a hidden "from where" buried inside it, and once you say that second part out loud, "our server's IP" is a humiliating answer, because your customers are not, as a rule, sitting inside your server rack.
So the num=100 change stung because it made a cheap illusion expensive. The illusion was never worth much. A keyword's position was never a number you look up; it's a range you sample, and the markets you never bothered to sample are usually the exact ones quietly beating you.