The cheapest way to say it changes at the border

For each of eight cities we read two phrasings of the same intent — “seo agency [city]” and “seo company [city]” — in that city’s own country database, between 27 and 28 September 2026. Volume, keyword difficulty and cost per click are as Semrush displayed them.

One rule governs everything below. Difficulty is scored against the set of sites competing inside a single country database. A 35% in Australia and a 36% in the United States are two different contests, not two points on one scale. So this piece never compares a difficulty across a border. It compares the direction of the gap inside each market — which phrasing is easier there — and that comparison is like-for-like.

We are careful about this because we got it wrong once already this month, comparing an Indian difficulty score with an American one and drawing a conclusion from it. The correction is published here.

Australia and Canada favour “agency”. Britain and the United States favour “company”. And the size of the gap ranges from nineteen points down to one.

Miami is the extreme case: “seo company miami” measures 17% difficulty against 36% for “seo agency miami”. Same city, same service, same buyer — the words are the only variable, and they halve the difficulty.

Chicago is the cleanest case, because there is no trade-off at all. Everywhere else the easier phrase has less volume; in Chicago “company” has 600 more searches a month and is seven points easier. Nothing to weigh up.

And Houston is the case that matters most for anyone hoping to shortcut this. It sits in the same database as Chicago and Miami, and its two phrasings are one point apart — 39% against 40%. There is no gap to exploit. So the flip is not even consistent within a single country.

The obvious explanation is advertiser behaviour, and the cost-per-click figures are at least consistent with it.

In six of the eight cities the “company” phrasing costs more per click. If advertisers judge someone searching for an SEO company to be closer to signing, they bid harder — and the agencies bidding hardest tend to be the ones investing most in organic for the same phrase. That would make the more valuable phrase the more contested one.

It fits Australia neatly. It does not fit Miami or Chicago at all, where “company” is both more valuable and easier. So either the mechanism is different in different markets, or something else is driving it entirely.

We have not measured why, and we are not going to assert a cause we cannot show. What we can show is the effect, and the effect is enough to act on.

Take the ordinary way a city page gets planned: pull both phrasings, pick the one with more volume, build for it.

In Miami that produces a page aimed at 1,300 searches a month at 36% difficulty, when 1,000 searches a month were available at 17% — and the cheaper term pays more per click, $15.86 against $13.92. You would have taken roughly double the difficulty for a 30% volume gain and a less valuable visitor.

In Chicago the volume heuristic gets it backwards outright: it points you at the 1,300/36% term while the 1,900/29% term sits next to it.

In Sydney the same heuristic happens to be right — “agency” has both the volume and the lower difficulty. A rule that is right in one market and expensively wrong in another is not a rule. It is a coincidence you have not noticed yet.

This is worth saying plainly because it is the point of the piece.

On the morning of 28 September we measured the three Australian cities, found “agency” easier in all three, and wrote it onto our Sydney page as a rule: build for “agency”, buy clicks for “company”. It looked like a clean three-for-three pattern with a plausible mechanism behind it.

That afternoon we measured Manchester, Chicago, Miami and Vancouver to check whether it travelled. It did not. Two of the four reversed, and by bigger margins than the Australian gaps we had generalised from.

The Sydney advice was not wrong — it was measured in Australia and it holds in Australia. What would have been wrong is the article we were about to write. Three samples from one country looked like a pattern; four more from three countries showed it was a local one. That is the second time this month a habit of carrying a number across a boundary nearly produced a false finding on this site.

The fix is not cleverness. It is a column in the spreadsheet recording which database every reading came from, and a rule that no comparison crosses it.

• Measure both phrasings, in the country you are actually selling into, before you write anything. Two lookups. It is the cheapest decision in the whole project and it moved difficulty by nineteen points in one of our eight cities.
• Do not carry the answer to the next market. Not to the next country, and on this evidence not reliably to the next city either.
• Read difficulty and cost per click together. In six of eight cities the two phrasings differ in value as well as difficulty, and sometimes the cheaper one to reach is also the better paid.
• Keep a market column in your keyword sheet. Without it, half your comparisons are unsafe and you cannot tell which half.
• Accept the null results. Houston’s one-point gap is not a failed measurement. It is the answer: there is nothing to optimise there, and knowing that took one lookup rather than a quarter of content.

We will re-measure all eight at ninety days and publish the movement, including wherever it shows nothing.