Ready to buy
Core service terms from people who already know what they need. Highest intent, highest bids, tightest landing pages.
Case study · Paid search and CRO
Making paid search testable
Four campaign groups split by what the searcher actually wants, one variable changed per test, and every lead followed through to whether it was a real buyer.
4campaign groups, split by intent
1variable per test cell
4things tested: keyword, message, page, creative
End to endclick through to closed deal
Problem
One big campaign teaches you nothing.
When keywords with different intent share a campaign, the budget drifts to whichever term is cheapest, not whichever term brings buyers. And when a campaign changes its keywords, its copy, and its landing page in the same week, and results move, nobody can say which change did it.
Paid search runs on this pattern more often than not. It looks like optimisation. It is guessing with a budget attached.
Four groups, by what the searcher wants
Technical buyers
Specific method and technique terms. Fewer searches, far better fit, and copy that has to be accurate rather than persuasive.
By industry
The same service framed for a sector. Separated because the language, the objections, and the deal size all differ.
Competitor terms
People comparing options. Different message entirely, and worth isolating so its cost per lead never gets averaged into the rest.
Method
Change one thing per cell.
Each test cell varies exactly one of four things and holds the rest still:
- The keyword theme.
- The message angle: speed, scale, accuracy, or price.
- The landing page.
- The ad creative set.
It is slower. It is the only way the result means anything. A search terms report every fortnight feeds a negative keyword list, so the money stops going to searches that were never going to convert.
The part most teams stop before
A conversion is not a lead.
The ads platform reports form fills. A form fill is not a buyer. So every paid search lead over six months was read one by one: who they were, what they wrote, whether the company was a plausible customer, and if the deal did not close, why not.
That review changed the campaigns more than any bid adjustment. It showed which keyword themes brought people looking for work rather than looking to buy, and it showed which landing pages attracted the wrong reader entirely. It also produced the rules behind the lead qualification gate, which now filters those out automatically.
Keywords, competitor names, spend, conversion figures, lead names, and company names are withheld. The structure and the testing rules are the transferable parts.