GEO groundwork and measurement for a SaaS product

Three months in, the answer to "did it improve?" rests on impression if nothing was recorded at the start. Impression does not defend a budget: work that cannot show what it returned does not reach a second year, even where it genuinely produced a result. A baseline closes that argument up front, because it leaves a reference to compare against.

That the measurement is reproducible matters as much as taking it. A score read only from publicly available files means anyone running it on the same domain gets the same result — the client, the competitor, and the next agency. A reporting method that only means something internally becomes indefensible at precisely the moment it needs defending.

The page a software company puts most effort into is usually the product page — yet what gets typed into a search box is not the product's name but the job to be done. Someone asking "how do I find leads" has no product name in mind yet; where they will learn it is inside the answer to their own question.

So the content plan is built around the problem rather than the product. Each page owns one real question and finishes answering it; the product appears inside that answer as the solution that gets named. The product page does not disappear — it simply stops being the page that brings anyone in.

The first thirty or forty points of a GEO score mostly come from one-off work: the site being open to engines, introducing itself, saying what it is. Once built, none of it is built again, and the score rises quickly when it lands. In this engagement most of the movement from 34 to 67 across four months came from that layer.

The remaining half is won page by page. Every new page has to own its own question, carry its own answer and bind to its own sources — there is no shortcut, and no part of it an agency can speed up. A slowdown in the later months is not a sign of trouble; it is the sign of having reached the second half of the work.