--- title: "We sized our market with keyword volume for two years. Here is what we found when we stopped." url: "https://marketermagazine.co/insight/we-sized-our-market-with-keyword-volume-for-two-years-here-is-what-we-found-when-we-stopped/" author: "Abhishek Shah" published: "2026-09-23" updated: "2026-09-23" --- # We sized our market with keyword volume for two years. Here is what we found when we stopped. Every planning cycle at Testlify used to include the same slide. Somebody totaled the monthly search volume across our keyword set and presented it as the addressable market. Nobody objected, myself included, because the number had a source and a decimal place and had come out of a tool we paid for. The media plan then got built on top of it. The number was real. It was not measuring what we thought it measured, and it took us longer than it should have to notice. ## **What we were actually looking at** Search volume counts people who have already worked out that they have a problem, decided the problem has a category of solution, and learned the vocabulary that category uses about itself. Three conversions happen before anyone types anything. Our slide was a census of late-stage demand and we were treating it as the size of the opportunity. The buyer I most wanted was never in that count. Picture a head of people who has watched three recent hires not work out and has not yet framed that as an assessment problem at all. She asks two peers in a group chat. She describes the situation in her own words to an assistant and reads what comes back. She sits in a session at a conference. None of that produces a query, so none of it reached our planning sheet, and we could not bid on any of it. ## **How the error reached our spend** The mechanism is not stupidity. It is incentive. You can only buy demand you can observe, observable demand sits close to purchase, so our spend concentrated on the terms nearest the transaction. Those terms returned our best measured cost per acquisition. The report then confirmed the allocation had been correct. It confirmed nothing of the sort. A good share of that demand would have found us regardless, and the work that might have created demand earlier was the first thing cut whenever a quarter got tight, because it could not produce a comparable number. Our measurement system was selecting for the spend it could measure, and then handing that spend a larger share next quarter. ## **The three things we changed** We resized the market from something independent of our own vocabulary. Employer counts and hiring volume in the segments we sell to, rather than the sum of the queries people use to describe us. The gap between that figure and our keyword total became the part of the plan we had to reach some other way. [Insert your resized TAM figure and the keyword total for contrast if you want the number on the page.] We started tracking whether we appear inside generated answers, across ChatGPT, Perplexity and Google's AI Overviews, on a fixed set of the questions our buyers actually ask. When we began we showed up in roughly eleven percent of those prompts. Four months later it was closer to twenty-seven percent. That is our own prompt set and our own methodology, so it is directional rather than a benchmark anyone should borrow. Then we built for the thing that moved it. Original research did most of the work. One report of skills-based hiring statistics has earned somewhere past a hundred and eighteen referring links, most from writers we never contacted, and it also turns out to be the page most often lifted into generated answers. Clean attributable figures are easy for a person to cite and easy for a retrieval system to cite, which in hindsight was not a coincidence. We also spent real time being genuinely useful in the communities where our buyers ask each other questions, Reddit in particular, which produced a measurable share of pipeline and almost no attributable clicks. ## **What we got wrong on the way** We treated answer visibility as an SEO project at first and handed it to the same people with the same monthly reporting rhythm. Wrong. The lever is what gets published and who cites it, not what gets optimized. We also expected the paid line to fall once the earned line rose. It did not, at least not on the timescale I predicted. [Add what actually happened to your paid budget here if you are comfortable publishing it.] ## **What I would tell another founder** Take last quarter's won deals and ask each buyer how they first heard of you. Not what the attribution model says. What the person says. Then put that ratio next to the ratio in your media plan. When we did this, the two documents did not resemble each other at all. That was the meeting where the keyword slide finally came off the deck. --- Abhishek Shah, Founder, [Testlify](http://www.testlify.com)