A keyword tool gives you a hypothesis, not a target
A research tool tells you what a phrase is worth if people use it for your thing. It cannot tell you whether they do.
A site retargeted its main page onto a term a research tool rated at 2,900 searches a month and modest difficulty. It looked like the obvious choice.
Three weeks of real search data later, that phrase had produced a single impression. Meanwhile two clusters nobody had chosen were producing roughly 150 impressions each, entirely on their own.
Why a good-looking term can be the wrong one
Volume is measured against the phrase, not against your intent. A phrase can be popular for a neighbouring meaning, be dominated by informational rather than buying intent, or simply not be how your particular buyers describe the thing.
The clearest example from that site: the page targeted one word for a device while the largest genuine term for it used a different word entirely. Same product, different vocabulary, and the page did not contain the buyers' word even once.
The two data sources answer different questions
- A research tool answers: how big is this phrase, and how hard is the competition.
- Your search console answers: which phrases are you actually being shown for, and at what position.
Neither is sufficient. The console cannot show terms you have no presence for. The tool cannot tell you which phrasing your buyers use. Used together, the console proposes the vocabulary and the tool sizes it.
Pull your impression data first and let it suggest the phrasings, including the ones you would not have chosen. Then price those with a research tool. Choosing the target from the tool alone is how you end up optimising for a phrase nobody types.
Do not dismiss small volumes
The same exercise turned up terms at 20 to 50 searches a month carrying advertising costs of £50 to £85 a click, at the lowest difficulty on the page. Low volume and high commercial value often travel together in business-to-business niches, and a volume-ranked list buries them. Sort by value and difficulty as well as by size.
Revisit the decision
Record why a target was chosen, and check it once real impressions exist. A target picked from a tool is a hypothesis. A few weeks of live data is the experiment. Nothing is wrong with having been wrong, only with not going back to look.
Need help with any of this?
These notes are free and always will be. If you would rather someone just set it up, or you are stuck on something similar, get in touch at hello@opsira.io.