Identify exactly what the number represents.
Before sorting a research sheet, read the provider’s current definition of its metric. It may be an index, estimate or other modelled signal. Record the source, scale, store, market and date. An unexplained column headed “volume” is not enough context.
This guide does not assume every tool calculates demand the same way. If the definition is unclear, mark the limitation instead of inventing a meaning.
Compare like with like.
A score from one provider should not be treated as numerically interchangeable with a score from another. Country, platform and collection period can also change the comparison. Keep these dimensions consistent when evaluating relative demand.
The chart is a fictional index example. A score of 60 is not presented as 60 searches or as proof of twice the demand of a score of 30. The scale’s meaning must come from its source.
Keep relevance ahead of the sort order.
A high-demand phrase for a product you do not offer is not automatically an opportunity. For a fictional receipt scanner, an appealing payroll term still needs to pass the product-fit check. More exposure to unsuitable users can create misleading expectations.
Use relevance and intent before comparing demand among suitable candidates.
Do not skip the steps between demand and value.
Even if demand were known precisely, visibility, position, audience response and product fit would still affect what happens next. A forecast that multiplies a score by an assumed conversion rate can create false precision.
Use scenario arithmetic only when the inputs and assumptions are explicit. The conversion calculator illustrates arithmetic, not the performance of a keyword.
Use the signal as one input to a decision.
Record why a candidate matters, how strong the evidence is and what other constraints apply. A lower estimated-demand term with clear product fit may deserve investigation ahead of a broad phrase with little relevance. The decision depends on the app.
Read difficulty estimates alongside this guide, then combine the evidence through prioritisation. Preserve unknowns rather than replacing them with zeroes.
