Search spending is usually discussed as a single channel, but the queries inside it differ enormously in what the searcher wants. That difference drives most of the cost variation.
Intent sits on a range
Some searches are informational, seeking an explanation. Others are comparative, evaluating options. A smaller set is transactional, naming a product or including words like pricing or near me.
Conversion rates rise sharply along that range, and so does competition. Transactional terms are the ones every competitor bids on, which is reflected in what they cost.
A budget concentrated at the transactional end buys efficiency and very little volume. Concentrated at the informational end it buys reach and a long lag before revenue.
Branded terms are cheap and misread
Searches containing a company's own name convert at high rates because the searcher already decided. Cost per acquisition on these terms looks excellent.
Much of that demand would have arrived anyway through organic results or direct navigation. The spend captures existing intent rather than creating it.
Companies that judge the channel on blended cost per acquisition therefore overstate its effect, because branded volume subsidizes the average.
Upper funnel spending pays late
Informational content and broad campaigns reach people months before a purchase. Attribution windows commonly close before that purchase occurs.
The measured result is that upper funnel activity looks unproductive, which leads to cuts. Demand at the transactional end then declines with a delay nobody connects to the cut.
This cycle repeats in many companies. The spending that fills the top is removed on evidence produced by a measurement window that cannot see its effect.
Competition varies by segment, not just term
Costs depend on who else is bidding. A term contested by well-funded national brands is expensive regardless of how well a smaller company targets it.
Narrower queries with clear qualifiers, such as an industry or a geography, often carry lower competition and higher relevance for a specialized seller.
Finding those pockets is slow manual work rather than an automated optimization, which is why it survives as an advantage for smaller advertisers.
Allocation should follow the sales cycle
A short-cycle consumer purchase can be served largely at the transactional end, because the interval between discovery and buying is small.
A long enterprise cycle cannot. Buyers spend months in the comparative range, and a company absent from it is not considered when the transactional search finally happens.
The split is therefore determined by how the category is bought rather than by which end reports better numbers in the current quarter.