CPC

Search advertising grew out of the generalised second-price auction, and the intuition from it still holds even though the mechanics have become far more complicated. You do not pay your bid. You pay roughly the minimum required to keep your position against the advertiser below you, which is why actual CPC almost always sits under max CPC and why raising a bid does not necessarily raise what you pay.

The comparison is not between bids alone. Ad Rank combines the bid with ad quality, the Ad Rank thresholds, the competitiveness of that particular auction, the context of the search and the expected impact of assets. An advertiser with better quality can hold position above one bidding more, and pay less for it.

The practical version of that: improving relevance is the only lever that lowers cost without losing volume. Everything else lowers cost by buying less: cutting bids, narrowing match types, pausing keywords.

Auctions are also recomputed per search rather than set per keyword. The same term can cost very differently across device, location, time of day and audience signals, so an average CPC is an average over a wide distribution rather than a price.

Most CPC movement originates outside the account. A competitor entering the market, a rival raising budgets for a product launch, a seasonal peak in the category: all of these raise what it costs to win the same auction, and nothing on your side changed.

Category seasonality is the most predictable version and still catches agencies out in client conversations. Insurance in January, retail through Black Friday and into December, fitness in the new year, legal services after a regulatory change. CPC in those windows reflects who else turned up.

Automated bidding introduces a second source of movement that looks the same but is not. Target CPA and Target ROAS deliberately pay more for clicks they predict will convert, so an account whose conversion data has improved will often show rising CPC and improving results simultaneously. Reading that as a problem and cutting bids undoes the gain.

The diagnostic that separates the two is simply whether cost per conversion moved with it. CPC up, CPA flat: the system is paying more for better clicks. CPC up, CPA up: something outside the account got more expensive, or something inside it got worse.

CPC is an input, not an outcome, and treating it as a target has a reliable failure mode. The cheapest clicks in any account are the least commercially valuable ones. Broad informational queries, low-intent placements, audiences nowhere near a purchase. Push average CPC down and you have systematically shifted budget toward them.

It happens most often when CPC appears prominently in a client report without a cost-per-result figure beside it. The client reasonably asks why clicks are getting more expensive, the agency reasonably responds by lowering bids, impression share on the converting terms falls, and two months later conversions are down for a reason nobody connects to the original question.

There is one case where CPC is genuinely the metric: campaigns whose objective actually is traffic, such as content distribution or top-of-funnel awareness where no conversion is expected. Even then, cost per engaged session is usually the better measure, because a click that bounces in two seconds cost the same as one that did not.

A search click from someone typing "emergency plumber near me" and a display click from someone who mis-tapped an ad on a news site are both one click, and they are not the same purchase. Comparing their cost tells you nothing except that one channel intercepts intent and the other creates awareness.

The same applies within a platform. Branded search is cheap because nobody else can be as relevant to your client’s own name; generic head terms are expensive because everyone in the category is bidding on them. An account average across both is a number that moves whenever the mix moves.

The comparison that means something is CPC for the same campaign against its own recent history, or CPC across competitors for the same keyword, which is what auction insights is for.