CAC starts climbing after share of search turns.
Every performance team I have sat with has the same explanation for a rising cost per acquisition. The auction got more competitive. It is the most reassuring sentence in the discipline, because it blames no one in the room and it comes with a ready fix: ask for more budget. Sometimes it is even true. More often it is a brand problem wearing a media costume, because the whole thing shows up in metrics a performance buyer already trusts.
I spent years watching that number climb and calling it the auction. Here is what was actually happening, in the dashboard we stared at every day.
The auction charges you for being unknown
Google’s ad rank rewards expected click-through, so a brand people recognise pays less for the same position and an unknown one pays a premium. The auction prices your familiarity.
When you bid on a keyword you do not simply pay your bid. Ad rank folds in the click-through rate the system expects your ad to earn, and a brand people recognise gets clicked more. That expected click-through feeds Quality Score, and a higher Quality Score lowers what you actually pay for a given slot. A brand nobody has heard of clears the same auction, for the same position, at a higher price.
That gap is not waste you can trim out of the account. It is the market putting a price on your familiarity. A weak brand pays more per click before anyone touches a bid.
Branded search is the tell
The cheapest, warmest traffic you own is people typing your name. When that stream thins, you make up the volume on generic terms where you have no advantage, and blended CAC climbs.
The cheapest and highest-converting traffic in almost any account is people searching the brand name. They already decided somewhere else; the click is a formality and it costs pennies. That volume is a direct readout of demand you built earlier.
When branded search flattens or slips, the shortfall does not go away. You cover it on generic, non-brand terms, where you are one of eight near-identical bidders and every click is a cold stranger who has to be convinced from scratch. Blended cost per acquisition rises, and the mix shifted underneath it while the headline bid stayed put. Les Binet’s work on share of search found that a brand’s share of category search usually moves ahead of its market share, so a falling share of search is a warning you can read months before the revenue confirms it.
A cold audience converts worse, and that is CAC too
Conversion rate is a brand metric in disguise. Warm traffic converts on the first visit; cold traffic needs more touches and a discount, so the same media buys fewer customers.
Cost per acquisition is spend divided by customers, so anything that drags the conversion rate down pushes CAC up without a single cost-per-click moving. A visitor who already trusts you converts on the first visit. A visitor who has never heard of you comparison-shops, waits, needs three touches and a code before they commit. Same media, fewer customers per pound. On the dashboard this shows up as a conversion-rate problem, so you go off testing button colours, when the thing that actually moved is whether people arrive already wanting you.
The auction is a mirror
Higher CPCs on your own name, a thinning stream of cheap branded clicks, a colder audience that converts worse: each is your brand strength, priced in real time.
Put those together and the auction is reporting on you. A strong brand carries a standing discount into every auction it enters. A weak one pays what I have come to think of as the weak-brand tax, and it compounds: the erosion behind this quarter’s CAC is still working on next quarter’s share of search.
A strong brand carries a discount into every auction it enters. A weak one pays the difference, click by click.
Sometimes it really is the auction
Category-wide inflation and deep-pocketed entrants are real. The test is whether the whole category’s costs moved, or only yours, measured against your own share of search.
Not every rising CAC is a brand problem, and I am not going to pretend it is. A new, well-funded competitor can lift prices for a whole category. Platform changes and privacy-driven measurement loss can inflate reported CAC on their own, with nothing real moving underneath. So before you reach for the brand explanation, check whether costs rose across the category or only for you, and check it against your share of search. If the whole category got more expensive, it is the auction, and more budget really is the answer for now. If your costs climbed while your share of search slipped, the auction is just the messenger. The real test underneath all of it is incrementality: strip out the customers who would have found you anyway, and see what your media actually bought. I make that case in full in the first essay.
Why no lever fixes it
Fewer people know the brand than a year ago, so fewer arrive already sold, and the account pays for that without being able to fix it. Every lever you hold, bids, budgets, creative tests, audience refinement, works on the people already showing up, and the cause sits upstream of all of them. You can run the most disciplined account of your career and still watch CAC climb, because what made your clicks cheap was people knowing the name before they searched.
The levers and the cause are in different places.
Pull last year’s share of search before anyone signs off more budget. It costs an afternoon. If it fell while costs rose, no amount of bid tuning brings it back: you are paying for a brand you stopped building, and the auction is only the invoice clerk.
Notes & references
- Google Ads, how ad rank and Quality Score work: expected click-through rate, ad relevance and landing-page experience determine Quality Score, which affects the price paid for a given position.
- Les Binet, share-of-search research for the IPA (from c.2020): share of category search correlates with, and tends to lead, market share. See also the IPA effectiveness work with Peter Field on brand and activation.
- On incrementality and holdout testing, and the limits of last-click attribution, see the first essay in this series, Performance harvests demand.