The price increases were engineered — and you were meant to read them as noise

Search advertising did not simply get more expensive. A federal court found the increases were deliberate, sized to be mistaken for ordinary auction fluctuation, and that 85% of advertisers blamed themselves.

HOW IT PRESENTS

Cost per click drifts upward. You assume it is competition, seasonality, or something you changed. Almost every operator we meet has concluded the problem is on their side of the account.

THE EVIDENCE

What the court actually found

01

Google influences the final price through what it internally calls “pricing knobs” — and refers to using them as “intentional pricing.”

United States v. Google LLC, No. 1:20-cv-03010 (D.D.C.), Findings of Fact ¶243, opinion of 5 August 2024
02

Google “endeavored to raise prices incrementally, so that advertisers would view price increases as within the ordinary price fluctuations, or ‘noise’, generated by the auctions.” Internal documents describe a 10% CPC increase as “safe” because it is “within usual WoW noise.”

Same opinion, FOF ¶264, quoting Google’s own internal documents
03

Google’s own studies found advertisers “dominantly attribute[d] these shifts to themselves, competition[,] and seasonality (85%) — not Google.”

Same opinion, FOF ¶266
04

Increases were applied “often between 5% and 15% at a time, without a significant shift in advertiser spending.” Text-ad cost per click roughly doubled between 2013 and 2021.

Same opinion, conclusions of law p.259; FOF ¶186 (plaintiffs’ economic expert, trial testimony)
05

The mechanism is structural as well as deliberate: “Because broader matching enters more advertisers into an auction, it leads to thicker auctions… which creates upward pricing pressure.”

Same opinion, FOF ¶278

SIZED AS NOISE

Why the increases were sized to look like noise

The increments were chosen so they would read as ordinary auction movement rather than as a decision. The findings quoted above are the mechanism in the platform's own words, and the last of them is the part worth sitting with: the great majority of advertisers attributed the movement to themselves, their competitors or the season.

HOW IT PRESENTS IN YOUR ACCOUNT

How this shows up in your own account

If your cost per acquisition has drifted, the first hypothesis should not be that your account degraded. A court has found that the platform raised prices in increments specifically calibrated so you would attribute them elsewhere — and that it worked on roughly six advertisers in seven. That does not make the increase reversible, but it does change what you should be measuring and what you should be blaming.

REMEDY AND LIMITS

What to do about it, and what cannot be undone

What cannot be undone: the price paid. Historical spend is spent, the auction cleared at the price it cleared at, and no operational change recovers it. Anybody offering to is selling something.

What can be done is narrower and worth doing. Rebuild your own cost history from the ledger rather than from platform reporting, so you can see what actually happened to your cost per new patient over the period rather than what a dashboard said about clicks. Separate the price effect from the volume effect — a rising cost with flat conversion is a market condition; a rising cost with falling conversion is your own funnel, and only one of those is anybody else's fault. Find out how much of your spend is reaching people who already knew your name, because that is the portion most exposed to auction price and the least productive. And plan on the assumption that the mechanism has not been reversed by anything published since.

The reason to read this is not grievance. It is that an operator who blames himself for a market-wide price movement makes worse decisions about next year's budget than one who has priced it correctly.

WHAT WE ARE NOT CLAIMING

We do not claim costs rise every year. Alphabet’s own filings show cost per click falling in several years — reported CPC is a blended figure distorted by device and product mix. The court findings describe the auction itself, which is why they are the stronger evidence.

PROVENANCE

Sources

Every figure above is public, cited, and checkable. None of it is a client's numbers — those stay sealed. How we rank a source, and which sources we refuse to use, is set out at /sources.

APPLY — 6 / QUARTER →