Ad Rank

Ad Rank is the score that orders the ads in a given auction and decides which of them appear at all. Google documents six inputs, and it is worth reading them as a list rather than as a formula, because the weighting is not published and has changed repeatedly.

The structural point is that only one of the six can be changed from the bidding screen. The rest are either properties of the ad and the landing page, which take work to move, or properties of the search itself, which cannot be moved at all.

Thresholds are the part that surprises people, because they break the mental model of advertising as a ranked list. An ad is not shown last when it is worst. It is not shown at all when it fails to clear the reserve for that context.

This is the mechanism behind the most common panicked client email an agency gets: the ad stopped serving and nobody touched the account. The threshold for that context moved, or the ad quality feeding into it slipped, and eligibility went with it.

Thresholds vary by position too, which produces behaviour that looks irrational until you know the cause. An ad can be eligible for the bottom of the page and not the top, so the same keyword at the same bid shows for some searches and not others.

They exist because Google is protecting the results page rather than maximising revenue on any single query. A weak ad that clears on bid alone costs more in user trust than it earns, so there is a floor it cannot buy past.

The intuition that whoever pays most wins comes from media buying, where inventory really is sold that way. In an auction weighted by quality it is close to backwards: the advertiser with better quality clears the threshold at a lower bid, so they can hold the position more cheaply than you can take it.

What bidding more reliably does is raise costs in every auction you were already winning, because it applies to all of them and not only the contested ones. Agencies that answer a competitor by lifting bids across the board usually find CPC up within days and position roughly where it was.

The lever that compounds is relevance. Tighter ad groups, copy that matches the query, a landing page that delivers what the ad promised. Each lowers the bid required to clear the same threshold, and it keeps working after the competitor leaves.

None of that makes bid a bad lever. It makes it the right lever for something short and specific, such as defending a brand term through a launch, and the wrong one for a standing problem.

Google removed average position in September 2019, and the reason it went is the reason not to reconstruct it. It described where an ad sat among the ads shown, not where it sat on the page, so an average position of 1.0 was perfectly compatible with appearing at the bottom every single time.

The replacements describe the page. Top impression rate is the share of your impressions that appeared above the organic results, and absolute top impression rate is the share that landed in the very first slot. Between them they answer what a client is actually asking when they ask about position.

Paired with impression share, they answer the follow-up as well. Impression share says how often you were eligible and showed; the lost-to-rank and lost-to-budget splits say which of the two problems you have.

This matters for agency reporting in particular. A client comparing your report against someone who still quotes average position will assume you removed something. Explaining once, in writing, that the metric was retired is cheaper than explaining it every quarter.