Quality Score

Google documentation is unusually direct on this point: Quality Score is intended to help diagnose whether ads and landing pages are relevant, not to be optimised or reported as a performance measure. It is not used directly in the auction either; what enters the auction is a real-time quality assessment for that specific search.

The consequence is that an account-level average Quality Score is a number with no operational meaning. It aggregates a diagnostic across keywords with wildly different volumes and purposes, and produces a figure that cannot be acted on and will nonetheless get asked about every month once it appears in a report.

The useful reading is always component-level. Each of the three is reported as below average, average or above average, and the point is to see which one is dragging. A keyword at 4 out of 10 because of landing page experience needs an entirely different intervention from one at 4 because of ad relevance.

Expected click-through rate predicts how likely the ad is to be clicked when it shows for that keyword, adjusted for position and format so a low historical CTR from a poor position does not count against you twice. It is the component most responsive to ad copy that matches the search.

Ad relevance assesses how closely the ad text matches the intent behind the keyword. It is the one most often fixed by structure rather than writing: an ad group containing twenty loosely related keywords cannot have copy that matches all of them, and splitting it is usually the whole fix.

Landing page experience assesses whether the page is relevant, useful, easy to navigate and reasonably fast. It considers the page the click actually reaches, not the domain in general, which is why sending every campaign to the homepage damages it across the account.

Quality feeds Ad Rank, and Ad Rank decides both whether an ad shows and what a click costs. Because quality is a multiplier rather than an afterthought, an advertiser with a better assessment can outrank one bidding more, and pay less for the position.

That is the commercial case for the unglamorous work. Every other way of lowering cost per click works by buying less traffic: lower bids, tighter match types, fewer keywords. Improving relevance lowers what you pay for the traffic you already have.

It also compounds slowly rather than immediately. The components are built from history, so a rewritten ad needs impressions to accumulate before the assessment moves. Expecting a change the same week is the most common reason people conclude the work did not help.

Landing page experience sits on the client side of the boundary in almost every agency relationship. The agency runs the ads; the client owns the site, the developer and the release schedule. When this is the weak component, the agency can diagnose it and cannot fix it.

The failure mode is predictable. The issue gets mentioned once in a monthly report, nothing happens because nothing was asked for specifically, performance stays flat, and six months later the agency is explaining underperformance caused by a page it was never able to change.

The version that works is to raise it as a named dependency in writing, with the specific page, the specific problem, and what would need to happen. That converts it from a complaint into a decision the client has either taken or declined to take, which is a very different position to be in at renewal.