What to actually look at in a Google Ads report

The report is fifteen pages and it is sitting on the table between us. Impressions up 22 percent. Click-through rate up a tenth of a point, highlighted in green. A pie chart of device split. Four pages of keyword tables. The owner has read all of it and asks the only question that matters, which is whether he should put more money in next month, and nothing in those fifteen pages answers it.

We have been in that meeting more times than we would like. The report was not dishonest and nobody padded it on purpose. It contained every number the platform exports, arranged neatly, because that is what the export does and because thoroughness is easy to demonstrate.

A simple test: if a number goes up and you would not act differently, and it goes down and you would not act differently, it is decoration.

There are only five decisions

Spend more. Spend less. Move money between campaigns. Change the offer or the page. Turn something off.

That is the full list for a paid search account. Every row in a report should be there because it feeds one of those five. Read your last report backwards and ask of each line which decision it informs. Whatever survives is your real report.

This is why a report that opens with impressions and click-through rate is already telling you something about who wrote it. Both numbers are real and measurable. Neither tells you whether to spend more money.

What survives

Cost per acquisition, split by campaign and device. The blended figure hides the campaign quietly taking a third of the budget.
Conversion volume against your capacity. Thirty leads a month is excellent or catastrophic depending on how many your team can actually ring back the same day.
Search impression share lost to budget, and lost to rank. Different signals with different fixes. Lost to budget means demand exists and you are declining to buy it. Lost to rank means the auction has judged your bid or your relevance and found it short.
The cost per acquisition trend across ninety days. A single month contains too much noise in any account below serious volume, and month-on-month comparison invites you to explain randomness.
What happened to the leads, from your own records. Not from Google.

What is actually in the Conversions column

Two things have changed underneath that column, and most reports do not mention either.

First, attribution. Google Ads now offers only two attribution models, last click and data-driven, and data-driven is the default for most conversion actions. First click, linear, time decay and position-based are gone, and conversion actions that used them were migrated to data-driven automatically. If your year-on-year comparison straddles that migration, you are comparing two different accounting methods and calling the difference performance.

Second, modelling. Where consent mode is running and the account clears Google’s quality bar (700 ad clicks over seven days, per country and domain grouping), Google estimates the conversions it could not observe, by learning from the users who did consent. Those estimates appear in the Conversions column alongside the observed ones, with no label distinguishing them.

We genuinely cannot decide how we feel about this. The alternative is worse: without modelling, the column simply under-reports by whatever share of your visitors declined cookies, and bidding optimises against a distorted picture. Google is probably better at estimating that gap than we would be. And we are being asked to make budget decisions from a figure that is part measurement and part inference, in a column that does not say which part is which, produced by the company selling the clicks. Both of those statements are true at the same time and we have not found a way to stop them both being true.

What we do about it: treat the Conversions column as the steering input for bidding, and your own sales records as the scoreboard. Never let the platform grade its own work.

What wastes your attention

Click-through rate is the most over-discussed number in paid search. It earns its place in exactly one comparison: two ads for the same query in the same ad group. Across campaigns it means nothing, because a brand campaign will always beat cold prospecting and that tells you only that people know your name.

Quality Score is a diagnostic rather than a target. Worth a look when one ad group sits far out of line with the rest, because that usually points at a mismatch between the query and the page it lands on. Chasing it account-wide is work that looks like work.

Conversion rate alone is the one that does real damage. Loosen the definition of a conversion and it rises. Count every phone-number click and newsletter signup and you can double it inside a week without earning a lev. It only means something sitting next to what a conversion is worth, which is the argument for tracking real inquiries rather than events.

Optimisation score belongs in the same bin. It measures your compliance with Google’s recommendations, which is a different thing from your account’s profitability, and the two diverge most sharply in small accounts.

The lines nobody puts in the report

The important numbers are the ones Google cannot see. In the lead generation accounts we work on, the difference between a good month and a bad one sits on the sales side far more often than in the auction. Same cost per lead, half the close rate, and the ads get the blame.

Add three columns to your monthly review: leads received, leads actually reached, deals closed. By campaign if you can. Once you can see that campaign A produces cheap leads who never pick up and campaign B produces expensive ones that close, the budget decision has already made itself, and no amount of in-platform optimisation would ever have found it. This is most of what we do in a marketing optimisation engagement, and it is usually a spreadsheet rather than anything clever.

A one-page review

A fixed window. Last 30 days against the previous 30, plus the same 30 days last year where the business is seasonal.
One row per campaign: cost, conversions, cost per conversion, impression share lost to budget, impression share lost to rank.
Underneath, from your own records: leads, contacted, quoted, closed, revenue.
A short list of what changed in the account during the period. Bid strategy edits, new assets, budget moves, page changes.
One sentence per campaign stating what you will do about it. If that sentence reads “keep monitoring” for the third month running, the campaign needs a decision rather than another month.

Twenty minutes a month, one page.

Point four is the one people skip and the one we would keep if we could only keep one. Change history is the least-used screen in the interface. When a metric moves, the first question is what we did and the second is what Google did, and both leave a trail. Read an outcome without its cause and you will invent a cause, usually a flattering one, and then reorganise the account around it next quarter.