Vanity Metrics in Marketing Reporting: What to Refuse
A number belongs in a report only if a plausible change in it would change a decision. Impressions, clicks, reach and score-style metrics almost never pass that test as headline results, but most of them are useful diagnostics inside the account. The distinction that matters is not which metric it is, it is what job you are asking it to do: a diagnostic answers "why did that happen", and a result answers "what did we get for the money".
Almost everything written on this subject hands you a list of banned metrics. The list is not wrong, and it also does not survive the meeting, because the person asking for impressions is asking for a reason and a list does not give you anything to say back. What follows is the test, the small number of things worth refusing outright, and the language for refusing them without sounding evasive.
This page is about paid campaign reporting: what goes in the monthly deck, what goes to somebody who does not run campaigns for a living.
The test, in one question
Before a number goes on the page, ask: if this moved 30% in either direction, what would we do differently?
If the answer is a specific action, it is a result or a live diagnostic and it belongs somewhere. If the answer is "we would talk about it", it is decoration. If the answer is "we would investigate", it is a diagnostic and it belongs in a tool rather than in a report that nobody can act on while reading.
Three failure modes fail that test for different reasons, and they need different handling.
| Failure mode | Example shape | Why it fails | What to do instead |
|---|---|---|---|
| Measures activity, not outcome | Impressions, reach, posts published, hours spent | Rises with effort regardless of result | Report the outcome the activity was supposed to produce |
| Measures a step with no known relationship to the end | Clicks alone, sessions alone | Real, but not connected to anything anyone buys | Report it only as part of a path with the end attached |
| Looks like an outcome and is not | Score-style metrics, blended totals with no definition, percentage change on tiny numbers | Reads as a result and cannot be acted on | Refuse it, and explain the mechanism |
The third row is where the money is, because the first two are already on everybody's list and the third one is not.
Diagnostics that get promoted to results, and should not be
The most common reporting error in this territory is not somebody reporting likes. It is somebody reporting a genuinely useful internal diagnostic as though it were a performance outcome.
Quality Score is the clearest case, and Google says so itself. Its documentation states that "Quality Score is a diagnostic tool meant to give you a sense of how well your ad quality compares to other advertisers", that it "is not an input in the ad auction", and, most directly, that it "is not a key performance indicator and should not be optimized or aggregated with the rest of your data" (Google Ads Help, "About Quality Score", checked 2026-09-07). An account-average Quality Score on a monthly slide is a number Google explicitly says not to aggregate. What the score is good for, and how to read the component statuses that actually tell you something, is in what Quality Score is and is not.
Impression share is the second case. It is a genuinely valuable diagnostic when you are answering a specific question about constraint at a specific moment, which is why impression share lost to budget is worth understanding. As a trend line reported to people who cannot act on it, it produces the wrong conversation, because the natural response to a low number is "spend more" regardless of whether the traffic being missed was worth having.
Search terms are the third. The search terms report is one of the most useful screens in an ad account and one of the worst report tiles, because its value is in reading the long tail and acting on it, not in a count of how many terms appeared.
The pattern: a diagnostic is read by someone who can change something in the next hour. A result is read by someone deciding whether to keep funding the work. Moving a diagnostic into the second context does not elevate it, it just guarantees a bad instruction.
Numbers to refuse outright
Short list, and each one is refused for a mechanism rather than a preference.
A conversion total summed across platforms. Each platform can legitimately claim the same sale under its own crediting rules, so the sum overlaps and exceeds reality. There is no correct way to add them. Report per platform with labels, or report the figure from the system that issues invoices.
Any conversion number without its source and date basis attached. The same word covers different objects in different products, on different clocks, as key events, conversions and the words GA4 keeps renaming sets out. A bare "Conversions: 214" is not a fact, it is a fact-shaped object waiting for an argument.
A percentage change on a small denominator. Going from three leads to five is a 67% increase and is also probably noise. Where volumes are small, report the counts and let the reader see the size, because a percentage hides it deliberately.
An average that hides a segment. A single blended cost per result across campaigns with wildly different economics is the most professional-looking way to make a bad campaign invisible. If the segments differ materially, the average is the number to refuse.
A channel breakdown with the unattributed portion quietly removed. Every breakdown has a bucket where the source was not recorded, and it is usually not small. What is inside it is described in what direct traffic in GA4 actually is. Showing the breakdown without that bucket overstates everything else on the chart in a way nobody can see.
What to say when someone senior asks for impressions
Refusing a metric in a document is easy. The reason this page exists is that you have to do it in a room, to somebody who signs off your budget, and "that's a vanity metric" is a phrase that makes you sound like you are avoiding a question.
Three things work better than the label.
Ask what the number would be used for. Very often the request is a proxy for a real question, usually "are we still visible" or "is the money doing anything". Answer that question with a number you trust and the request for impressions disappears. Occasionally the answer is that a brand awareness objective genuinely exists and was never written down, in which case impressions are not a vanity metric at all, they are the closest available measure of the agreed goal, and the honest move is to say so.
Report it, and attach the consequence. "Impressions rose 40%, results were flat, so we reached more people who did not want this." That sentence uses the number the person asked for and still moves the conversation to the decision. Withholding it just makes you look defensive.
Name what the number cannot do, before recommending anything. A metric with a stated limit is more persuasive than a metric presented as clean, and it protects you when it later moves for a reason nobody predicted.
The honest counter-argument
Two things every article on this topic gets wrong by omission.
Not all upper-funnel measurement is vanity. If the goal is genuinely awareness, then reach and impressions are the measure of the goal, and dismissing them is as sloppy as reporting them without context. The failure is not the metric, it is reporting a proxy for one objective as though it were the result of another.
Some things that matter cannot be counted at all. Whether a sales team trusts the leads it is being sent, whether a message is landing with the people you want, whether an offer is right: none of these produce a number, and a report that contains only countable things will systematically underweight them. The answer is a written sentence next to the numbers, not a metric invented to fill the gap.
Nothing in this article is a claim about your account, and we have no client data or test to cite here. The one product statement above is quoted from Google's own documentation and dated to 2026-09-07. Google renames and reframes reporting metrics on its own schedule, so check that page before quoting it much later.
FAQ
What is a vanity metric in marketing reporting? A number that reliably moves with effort rather than with results, and that nobody would act on differently if it changed. The practical test is whether a large move in it would change a decision. If not, it is decoration whatever it measures.
Are impressions always a vanity metric? No. Impressions are a diagnostic inside an account and a legitimate measure when awareness is the agreed objective. They become a vanity metric when they are presented as the result of a campaign whose actual goal was something else.
Should I report Quality Score to management? Google's own documentation says Quality Score "is not a key performance indicator and should not be optimized or aggregated with the rest of your data" (checked 2026-09-07). It is useful for diagnosing a specific ad group, not for describing performance to somebody who cannot change one.
What should replace the metrics I remove? One number per question the audience actually asks: what was spent against plan, what it produced, what it cost per result with a stated definition, and what changed against a comparable period. Plus one sentence saying why.
My boss wants a metric I think is meaningless. What do I do? Give it to them with the consequence attached rather than refusing it. Ask what decision it supports, answer that underlying question with a number you trust, and state plainly what the requested metric cannot tell either of you.
Sources, read 2026-09-07
Google Ads Help: "About Quality Score" (support.google.com/google-ads/answer/6167118).
Reporting metrics get renamed and reframed regularly. This page describes the test rather than a fixed list for that reason.