What 14,290 placements actually look like: anatomy of an unaudited display account

Almost everything written about made-for-advertising sites is written in the abstract. Percentages with no account behind them, or vendor numbers you cannot check. So here is one real placement report, pulled apart in public.

One caveat before the numbers: this is a single account, not a study. A German advertiser, mixed Display and PMAX, two months. It is not representative of your account, and I have no idea whether it is representative of anything. It is one honest look at what a placement report contains when nobody has ever audited it.

No domains are named. The findings are patterns and totals.

The account

By any normal reading this is a healthy account. 1,244 conversions at €7.96 is a result most advertisers would sign for. Nothing on the dashboard says anything is wrong.

Finding 1: most of a placement report is not about money

8,974 of the 14,290 placements, or 62.8%, cost nothing at all.

They took impressions and were never clicked. Only 5,316 placements had any cost attached.

This is the first thing that surprises people who open a placement report expecting a bill. Most rows are not a bill. They are a record of where your ad was shown, and the great majority of those showings cost you nothing directly.

It has a practical consequence: if you sort by cost and audit the top of the list, you will finish very quickly and learn very little about your inventory. Two-thirds of your placements do not appear there at all.

It also means the raw row count is a useless measure of anything. "We have 30,000 placements" describes a spreadsheet, not a problem.

Finding 2: the money is concentrated, and so is the temptation to stop early

Fewer than 300 rows out of fourteen thousand carried two-thirds of the budget. That is genuinely useful: it means a partial audit is worth doing, and it is why our free audit starts at the top of the list rather than pretending to do everything.

Now look at the other end. 4,064 placements cost under €1 each, and together they came to €1,435 — about 14% of total spend, spread so thinly that no individual row would ever catch your eye. This is where made-for-advertising inventory lives. No single row is big enough to notice, which is precisely why nobody notices them.

Individually each of these looks like a rounding error. Four thousand of them is 14% of the budget.

Finding 3: €3,644 produced nothing

4,759 placements took real money and returned zero conversions. Together: €3,644.23, or 36.8% of total spend.

That number needs handling carefully, and I want to be honest rather than dramatic about it.

Zero conversions on a placement is not automatically waste. It is entirely normal for upper-funnel activity, for products with long consideration cycles, and for placements that contribute to a decision made somewhere else. Attribution windows hide real contribution constantly. Anyone who tells you 36.8% of this budget was thrown away is selling you something.

But it is also true that you cannot tell the difference from the dashboard, and almost nobody looks. A placement that genuinely assists and a placement that is pure arbitrage produce an identical row: some spend, some clicks, no conversions. The only way to separate them is to look at what the site actually is.

That is the whole argument for auditing placements, and it is a narrower argument than the industry usually makes.

Finding 4: the clearest warning sign was a click-through rate that was too good

This is the finding I did not expect, and it is the one worth taking away.

The account averages a 1.27% click-through rate. Here are the five most extreme placements, anonymised:

On that placement, 174 of every 207 people who were shown a banner clicked it. Display advertising normally works at fractions of a percent, because people do not click banners on purpose. A rate like this means the ad is placed where it gets hit by accident, jammed against a menu item or under a close button on mobile, or that something automated is clicking it. Either way you paid for every one of those clicks, and not one of them converted.

If you come from search advertising this is counter-intuitive to the point of being uncomfortable. High CTR is the thing you optimize toward. On display it is frequently a symptom.

Two of those five placements were sites whose entire purpose was giving something away free, the kind of page that exists to be clicked off. None of them produced a single conversion.

Finding 5: what the flagged placements had in common

Our scoring flagged 444 placements (3.1%), of which 36 were high confidence. Together they accounted for €1,758.82, or 17.8% of spend.

Grouped by what they actually were, without naming any:

• Free-download and free-content farms. Ringtones, wallpapers, file downloads. Pages that exist to be arrived at and left. These produced the most extreme click-through rates in the entire account.
• Job-alert aggregators. A large category by volume, particularly in South Asian traffic. Real sites, real visitors, no relationship whatsoever to what this advertiser sells.
• Foreign-language content farms. Several with 30%+ click-through rates and no conversions, serving audiences the campaigns were never meant to reach.
• Cheap-TLD domains. .top, .live, .fun, .icu. Individually meaningless, but heavily over-represented among everything else that was wrong.
• Parked and for-sale domains. Pages with nothing on them at all, still taking impressions.

Notice what is not on that list: nothing offensive, nothing extremist, nothing a brand safety vendor would flag. Every one of these would pass a standard brand safety check, because there is nothing unsafe about a ringtone site. They are a suitability and quality problem, not a safety problem, and that distinction is where most display budget quietly goes wrong.

What we deliberately refused to count

One finding came out of building this, not from the data.

An earlier version of our scoring treated numbers in a domain name as suspicious. The reasoning was that auto-generated domains often carry digit runs. Against this account it flagged a well-known German business directory whose name is a phone number, an online language school, and a major car marketplace. In German, a trailing 24 is an ordinary naming convention, not a signal.

We removed digits from the model entirely. A number in a domain name tells you nothing about quality, and scoring it produced more false positives than every other signal combined. The full ruleset is published if you want to argue with the rest of it.

The general lesson is worth more than the specific one: a scoring model that flags things you can see are fine will be ignored entirely, and correctly so. Precision matters more than recall when a human has to trust the output.

What this does and does not tell you

It does not tell you that 36.8% of your budget is wasted, that a third of display spend goes to MFA sites, or anything else about your own account. This is one advertiser, one country, one vertical, two months.

What it does show is the shape of the problem. Most of the report is noise. The money concentrates at the top. The junk hides in a long tail of sub-euro rows. The loudest signal is an engagement rate that is too high rather than too low.

It also shows that none of this is visible from a campaign dashboard. Every number in the first table looks fine.

Doing this on your own account

Pull Reports → Predefined reports → Other → Campaign URL performance in Google Ads, over at least 90 days.

Then, in rough order of effort:

1. Sort by cost and read the domain names. If most are unfamiliar, that is the finding.
2. Sort by click-through rate, filtered to placements with at least 50 impressions. Anything several times your account average deserves a look.
3. Check how much spend sits below your top 100 rows.

The free placement report tool does all three in your browser, uploads nothing, and produces an exclusion list at the end. If you would rather have the sites actually visited and scored rather than pattern-matched, send it to us and we will audit the top 100 free.

Either way, look at the report. It is the only place where any of this is visible.