Display advertising has a traffic quality problem that most dashboards hide. This is a field report from a campaign where we could see it clearly. If you want to build a curated list like the one described here, the practical inclusion-list guide covers the method and includes a worksheet.
A large European transportation brand with multi-million euro annual display budgets had been running GDN campaigns for years on a placement list that had never been systematically audited. We cleaned that list. Then we ran two concurrent campaigns: one on the curated list, one on the historical placements. Same brand, same period, different placement environments.
The cleaned campaign got 52% fewer visits. It also generated 7x more revenue per visit — same brand, same period, same creative.
Here's what we removed, and why we think it mattered.
What was in the placement list
When we audited the brand's historical placement data, we found three broad categories of problematic domains. Each is worth understanding separately, because they cause different kinds of damage.
Made-For-Advertising sites. The most well-documented category. Sites built to generate ad revenue rather than serve real readers: thin or AI-generated content, excessive ad density, clickbait headlines engineered for passive clicks. On a large GDN placement list that has never been reviewed, these are routine. We identified and removed hundreds.
Clickbait aggregators and news imitators. Distinct from pure MFA sites in that they have more content volume, but their audience is there to be entertained or outraged, not to research a purchase. High traffic, zero intent. A user who clicked an ad while scrolling a "you won't believe" article is not in a buying mindset. The click looks identical to any other in the dashboard. The person behind it is not.
Copycat and ad arbitrage sites. This is the category that surprised us most, and it's almost never discussed in the brand safety literature.
For a transportation brand, there exists an entire ecosystem of sites that imitate or parasitically target the brand's own product category. These sites appear in search results for travel-related queries, look like legitimate booking resources or comparison tools, and attract users who are genuinely close to a travel purchase. Their business model is capturing that high-intent audience and then monetizing it through display advertising. Including ads from the original transportation brand.
The result is something close to circular waste: the advertiser was paying to run ads on sites designed to intercept their own potential customers. A user arrives at a copycat travel site through a search or recommendation, sees a display ad for the original brand and clicks through, but they arrive in a disrupted state, on a page that had already partially muddied their intent. They were never cleanly yours to begin with.
Brand safety risks. Politically sensitive, controversial, or reputationally damaging content. Removed.
What remained after cleaning was a list of authoritative editorial sites, reputable publishers, and content environments with genuine, engaged audiences.
The data
We want to be transparent about the methodology upfront. The two campaigns ran with different optimisation goals. Because of that, we're not comparing total traffic volumes or costs, where the optimisation strategy has a direct effect. We're looking at what happened downstream: what the traffic actually did once it arrived on site.
Compared to the uncleaned campaign, visitors from the curated placement campaign:
- Generated 7x more revenue per visit
- Converted at 4x the rate (visits to orders)
- Produced 2.5x more orders in total
- Contributed 3.3x more total revenue
- Spent 32% more per order
- Spent 6% more time on site per visit
- Bounced at a 5.5% lower rate
The cleaned campaign drove fewer total visits. Every downstream signal says those visits were worth far more.
Why we think this happened
Three mechanisms, probably operating simultaneously.
Invalid and bot traffic removal. MFA sites have well-documented bot and invalid traffic problems. High impression and click counts don't represent real humans. Removing these placements removes a layer of artificial engagement that inflates click metrics while producing zero downstream value. This explains part of the conversion gap, but not all of it.
Click intent mismatch. Clickbait and passive-scroll environments generate clicks from users who are not in a purchasing mindset. The click looks the same in the platform. The person is in a completely different mental state. Curated placements on editorial content tend to produce clicks from users who were already engaged: reading something, considering something, researching something. That intent carries through to the landing page.
The copycat site effect, and why it explains the basket value gap. Bot filtering and intent mismatch explain higher conversion rates. They don't fully explain why whitelist buyers spent 32% more per order. That's not a fraud problem. That's an audience quality problem.
Our interpretation: the copycat and ad arbitrage site category attracted users who were genuinely interested in travel, but arrived via a detour that fragmented their intent. They may have already partially engaged with a competitor-adjacent product, or arrived in a comparison mindset rather than a booking mindset. Users arriving from authoritative content environments such as travel journalism, news and reputable publishers came in with cleaner intent and completed higher-value bookings.
This is speculative. But it's consistent with the data, and we think it's an underappreciated mechanism in how placement quality affects not just conversion rate, but purchase value.
What this comparison can and cannot show
Start with the part that is not in doubt, because it is the finding: traffic halved and sales went up. The curated campaign took 52% fewer visits and still produced 2.5x the orders and 3.3x the total revenue. That is the opposite of what almost everyone expects when traffic falls, and it holds regardless of how you attribute it. Cutting volume usually costs you sales. Here it did not, because the volume that disappeared was not doing anything.
What the comparison does not isolate is how the gain splits between the two things that changed. The curated list was one; the optimisation goal was the other, and a different goal buys different people. So this shows what the package did, not how much of it the placement list earned on its own. Separating them needs the same goal, the same budget, the same period and a geographic split, which takes more volume than most accounts have.
That distinction matters mainly if you are trying to forecast. The direction is well supported: six independent downstream measures moved the same way, which is harder to explain by audience differences alone than any single one of them would be. The multiples are the shape of the effect in one account, not a number to put in your own plan.
It is the same habit worth applying to brand search. A brand campaign reporting a return of 300 is reporting something real, and still says nothing about how much of that revenue would have arrived without it. Numbers from a single arrangement describe that arrangement. The way to find out what a change is worth is to change one thing and watch what actually moves.
What this means for display advertisers
The standard metrics of impressions, clicks, CTR and CPM do not tell you whether the traffic you are buying is capable of converting. A high-CTR placement list looks healthy. It may be producing the worst-quality traffic in your mix.
The metrics that reveal placement quality are downstream: conversion rate, revenue per visit, bounce rate, time on site, average order value. Most advertisers don't cut these metrics by placement. Most placement lists never get audited at all.
The case for placement auditing has been made before, usually in terms of brand safety and fraud prevention. What this data adds is a more specific mechanism: the copycat and arbitrage site problem is structural, not incidental. For any brand with a recognizable product category, whether travel, financial services, insurance or retail, there is almost certainly an ecosystem of sites built to intercept your potential customers and sell them back to you as ad impressions. Standard brand safety tools, most of which were built for programmatic DSP buying rather than GDN's placement model, don't catch this.
If your GDN or PMAX placement list has never been systematically reviewed, it almost certainly contains all three categories above. The ratio varies. The direction doesn't.
DisplayGateGuard audits GDN and PMAX placement lists for MFA risk, brand safety, and placement quality. Analyze your placements free with 100 credits →