In digital advertising, we usually buy on performance, for example based on CPA. Broad targeting makes sense in that context. The algorithm receives a clear signal — a purchase or a lead — and finds the people most likely to convert. Reach and branding campaigns are different. We buy those on CPM, which gives the algorithm just one goal: as many impressions as possible for as little money as possible. It has no signal for quality.
As a result, the algorithm pushes you towards placements that are cheap or easily available. That’s why CPM is a poor metric when you’re buying junk, or even bots. On YouTube, we see this in the channels where ads appear. When buying reach, you need to think like a marketer again. Humans must remain in control of quality, not the algorithm.
Scale makes that difficult. There are an estimated 115 million-plus YouTube channels worldwide, around 60 million of which actively publish new videos (DemandSage, based on Social Blade). New channels appear every day. Assessing and configuring those channels one by one is practically impossible for a human. That’s precisely why most advertisers leave it to the algorithm — and precisely where things often go wrong.
On YouTube, the risk is greater than many advertisers realise. When buying open reach, you’re primarily buying into the long tail: hundreds of thousands of channels that nobody has selected and nobody monitors. This became clear when Draft Digital and Spektre inspected the placements of a Dutch direct-to-consumer brand, channel by channel. More than a quarter of a million channels over five weeks, with 99% of them receiving fewer than 100 impressions.
We reversed the buying approach: quality first, price second. The surprising result was that quality didn’t become more expensive — it became cheaper. Fewer channels, better context and almost half the cost per thousand people reached. In this article, we explain how we did it, what we learned along the way and what every YouTube advertiser can do tomorrow.
Why the Algorithm Sends Reach Campaigns in the Wrong Direction
CPM measures the price of an impression, not its value. An impression is a technical event: a video starts playing on a screen. The price doesn’t tell you whether a person is sitting in front of it, whether they’re watching, or whether the environment helps or harms the brand.
In a performance campaign, the results naturally correct this. A placement that doesn’t convert automatically receives less budget. In a reach campaign, that safeguard is missing. The algorithm only sees that an impression is cheap, and cheap impressions are often found precisely where nobody wants to advertise.
YouTube: The 45-Year-Old Woman Who Is Actually a Child
On YouTube, this is most evident in the channels themselves. A large share of cheap reach comes from children’s channels. Children often watch through their mother’s or father’s account. In the targeting data, they appear to be a 45-year-old woman — precisely the audience many brands want to reach. In reality, a child is watching the ad.
The targeting looks correct on paper and the CPM looks attractive, but the brand isn’t reaching anyone who will ever buy its product. On top of that, there’s mass-produced AI content without editorial oversight and traffic that may originate from the wrong IP addresses. In these environments, a low CPM isn’t an advantage; it’s a warning sign.
Think Like a Marketer Again
Attention is scarce. People see hundreds of ads every day and give each one only a few seconds on average. Brand recognition only begins after approximately two and a half seconds of viewing. A cheap impression that nobody consciously sees isn’t a cheap contact. It’s no contact at all.
That’s why we ask a different question when buying reach: how much does it cost to reach a real person, in the right context, who actually sees the ad? The algorithm can’t answer that question. The marketer can. It requires insight at the channel level and decisions made by people, rather than leaving those decisions to the system.
The Long Tail Is the Most Expensive Reach You Buy
Open targeting feels efficient, but the algorithm spreads your ad across an enormous long tail. In the campaign flight we analysed, the In-stream ad appeared on 264,129 channels. Of those, 261,338 received fewer than 100 impressions.
That long tail was also more expensive than the rest of the campaign. Its CPM was approximately 19% higher than the CPM across the remainder of the flight. You’re paying more for channels you don’t know, can’t assess and wouldn’t have chosen. That’s precisely the opposite of what you want to buy.
You only see what’s in that long tail when you look closely. Much of the content is made for teenagers, much of it is in other languages, and a growing share consists of AI-generated content without editorial oversight. For example, foreign-language content ran uncontrolled across more than 3,000 Turkish channels, without anyone having deliberately chosen those placements.
The highest-volume channels also told a story. At the top were mainly young creators and entertainment channels. Big names, but not the environments in which you want to show a brand to adult buyers.
From Open Reach to Inclusion
We now buy In-stream exclusively through inclusion: a list of channels we’ve selected and assessed in advance. The algorithm still optimises, but within an environment we control.
The list consists of channels with editorial context. Think TV broadcasters and their digital channels, regional broadcasters, news brands, satire and talk shows, and specialist creators with adult audiences. These are channels where a brand can be proud to appear.
The setup runs across three content tracks that we manage separately: Dutch, English and a deliberately selected language track for a specific audience. Previously, that last track ran uncontrolled through open targeting, spread across more than 3,000 channels. It now has its own ad group with 50 selected channels, allowing us to decide how much volume it receives.
Each track also runs separately on CTV — the YouTube app on television. This allows us to see how people watch each track on each screen and adjust budgets where it works. On regular In-stream, between 66% and 69% of viewers watch the entire ad, with almost identical results across the three tracks. This makes the tracks comparable and gives us control over how we manage them.
Quality Turned Out to Be Cheaper Too
The CPM for In-stream fell by 43%, while we bought 12% more reach per day. We compared the previous campaign flight using open targeting with the first three weeks of the new structure. Because the flights were not the same length, we compared performance on a daily basis.
In-stream results
| Metric | Change |
|---|---|
| Active channels | From 264,129 to 8,814 (−97%) |
| Impressions per day | +12% |
| CPM | −43% |
| CPM on TV broadcasters and news brands | −61% |
| Share of reach from broadcasters and news brands | From 3.5% to 5.6% |
| View-through curve across quartiles | Virtually unchanged |
That’s the key point. We set out to buy quality and accepted that it might cost more. It didn’t. By removing the long tail, we eliminated the most expensive and least controllable part of the media buy.
Viewing quality remained intact. The view-through curve is virtually identical to that of the previous campaign flight: the same viewing behaviour, at a much lower price. And the CPM hasn’t reached its full potential yet. After the first few days, it has fluctuated within a range of around 10%, with the most recent days being the cheapest of the entire flight.
Within the inclusion set, the best channels are slightly more expensive per impression than the average. That’s a deliberate trade-off. Shifting spend towards those channels costs us a little reach but delivers better context. That’s why we’re doing it gradually, with maximising reach remaining the second objective.
CTV: The Calmest Screen in the House
More than a quarter of our In-stream reach now comes through CTV, the YouTube app on television. Here, YouTube isn’t a feed on a phone; it’s a programme in the living room. Four out of ten viewers watch the entire ad.
That screen does come at a higher price: the CPM is approximately 36% higher than regular In-stream. This is exactly where CPM demonstrates why it’s a poor measure of value. A large screen, the sound on, often with several people on the sofa — that’s a different kind of contact from a video that starts playing while someone scrolls past it.
Clicks tell you nothing here. This is reach you evaluate based on viewing, not clicking.
Measuring What Really Matters
Google Ads measures performance using last-click attribution and systematically underestimates YouTube’s contribution during the awareness stage. Someone who sees an ad on TV and searches for the brand a week later doesn’t count towards YouTube’s contribution in that model. If you make decisions based on those figures, you cut precisely the reach that helps a brand grow.
That’s why we assess YouTube’s contribution through Billy Grace UMM, which measures incremental impact across all channels. A model like this needs time to learn, so we only evaluate a new structure once enough data has accumulated.
The placement data itself also needs to be interpreted carefully. Google doesn’t show all inventory at channel level. In the previous campaign flight, 72% of the media spend could be traced back to a channel. In the new structure, that figure rose to 94%. That’s an improvement in itself: you finally know where your ad is running.
Making Quality Measurable with Spektre
An inclusion list is only as good as the method used to assess its channels. A list based on intuition quickly becomes outdated and doesn’t scale. That’s where Spektre comes in — our partner for brand suitability and channel curation on YouTube.
Spektre uses AI to assess YouTube channels and assigns each channel an ASPI score. This makes it possible to maintain quality at a scale that would be impossible for a person. While a marketer might be able to review a few hundred channels manually, AI can assess content at the scale of YouTube itself.
More Than Brand Safety
ASPI goes beyond simply asking whether a channel is safe. Spektre assesses channels across several dimensions at once:
- Language Performance Breakdown: The language of the content and how each language performs.
- Context Per Impression: The subject matter and context in which impressions appear.
- GARM: The brand-safety standard established by the Global Alliance for Responsible Media.
- Attentional Mode Split: Whether viewers are actively watching or the video is typically playing in the background.
- Production Type Breakdown: Whether content is made by people or mass-produced automatically.
Spektre also identifies the intended audience for the content: children, teenagers, young adults or a broader audience. This is precisely where the problem of children watching through their mother’s account becomes clear. The targeting says a 45-year-old woman; the content says “Made for Kids”.
From Tier S to Tier D
All these dimensions come together in a consistent quality scale. It helps determine which channels justify a higher price and which are best avoided altogether.
| Tier | What it means | How we use it |
|---|---|---|
| S | The best human-made channels, with the highest attention levels and strongest emotional connection. | Premium value; we scale up here. |
| A | High-quality professional content that viewers watch actively and attentively. | The core of the inclusion set. |
| B | Reliable, mainstream channels with consistent views, often featuring a fixed or recurring format. | For volume at a competitive price. |
| C | Safe content that mainly plays in the background. | Only for mass reach. |
| D | Unsafe content, AI slop or content watched by children through adult accounts. | Always excluded. |
From Scores to Action
We use these scores to build clusters based on quality and context, rather than relying on one long list. These include TV broadcasters and news brands, the highest-tier channels, thematic clusters relevant to the brand’s category, and language tracks for audiences we deliberately want to reach.
Each cluster receives its own budget. This lets us see what each cluster delivers and scale up what works, rather than optimising around an average. That’s where the combination becomes powerful: Spektre assesses the quality of the environment, while Draft Digital manages the buying and measures the impact.
Seven Steps Every YouTube Advertiser Can Take
- Export your placements. Look at the channels where your ads appear and check how many receive fewer than 100 impressions.
- Look beyond the CPM. Ask who actually watched, where the ad appeared and whether a real person was in front of the screen. A 45-year-old woman in the targeting data could be a child using their mother’s account.
- Buy In-stream through inclusion. Select your channels in advance instead of leaving the long tail to the algorithm.
- Score quality, not just price. Have channels independently assessed for brand suitability, and compare those scores with reach and viewing behaviour.
- Optimise by clusters. Group channels by quality and context, give each cluster its own budget, and scale up what works.
- Use CTV deliberately. Run In-stream separately on television and evaluate that reach based on viewing, not clicks.
- Measure incrementally. Last-click attribution underestimates YouTube’s contribution. Use a model that measures impact across all channels, and give it time to learn.
Fewer Channels, Better Context
The lesson from this analysis is simple: a low CPM is only valuable when you know what you’re buying. By prioritising quality from the outset, we eliminated the long tail, improved the context and reduced the cost per thousand people reached by almost half. Quality and efficiency turned out not to be opposites.
YouTube is one of the largest video channels in the Netherlands, and it only delivers value when you know where your ad is running. Draft Digital and Spektre help brands take back control of that decision: measuring quality, selecting channels and directing media buying towards contexts that work.
Curious where your YouTube ads are appearing? Contact Draft Digital or Spektre for a placement scan.
Together, we’re setting a new standard.
“Dashboards often show everything in green, but the reality of campaigns is often very different. With our ASPI score, we cut through the noise and make it visible where budgets are actually going. Quality shouldn’t be the exception; it should become the new standard in media buying.”
— Co-Founder, Spektre

