Latest / The CMO Podcast with Fexingo: Marketing Leadership, Budgets, and Executive Strategy / The Brand Safety Paradox Inside Programmatic Advertising
Transcript
- Lucas: So back in 2017, Procter & Gamble pulled over a hundred million dollars in digital ad spend essentially overnight. Their CMO at the time, Marc Pritchard, said they'd discovered their ads were running alongside extremist content on YouTube. The brand-safety panic was real. And what's interesting is that nearly a decade later, the same problem keeps resurfacing — just with different names. Luna: I remember that. Every big brand had their 'YouTube moment.' But you're saying the fixes we put in place haven't actually fixed anything? Lucas: That's the paradox. The brand-safety industry — companies like Integral Ad Science, DoubleVerify, and others — is now worth roughly eight billion dollars. Yet in 2025, a major study found that over 20 percent of programmatic impressions still end up on sites that advertisers explicitly flagged as off-limits. We're spending more to verify than we are to actually reach quality environments. Luna: And that's the part I want to dig into. Because I've sat in those meetings where someone says, 'We need whitelists. We need certified safe publishers.' And it feels like a solid plan. But it clearly isn't working. Lucas: Look, if these marketing conversations have sparked something you've actually used — a smarter media plan, a rethinking of a budget, even just a phrase that won you an argument — a small handful of listeners keep this show going through buy me a coffee dot com slash fexingo. That's the whole economics of it. No ads, no sponsors, just people who find it useful chipping in a few bucks each month. Luna: And it's genuinely a small group. If everyone who listens forwarded that link once to a colleague who gets the same value, it'd be more than enough to keep us making these. Anyway — back to those whitelists. Lucas: Right. So the whitelist approach is seductive. You pick two hundred domains you trust — say, The New York Times, Bloomberg, ESPN — and you only buy inventory on those sites. But here's the problem: that inventory is expensive and limited. The open exchange is what's cheap and abundant. And programmatic systems are optimised to spend the budget. So if you set a broad targeting parameter but layer a whitelist on top, the algorithm will still chase low-cost placements that technically match the whitelist but appear on subpages you'd never approve. Luna: Meaning your ad ends up on a comment thread within a trusted domain that's filled with hate speech. Lucas: Exactly. That happened to major advertisers in 2020 on The Guardian's own site — their comment sections were toxic. And the brand-safety tools had flagged the domain as safe. So the system delivered impressions to a 'safe' URL that was anything but. This is the core failure: brand safety is measured at the domain level, not the page level. And even when it's measured at the page level, the technology relies on keyword matching and simple classifiers that miss context. Luna: I've seen that up close. A client of mine was buying programmatic inventory and their verification partner flagged a cooking blog as 'high risk' because it mentioned firearms — they were running a recipe for game meat. Meanwhile, actual hate speech on a mainstream news comment section sailed right through. Lucas: That's the classic false positive / false negative problem. And it's costing advertisers in both directions. False positives inflate CPMs because you're paying a premium for 'verified safe' inventory that's actually no safer than the open exchange. False negatives create reputational risk. Industry estimates suggest brand-safety verification adds between 30 and 40 percent to the effective CPM. Luna: So what's a CMO supposed to do? Just accept the risk? Lucas: Partially, yes. The most effective CMOs I talk to have moved away from the whitelist / blacklist binary. They're using a risk-tiered approach. Tier one is direct deals with premium publishers — you negotiate directly, you control the environment. Tier two is private marketplaces with pre-vetted curation. Tier three is the open exchange, but with a much looser set of exclusions and a higher tolerance for minor adjacency risk. The key is not to spend eight billion dollars pretending you can eliminate risk entirely. Luna: And that's a hard sell to a chief marketing officer whose CEO just saw a screenshot of the company's ad next to something horrible on Twitter. Lucas: It is. But the alternative is worse. Because the more you clamp down with automated tools, the more you narrow your reach and the more you pay. There's research from 2024 showing that brands with the most restrictive brand-safety settings saw a 60 percent reduction in addressable audiences and a 45 percent increase in cost per completed view. Meanwhile, their actual brand-safety incidents didn't improve measurably. Luna: So the tools are creating a penalty without delivering the benefit. Lucas: Exactly. And here's where it gets structural. The ad-verification industry has an incentive to keep the problem alive. If they actually solved brand safety, their revenue would collapse. So they sell a narrative of 'continuous improvement' that requires ever more sophisticated — and expensive — technology. Marc Pritchard himself said in a 2023 interview that the industry has created a 'complexity tax' that benefits the middlemen, not the advertisers. Luna: That feels like a damning quote. But if I'm a CMO, I still need to show my board I'm doing something. What's the practical play? Lucas: Three things. First, shift budget from verification to curation. Instead of paying a vendor to tell you where not to go, pay a premium to publishers whose environments you trust. That money goes to quality content, not to a middleman. Second, use contextual targeting instead of behavioral. Contextual tools look at the page content, not the user's browsing history. They're less prone to false positives. And third, accept a defined risk budget — say, 5 percent of impressions may appear in less than ideal environments, but you monitor those exceptions manually rather than trying to automate them away. Luna: I like the risk-budget idea. It's honest. It acknowledges that programmatic is a probabilistic system, not a deterministic one. Lucas: Right. And it avoids the worst outcome: that you spend millions on brand-safety tech, still get burned in a high-profile incident, and then have to explain to your CEO why you're paying for something that didn't work. The Unilever experience is instructive here. After the 2020 Facebook ad boycott, they pulled back on programmatic and invested more in direct publisher relationships. Their brand-safety incidents dropped, and their cost per engagement actually improved. Luna: So the answer is less technology, more relationships. Lucas: That's the contrarian take, but the data supports it. The brand-safety paradox is that the harder you try to eliminate risk through automation, the more you introduce inefficiency and cost without eliminating the risk. The smarter move is to accept a baseline level of adjacency risk and invest the savings into understanding your publisher ecosystem deeply. Luna: Which sounds like a more strategic role for the CMO anyway — not just buying ads, but building media partnerships. Lucas: Bingo. And that's where the best marketing leaders I know are headed. The question is whether the rest of the industry will follow, or whether we'll keep pouring money into verification tools that enrich vendors while frustrating advertisers. Luna: I think we know which way the smart money goes. Lucas: We do. And that's the conversation every CMO should be having with their procurement team right now.