Why CPMs are the single dumbest metric to KPI media buying (and buyers!) against
Let’s me be blunt. If your marketing team is still using Cost Per Mille (CPM) as its North Star metric, you are not just burning money; you are actively, systematically, and with quantifiable certainty, setting it on fire. You are paying for the privilege of being lied to. And in 2024, you and every other advertiser are projected to burn a collective $100 billion doing it [1].
For decades, the digital advertising industry has been held captive by this single, deceptively simple metric. It has become the default language of media buying, the bedrock of bad agency contracts, and the lazy KPI of choice for marketers who value the illusion of efficiency over the hard reality of effectiveness. And in doing so, it has single-handedly created a multi-billion dollar criminal enterprise, commoditized strategic thinking to nothing, and driven a crisis of confidence in the value of advertising itself.
The incentive structures it creates are criminally bad. It is a vicious, self-perpetuating cycle of perverse incentives that rewards fraud, penalizes quality, and ensures that advertisers, the very people footing the bill for the entire ecosystem, are the ones left with the worst possible outcomes.
In this article, we’re going to dissect this catastrophic failure. We’ll look at the metric’s inglorious history, the economic trap it sets, how it became the economic engine for global ad fraud, and how the pressure to lower it actively destroys business value. Most importantly, we’ll outline the only viable way out: a radical shift from buying impressions to buying outcomes.
The Original Sin: A Brief, Inglorious History of CPM
To understand how we got here, we have to go back to the primordial soup of the digital ad world, circa 1996 [40]. The internet was the Wild West, a chaotic mess of banner ads and pop-ups. In a desperate attempt to bring a semblance of order and accountability, the industry (led by the IAB) reached for what it knew: the models of traditional media. They needed a simple, standardized unit to buy and sell digital inventory, just like they bought column inches in a newspaper or 30-second spots on television.
Enter the CPM. Cost Per Mille, or Cost Per Thousand impressions.
On the surface, it seemed logical. It was a simple, auditable number. You pay X dollars, and in return, a publisher’s ad server serves your ad one thousand times. The problem, and the original sin from which all other dysfunctions flow, is in that one, seemingly innocuous word: served.
The CPM model measures the cost to request that an ad be displayed [5]. It says nothing about whether that ad was actually visible on the screen (aka viewability), whether it was seen by a carbon-based lifeform (aka ad fraud), whether that lifeform was in your target market, or whether it had any impact whatsoever (actual money in your bank). It measures a technical event in a server log, not a moment of human attention.
This was the fatal flaw. A metric designed to create accountability has, ironically, built an entire ecosystem that is fundamentally devoid of it. It allows every intermediary in the sprawling, murky programmatic supply chain to claim success—"We served the impressions!"—while the advertiser’s business goes nowhere. It created a system where everyone gets paid for showing up, and advertiser takes 100% of the risk that any of it actually worked [27, 32].
The CPM Trap: Optimizing for Efficiency, Achieving Ineffectiveness
The core of the problem today is that CPM has been twisted from a simple pricing model into a Key Performance Indicator (KPI). And when you make CPM your primary KPI, you fall into a devastating strategic trap. You begin to conflate efficiency with effectiveness [28].
Let me be clear: CPM is a metric of efficiency. It tells you, in theory, how cost-effectively you are reaching an audience. But efficiency is worthless if it’s not effective. You can efficiently buy a million impressions for $10, but if those impressions are served to bots in a server farm in Siberia or stacked invisibly behind 15 other ads on a pirated movie site, your "efficient" campaign has been 100% ineffective. You’ve efficiently set your money on fire.
This leads to a critical, and often ignored, bifurcation in the market. The term "CPM" is used to describe two fundamentally different products:
The Quality CPM: This is the cost of premium, viewable, human-seen inventory on legitimate platforms, targeted to a relevant audience. Its price is rising for valid economic reasons: a shift to high-impact video formats, better targeting capabilities, and more brands competing for finite premium space [18, 19].
The Junk CPM: This is the cost of fraudulent, non-viewable, bot-driven, or otherwise worthless impressions. This inventory is sold at a suspiciously low price specifically to attract buyers whose primary mandate is to lower their average CPM [21].
The industry-wide "average CPM" is a meaningless, murky blend of these two opposing markets. And here is the trap: A strategic focus on lowering your average CPM inevitably pushes your media buyers away from the quality market and directly into the junk market. If you ask media buyers to be effective whilst demanding the lowest CPMs, you have given them an impossible task.
When a CMO tells their agency, "Your primary goal is to lower our CPM by 15% this year," they have just given that agency a clear mandate: "Go buy more junk." The agency, to hit its target and keep the business, is now incentivized to seek out the cheapest possible impressions, which are cheap precisely because they are garbage. As one analysis bluntly puts it, a suspiciously low CPM means your ad is "showing up in some weird places" [21]. The KPI has been met, but the business objective has failed.
The Fraud Economy: How CPM Built a Criminal Empire
This isn't just about poor quality; it's about criminality on an industrial scale. The CPM model, by creating a payable event that is completely disconnected from human verification, provides the perfect business model for fraud. It is the economic engine of a global criminal enterprise.
The numbers are not just bad; they are apocalyptic.
A landmark Juniper Research study found that advertisers lost $84 billion to ad fraud in 2023. Other reports peg it at $87 billion [1, 9].
Projections show this isn't slowing down. The forecast is for losses to hit $100 billion in 2024, and a staggering $172 billion by 2028 [9, 11].
The World Federation of Advertisers has warned that ad fraud is on track to become the second-largest market for organized crime, trailing only the drug trade [9, 11].
Analyses suggest that advertisers lose $1 for every $3 spent on digital ads. Let that sink in. A third of your budget, gone [11, 15].
This isn't a rounding error. This is a systemic plundering of marketing budgets, enabled directly by the CPM. Fraudsters don’t need to trick a human; they just need to trick a server. They do this through a sophisticated arsenal of techniques:
Industrial-Scale Bot Traffic: In 2024, nearly half of all internet traffic (47-50%) is non-human, and "bad bots" make up almost a third of the total (30-32%) [6]. One analysis found bots account for 80% of web visits, meaning only one in five visitors is human [7]. These are automated programs designed for one purpose: to visit websites and load ads, generating fraudulent impressions that advertisers pay for. Some operations involve botnets of 1.7 million infected PCs generating 12 billion bid requests a day [9].
Ad Stacking: Multiple ads are layered in a single ad slot. Only the top one is visible (if that), but impressions are counted and charged for every single ad in the stack. This tactic alone was projected to cost advertisers $35 billion [7, 8].
Impression Laundering: Fraudsters set up non-brand-safe websites and use complex redirection schemes to make their inventory look like it's coming from a legitimate, premium publisher, tricking advertisers into buying worthless impressions at inflated prices.
Look at the parallel trends. On the one hand, CPMs on major platforms are soaring. Between Q4 2020 and Q4 2023, the average CPM on Meta rose from $13.28 to $16.31. On Google Display, it went from $15.28 to $21.28. On TikTok, it exploded from $2.93 to $10.43 [13]. At the same time, the money lost to fraud is accelerating at an almost identical rate.
This is not a coincidence. It is cause and effect. The system that rewards the serving of impressions is the system that funds the generation of fraudulent impressions.
The Vicious Cycle: How Procurement Kills Value
So how does this manifest in a real company? It begins with a well-intentioned but catastrophically misguided directive from leadership, which sets off a destructive chain reaction. Here’s a scenario to walk you through it:
Step 1: The C-Suite Mandate. The cycle starts in the boardroom. A CMO or CFO, looking at a line item for "digital media spend," declares that the goal is to drive "efficiencies." This is corporate-speak for "cut costs," and the chosen metric is, inevitably, the CPM.
Step 2: Procurement Gets Its Marching Orders. The mandate flows to the procurement department. The role of procurement in agency negotiations has exploded. Over a decade, its contribution rose by 90%, and today procurement leads 51% of all contract talks, while marketing's leadership has fallen to just 20% [24]. Their mission is simple and clear: negotiate lower CPMs and reduce agency fees. They are not tasked with evaluating strategic value or business impact; they are tasked with commoditizing a purchase. (Note: some procurement departments are amazing, and do not do this).
Step 3: The Agency is Handcuffed. The agency now sits across the table from a procurement team whose entire bonus structure may depend on hitting a cost-reduction target. The conversation is no longer about strategy, audience insights, or creative effectiveness. It's a "race to the bottom" on price [22]. The agency's role is debased from a strategic partner to a simple media broker.
Step 4: The Inevitable Purchase of Garbage. To meet the client's mandated low-CPM target, the agency has no choice. They must venture deep into the long tail of the programmatic ecosystem, the dark corners where the "Junk CPM" inventory lives. They buy the cheap, low-quality, high-fraud-risk media because it's the only way to blend down the average CPM to the number the client has demanded. As one analysis bluntly puts it, a suspiciously low CPM means your ad is "showing up in some weird places." You get what you pay for [21].
Step 5: The Terrible Results. The campaign runs. The agency hits its CPM target, and procurement celebrates a "win." But the client's business doesn't grow. Sales are flat. Leads are junk. The campaign was a commercial failure.
Step 6: The Cycle Repeats. The CMO looks at the poor results and concludes that digital advertising "doesn't work as well as it used to." Their perception of marketing as a cost center to be minimized is reinforced. And so, the next quarter, the mandate is even more aggressive: "We need to lower our CPMs even more." The vicious cycle begins anew, but this time with more intensity, driving the agency even deeper into the fraudulent ecosystem.
The Way Out: From Impressions to Impact
Breaking this cycle requires a fundamental, non-negotiable shift in thinking. It requires advertisers to stop asking, "How cheaply can I buy impressions?" and start asking, "How effectively can I buy business outcomes?"
This means abandoning CPM as a primary KPI and embracing metrics that are directly tied to business value: Return on Investment (ROI), Cost Per Acquisition (CPA), and, most importantly, incremental sales.
The gold standard for making this shift is Marketing Mix Modeling (MMM). MMM is a top-down statistical analysis that quantifies the impact of all marketing inputs on a key business outcome, like revenue. It moves beyond siloed, channel-specific metrics to give you a holistic, cross-channel view of what is actually driving your business forward. It accounts for external factors like seasonality and economic trends, providing a far more accurate picture of your true ROI [2, 3].
This isn't some academic fantasy. This is how the world's most sophisticated marketers are already operating, and the results speak for themselves:
A July 2024 survey found that 72% of marketers trust MMM more than any other method for identifying the drivers of business outcomes [2].
Coca-Cola used MMM to optimize its global marketing spend, discovering that digital channels delivered a higher ROI than TV in many markets. The result? A 20% improvement in media efficiency and significant incremental growth [2].
An FMCG brand used MMM to analyze three years of data, leading to a 15% increase in marketing ROI in the first year [45].
A major retailer used MMM to discover its digital ads had a higher ROI than its TV campaigns. Shifting the budget accordingly led to a 15% boost in sales [3].
A Deloitte analysis found that in over half of MMM projects, the contribution to the business increased by over 6% with the same advertising budget [46].
Yes, implementing MMM can be hard. It requires at least two to three years of clean, granular historical data [47]. It requires specialized data science expertise.
But these are not reasons for inaction. They are the necessary investments required to be a serious, data-driven business. Complaining that it's too hard is an excuse to continue lighting money on fire.
A Call to Arms for the Modern Marketer
The choice facing every advertiser today is stark. We can continue to worship at the altar of the CPM, a false idol that has led the industry into a crisis of fraud and ineffectiveness. We can continue the vicious cycle of pressuring your agencies to buy cheaper and cheaper junk, wondering why your results get worse and worse.
Or we can make a change.
You can redefine your North Star KPI to a real business outcome. You can re-mandate your teams to pursue value, not just cost, and to structure performance-based contracts where your agency partners share in the risk and reward. You can invest in a robust, independent measurement framework like MMM to get an honest, holistic view of what’s working and use that to enhance the perception of value across your organisation.
This is the only way forward. Stop buying impressions. Start buying outcomes. Your budget, your brand, and your business depend on it.
References
[1] Statista. (2024). Global digital ad fraud cost 2023-2028.
[2] eMarketer / Snap Inc. (2024). Marketing Mix Modeling: How Marketers Are Measuring the Drivers of Business Outcomes.
[3] Think with Google. (Various years). Marketing Mix Modeling case studies.
[4] Juniper Research. (2023). Digital Advertising Fraud: Key Trends, Competitor Leaderboard & Market Forecasts 2023-2028.
[5] Investopedia. (2023). Cost Per Mille (CPM).
[6] Imperva. (2024). 2024 Imperva Bad Bot Report.
[7] The Drum. (2023). 80% of web visits are from bots.
[8] Forbes. (2023). The $35 Billion Problem Of Ad Stacking And Click Spam.
[9] University of Baltimore. (2017). The Methbot Operation.
[10] Interactive Advertising Bureau (IAB). (Various reports). Programmatic Advertising and Ad Fraud.
[11] World Federation of Advertisers. (Various reports). Global Ad Fraud.
[12] Marketing Dive. (2023). Global media inflation 2023.
[13] Social Insider. (2024). Social Media CPM, CPC, & CTR Benchmarks.
[14] AdExchanger. (2024). Programmatic DOOH CPMs.
[15] Forbes. (2023). Advertisers Lose $1 for Every $3 Spent on Digital Ads.
[16] Statista. (2024). Projected ad fraud losses worldwide 2024.
[17] Marketing Evolution. (Blog). What is CPM?
[18] HubSpot. (Blog). What Is a Good CPM?
[19] Influencer Marketing Hub. (Blog). What is CPM? The Easy Guide to Cost-Per-Mille.
[20] Revealbot. (Blog). What is CPM in Marketing?
[21] Instapage. (Blog). What Is CPM? How Is It Calculated?
[22] Forbes. (2017). Is Marketing Procurement A Broken Model?
[23] World Federation of Advertisers (WFA). (2020). Global Guidance on Marketing Procurement.
[24] Marketing Week. (2015). The changing face of procurement.
[25] LinkedIn Marketing Solutions. (Blog). Marketing Procurement: A Strategic Guide.
[26] RightSpend. (Report). The Value of Data-Driven Procurement.
[27] Forbes. (2021). How To Structure Performance-Based Agency Agreements.
[28] Marketing Evolution. (Blog). CPM vs. CPC vs. CPA.
[29] HubSpot. (Blog). CPM: What It Is & How to Calculate It.
[30] Mailchimp. (Resource). CPM (Cost-Per-Mille).
[31] Investopedia. (2023). CPA: Cost Per Acquisition.
[32] Forbes. (2022). The Dangers Of Click Fraud.
[33] eMarketer. (2023). Enterprise Marketers' Priorities.
[34] Klipfolio. (Resource). What is Cost Per Mille (CPM)?
[35] Corporate Finance Institute. (Resource). Cost Per Mille (CPM).
[36] AdEspresso. (Blog). Why Is My CPM So High?
[37] StoreYa. (Blog). What is a Good CPM?
[38] Marketing Charts. (2020). Advertisers Can Save Up to 30% of Their Budgets.
[39] Nielsen. (Report). ROI Report.
[40] The New York Times. (1996). A New Way to Sell Ads on the Internet.
[41] LinkedIn Marketing Solutions. (Blog). CPC (Cost-Per-Click).
[42] McKinsey & Company. (2023). Marketing’s new key performance indicators.
[43] Forbes. (2023). The Shift To Outcome-Based Marketing.
[44] Recast. (Blog). What is Marketing Mix Modeling (MMM)?
[45] Analytic Edge. (Case Study). FMCG Brand ROI Improvement.
[46] Deloitte. (Report). Marketing Mix Modeling ROI.
[47] Neustar. (Blog). The Pros and Cons of Marketing Mix Modeling.
[48] Ekimetrics. (Blog). MMM Data Requirements.
[49] Measured. (Blog). The Limitations of Marketing Mix Modeling.
[50] Statology. (Resource). Correlation is not equal to causality.
[51] Google AI. (Blog). Robust and Reliable Causal Inference.
[52] INSEAD. (2017). The 10 Commandments of KPI.
[53] ClearPoint Strategy. (Blog). 17 KPI Mistakes You’re Probably Making.
[54] Geckoboard. (Blog). The Dangers of Inaccurate Data.
[55] Bernard Marr & Co. (Blog). The KPI Trap.