F8·Digital Marketing·Programmatic Reckoning Case

Procter & Gamble — The Digital Ad Fraud Reckoning

Covers lectures

F8-05 · F8-06 · F8-09

Procter & Gamble — The Digital Ad Fraud Reckoning

Module: F8 — Digital Marketing Type: Programmatic Reckoning Case Cross-references: F8-05 (programmatic supply chain), F8-06 (attribution illusions), F8-09 (integration beats separation)

The Situation

On 29 January 2017, in a conference room at the Interactive Advertising Bureau's Annual Leadership Meeting in Hollywood, Florida, Marc Pritchard walked to the lectern and began dismantling the digital advertising industry from the inside. Pritchard was Chief Brand Officer at Procter & Gamble, the single largest advertiser on Earth, with an annual media budget of roughly $8 billion and brands including Tide, Pampers, Gillette, Olay, Crest and Ariel spanning almost every household on the planet. When P&G speaks, agencies listen. When P&G threatens to stop paying, the entire ecosystem flinches.

Pritchard's speech, later reprinted almost verbatim in trade press, was unusual in both tone and substance. He opened by acknowledging P&G's own complicity. "We have a media supply chain that is murky at best and fraudulent at worst," he said. He then set out a four-point ultimatum that the industry had one year to meet: adoption of the Media Rating Council's viewability standard, implementation of third-party verified measurement, independent fraud audits, and transparent contracts with agencies that eliminated the opaque rebate and arbitrage practices that had become endemic in programmatic buying. The deadline was the end of 2017. After that, P&G would start cutting spend from any vendor or platform that failed to comply.

This was not grandstanding. Over the following eighteen months, P&G did exactly what Pritchard threatened. By mid-2018, the company had cut roughly $200 million of digital media spending that internal audits had classified as ineffective, non-viewable, fraud-adjacent or duplicated against other channels. The cuts included approximately $140 million from YouTube at peak and roughly $120 million from Facebook across 2017-2018. The industry held its breath. Analysts at Morgan Stanley and Bernstein modelled the expected damage to P&G's brand health, volume share and organic sales growth. Pritchard had just deliberately lit $200 million on fire, in public, on the assumption that it was already on fire and nobody was telling him.

What happened next was the detail that matters most for this case and for the broader F8 curriculum. P&G's earnings for the twelve months following the cuts showed no negative impact on brand health, volume share, penetration, or organic sales. In fact, organic sales growth accelerated modestly. Pritchard confirmed this publicly at the IAB's 2018 meeting and again at the Association of National Advertisers' Masters of Marketing conference the same year. Cutting $200 million of digital media spend from the world's largest advertiser had precisely the impact the internal models predicted: zero. The spend had never been doing the work the invoices claimed it was doing.

This is the reckoning. It is the single most cited empirical event in the argument that the programmatic digital advertising supply chain, as it existed between approximately 2012 and 2022, was structurally wasteful on an industrial scale.

The Data

The scale of the P&G cut

P&G's total annual advertising expenditure in fiscal 2016, the year before Pritchard's speech, was approximately $7.2 billion, of which roughly $2.4 billion, or around 35 per cent, was allocated to digital channels (company 10-K filings; Bernstein Research estimates). By fiscal 2018 the digital allocation had been rebalanced downward by approximately $200 million — a 7-8 per cent cut from digital specifically, equivalent to around 3 per cent of total media. Within that $200 million envelope, Pritchard identified roughly $140 million of YouTube inventory and $120 million of Facebook inventory as having failed his transparency tests at some point during the period; the two figures overlap because some spend was shifted before being cut and some was cut and partially reinstated once verification was in place.

The decision was informed by internal fraud audits, work by the verification vendor White Ops (now Human Security), and independent viewability analysis by Moat and Integral Ad Science. P&G reported that viewability on some YouTube ad inventory had measured as low as 50 per cent, meaning half the impressions it was paying for never actually appeared on a human screen for the MRC-defined minimum duration. On some programmatic display buys, viewability was below 30 per cent. Invalid traffic — bots, data centre traffic, click farms — was measured in some programmatic video buys at between 7 and 15 per cent of impressions.

The near-zero business impact

Across fiscal 2017 and fiscal 2018, P&G's organic sales grew by 2 per cent and 3 per cent respectively, both within or slightly above the company's guided range. Volume share was stable or modestly up in roughly three quarters of its core categories. Brand equity tracking (conducted internally via Kantar BrandZ and through P&G's own continuous tracking programmes) showed no degradation for any of the company's top ten brands through the period of the cuts. Pritchard, speaking at the ANA Masters of Marketing in October 2018, summarised the finding as follows: "We reduced wasted spend by $750 million in agency and production fees and over $200 million in ineffective digital media, and we grew the business." The $750 million figure included production and agency fee savings separate from the media cut; the $200 million is the media cut specifically.

The revealing detail is not that the cuts were neutral for the business. It is that the business grew. If $200 million of spend had been doing measurable work, removing it should have dragged growth down by at least a sliver. Instead, growth accelerated. The simplest explanation — the one Pritchard himself adopted — is that the spend had been substantially phantom: paid for, counted in reports, and never actually producing incremental reach, attention or sales.

The structural waste in programmatic

What the P&G reckoning revealed was not a rogue set of bad actors inside the digital supply chain; it was the supply chain itself. Independent analyses from 2017 onwards converged on a remarkably consistent figure: for every dollar an advertiser spent in programmatic open-exchange display and video, only between forty and sixty cents was actually reaching the publisher or platform where an ad could theoretically be seen. The remainder was absorbed by demand-side platforms, supply-side platforms, data management platforms, verification vendors, ad servers, exchange fees, agency trading desks, and the various arbitrage layers in between.

The definitive quantification arrived in June 2023, when the Association of National Advertisers, working with PwC, Kroll and the programmatic analytics firm Chalice Custom Algorithms, published its Programmatic Media Supply Chain Transparency Study. The study analysed $123 million of programmatic spend across 21 advertisers and 95 million log-level impressions. Its central finding: the programmatic open marketplace, across U.S. advertisers alone, wastes an estimated $20 billion annually — roughly one in three dollars spent. The study documented 44,000 websites receiving programmatic ad spend across just 21 advertisers; the median "made-for-advertising" site (low-quality content farms optimised to game viewability metrics) captured 15 per cent of open-web ad spend by volume. Ads ran on sites no brand manager had ever approved, next to content no brand manager had ever seen, measured by metrics no brand manager had ever validated.

The post-reckoning rebound and the new spend

P&G did not abandon digital. After the cuts, Pritchard and his team reallocated aggressively — into linear and connected TV, into retail media, into first-party data activation, into reach-based upper-funnel YouTube placements that had cleared verification, and into a tighter supply path that bypassed most of the programmatic middle layer. By 2022, P&G's digital share of total media was once again around 40 per cent, but the composition had fundamentally changed. Less open exchange, more direct buys. Less last-click attribution, more marketing mix modelling. Less audience "precision", more reach-based brand building. The lesson P&G took from the reckoning was not that digital does not work. It was that most of what the digital supply chain had sold P&G as "digital marketing" had not been digital marketing in any meaningful sense; it had been a fee-extraction layer sitting between P&G and actual audiences.

The Analysis

What the programmatic supply chain actually was

The F8 lecture on programmatic frames the central issue as structural. A traditional television buy involves a small number of intermediaries: the advertiser, the agency, the network, the measurement vendor. Fee take-rates between advertiser and publisher typically sit in the range of 10-15 per cent in total. A programmatic open-exchange buy, by contrast, routes a single impression through, on average, 6-10 technology vendors, each of which takes a fee. Industry estimates during the 2015-2019 period commonly placed the cumulative "ad tech tax" at between 40 and 55 per cent of the advertiser's gross spend, meaning that for every $1 of budget, the publisher received only 45-60 cents — and the remainder was absorbed not by the content the audience was there to see, but by the invisible plumbing that routed an impression to an auction.

On top of that take-rate problem, the open marketplace had three structural failure modes that compounded one another. First, non-viewability: large portions of bought impressions appeared below the fold, in stacked ads, in minimised tabs, or for durations shorter than the MRC's one-second-for-display, two-seconds-for-video threshold. Second, invalid traffic: sophisticated bot networks (Methbot, 3ve) had been quietly extracting hundreds of millions of dollars a year from the open exchange by simulating human ad consumption. Third, made-for-advertising sites: publishers that built no audience at all, whose entire content model was to create ad inventory optimised to pass automated viewability checks. None of these failure modes were hidden. All of them were documented in trade press and verification vendor reports throughout 2014-2016. What Pritchard's 2017 speech did was make the failure modes expensive to ignore.

P&G was the perfect advertiser to expose this, because P&G's media spend is so vast, its brands so category-leading, and its measurement infrastructure so mature that the company could actually test the counterfactual. Most advertisers cannot pull $200 million of digital spend and credibly measure "did this change anything?" P&G could, and did, and the answer was no.

Attribution as the accomplice

The F8 lecture on attribution argues that the deeper problem is not programmatic per se but the attribution models that justified it. Last-click attribution, which assigns all credit for a sale to the final advertising touch before purchase, made programmatic look effective because a user who had already decided to buy Tide, searched "Tide pods" and clicked a retargeting display ad on the way to the purchase page was counted as a "conversion" that the display ad "drove". The display ad did no such thing. The sale would have happened without it. But last-click attribution was how platforms and agencies reported performance, because it produced the flattering numbers advertisers wanted to see.

Marketing mix modelling, which P&G has used for decades and strengthened substantially post-reckoning, produces almost opposite answers. When MMM is applied carefully to a large brand like Tide, the short-term ROI of upper-funnel brand-building channels (linear TV, out-of-home, sponsorships) typically comes out two to four times higher than last-click attribution credits them, and the short-term ROI of lower-funnel click-based channels (search retargeting, programmatic display) typically comes out two to four times lower. The gap between the two methods is the measurement of the attribution illusion.

P&G's cuts, in other words, were not just about fraud in the narrow sense of bots and made-for-advertising sites. They were about the entire reporting infrastructure that had convinced brand managers that spend which was not doing any real work was doing the most important work. Remove the flattering attribution model and the spend's true value becomes measurable. P&G removed the spend and the true value became measurable by its absence: zero.

The Both/And Lesson

Digital advertising works. It can reach audiences linear channels cannot. It can produce measurable, incremental brand and sales effects when bought with discipline. The P&G reckoning is not an argument against digital. It is an argument against the specific assumption — pervasive in 2012-2017 and still common today — that digital is a separate discipline operating under separate rules, best optimised through platform-native metrics and last-click attribution, and best delivered through opaque programmatic supply chains run by parties whose economic interests are not aligned with the advertiser's.

The Both/And lesson is this: digital is a channel set, not a separate discipline. It must be planned, measured and audited using the same evidence-based fundamentals that apply to every other channel. That means reach-based measurement, not impression-based. That means mixed-model attribution, not last-click. That means direct or audited supply paths, not blind open-exchange buys. That means brand-building work and activation work balanced in the Binet & Field 60/40 range, not the 10/90 skew that programmatic optimisation tends to produce. When digital is treated as a channel set inside a fundamentals-led marketing operation, it delivers. When it is treated as a separate discipline that escapes the fundamentals, it delivers $200 million of spend that Procter & Gamble could throw away without noticing.

P&G did not retreat from digital. It integrated digital into the same accountability regime it had always applied to TV, print and in-store. The numbers got smaller. The work got better. The business grew.

Questions for Reflection

P&G could cut $200 million of digital spend without a measurable business impact, implying the spend had been phantom. How would you design a test that a smaller organisation could run to discover the equivalent in its own budget?
The ANA 2023 study found that one in three dollars spent in programmatic open exchange is wasted. What specific contractual or measurement changes would you implement to move your own media plan out of the wasted third?
Last-click attribution flattered programmatic and punished brand-building for roughly a decade. Whose economic interests did that serve? Whose did it damage?
Pritchard's 2017 speech was unusual because it came from the largest advertiser in the world. Why do you think no smaller advertiser had forced the same reckoning earlier? What does that tell you about the incentive structure of the agency-advertiser relationship?
If you inherited a media plan tomorrow that was 80 per cent digital and 20 per cent traditional, with performance reported entirely via last-click, what three changes would you make in the first month, and what single measurement would you use to prove you were right?