Marketing Attribution: Understanding What Actually Drives Growth

Industry & Competitive Context

The modern advertising ecosystem is built on an unresolved paradox: brands spend hundreds of billions of dollars annually on marketing, yet the systems used to measure what that spending actually produces are frequently unreliable, opaque, or exploitable. The problem predates digital media the observation that half of all advertising spend is wasted, though no one can identify which half, has long circulated in marketing circles but programmatic advertising, mobile app ecosystems, and multi-touch digital journeys made the stakes dramatically higher. A single consumer purchase path today can touch search ads, social media impressions, retargeting pixels, affiliate links, and organic discovery, all within minutes, and each touchpoint is often tracked by a different vendor with a commercial incentive to claim credit for the outcome.

This created an industry-wide trust gap between advertisers and the ad-tech intermediaries managing their campaigns. Programmatic buying introduced layers of exchanges, networks, and agencies between the advertiser and the consumer, each taking a share of the budget and each reporting performance using self-supplied data. Viewability (whether an ad was actually seen), brand safety (where an ad appeared), and attribution (which touchpoint caused a conversion) became three distinct, frequently contested categories of measurement. Regulatory change compounded the uncertainty: the EU’s General Data Protection Regulation took effect in 2018, and Apple’s App Tracking Transparency (ATT) framework, introduced with iOS 14.5 in April 2021, required explicit user opt-in before apps could track activity across other apps and websites using the device identifier for advertisers (IDFA). Both developments restricted the very data infrastructure that multi-touch attribution models depended on.

Two independent, publicly documented episodes at Uber in 2017 and at Procter & Gamble in 2017–18 became widely cited reference points in the industry because they offered rare, concrete evidence of what happens when a major advertiser interrogates its own attribution data rather than accepting it at face value. A third episode, Meta’s 2022 disclosure of the financial impact of Apple’s tracking changes, illustrated the opposite side of the same dynamic: how dependent the platform economy had become on the attribution infrastructure that regulators and device makers were now dismantling.

Brand Situation Prior to the Reckoning

By early 2017, Uber was one of the most aggressive digital advertisers in the technology sector, spending roughly $150 million annually on paid channels to drive rider app installs, according to Kevin Frisch, Uber’s then-head of performance marketing and customer relationship management. The company operated in a growth-at-all-costs environment in which marketing teams, agencies, and ad networks were incentivized primarily to report installs, not to verify whether those installs were incremental that is, whether the advertising had actually caused them. Uber was simultaneously managing a separate, unrelated reputational crisis (the #DeleteUber movement), which heightened internal scrutiny of spending and performance across the business.

Procter & Gamble, the world’s largest advertiser by spend, faced a different but related problem. As a long-standing proponent of mass-reach brand marketing, P&G had steadily increased its digital ad investment through the 2010s but had limited visibility into whether its digital placements were actually being seen by real, relevant consumers. Viewability data obtained in 2017 revealed that the average view time for an ad on a mobile newsfeed was just 1.7 seconds barely longer than a glance while poor targeting meant some individuals were shown a disproportionate volume of P&G ads. These findings surfaced during a broader period in which advertisers across the industry were contending with ad fraud and brand-safety failures, including ads appearing alongside inappropriate or extremist content on major platforms.

In both cases, the companies were not suffering from a lack of marketing activity or a poorly performing campaign in the conventional sense. They were suffering from a measurement system that had gone largely unquestioned, despite controlling hundreds of millions of dollars in annual decision-making.

Strategic Objective

Neither company set out to overhaul its entire marketing function when it began investigating attribution. Uber’s objective, as described by Frisch, was narrower: to understand whether a modest reduction in ad spend would produce a corresponding drop in app installs, as a basic test of marginal return on paid media. Procter & Gamble’s stated objective, articulated publicly by Chief Brand Officer Marc Pritchard at the Association of National Advertisers’ media conference in early 2018, was to use newly available transparency data to ensure media dollars were reaching real, engaged audiences, and to pressure digital platforms into providing the third-party verification standards that offline media (television, print) had long been held to.

In both instances, the explicit goal was efficiency and accountability rather than growth suppression. P&G was clear that its objective was not to reduce total marketing investment but to reallocate it toward channels and formats that could demonstrate genuine reach and engagement.

Campaign Architecture & Execution

Uber’s investigation functioned as a controlled, if improvised, natural experiment. Frisch first turned off 10 percent of Uber’s annual rider-acquisition ad spend about $15 million expecting a corresponding decline in app installs. According to his account, there was no significant drop in new rider acquisition following that cut. Based on this result, Uber went further and suspended a much larger share of its digital ad spend, ultimately turning off $100 million of the $150 million annual budget roughly two-thirds. Uber again saw no material change in the number of rider app installs.

The explanation that emerged was attribution fraud: ad networks were claiming credit, and payment, for installs that would have occurred organically regardless of any ad exposure. Fraudulent actors were reportedly generating clicks across large numbers of devices so that when a user later installed the Uber app on their own, the network associated with that device could claim the install as an ad-driven conversion. Uber subsequently filed suit against U.K. mobile ad agency Fetch Media over alleged fraud, including failure to return rebates and misrepresentation of ad effectiveness; that suit was later voluntarily dismissed. A separate 2019 suit named roughly 100 additional companies believed to be sub-publishers operating within Uber’s ad networks. In 2020, Uber reportedly reached a settlement with mobile marketing firm Phunware, under which Uber received $6 million.

P&G’s process was less adversarial and more structural. Rather than running a blunt spend experiment, the company used viewability and transparency data supplied by media partners to identify where ad dollars were being wasted on inventory that was never meaningfully seen. Pritchard stated that, armed with this data, P&G reduced spending with several major digital platforms by 20 to 50 percent and cut total digital ad spend by $200 million over the course of 2017, split roughly evenly between a mid-year reduction and a second wave in the second half of the year. P&G also suspended advertising on YouTube after its ads appeared next to objectionable content, and it pushed Facebook and Google (then Alphabet’s YouTube) to improve content-quality algorithms and reporting standards. Independent measurement firm Media Rating Council was engaged to audit ad metrics for both platforms.

Positioning & Consumer Insight

The consumer-facing insight in both cases was secondary to an internal, operational insight: that organic demand was stronger, and paid media’s marginal contribution weaker, than either company’s existing attribution models had suggested. For Uber, the insight was that a meaningful share of app installs attributed to advertising were, in fact, driven by word of mouth, brand awareness, or simple habitual demand for ride-hailing installs that would have happened with or without the specific ad exposure being credited.

For P&G, the insight was about attention rather than installs: that impressions served did not equate to impressions seen, and that optimizing for raw ad delivery volume, rather than verified human attention, had allowed media budgets to be allocated inefficiently for years without detection. Pritchard’s public remarks framed this as a transparency problem as much as a performance problem the company argued that without independent, standardized measurement, advertisers had no reliable way to compare the true value of one channel against another.

Media & Channel Strategy

P&G’s channel strategy shift is the more fully documented of the two. Reinvestment following the digital cuts went toward television, audio (including streaming audio platforms such as Pandora), and e-commerce media (including Amazon and Alibaba), with the company stating that these channels offered more reliable and verifiable reach. Despite cutting digital spend by $200 million, P&G stated it did not reduce total media spending; funds were reallocated rather than eliminated, and the company reported that the reallocation increased overall media reach by roughly 10 percent. In its fiscal year ending in mid-2017, P&G’s total advertising spend was reported at approximately $7.1 billion, down only marginally from $7.2 billion the prior year indicating the digital cuts were offset by investment elsewhere rather than representing an overall retrenchment. The digital ad reductions also coincided with a separate initiative, announced in January 2018, to cut the number of agency partners by 50 percent as part of a broader $400 million cost-reduction target within agency and production costs, itself nested within a larger $10 billion company-wide cost-reduction program.

Uber’s channel response was narrower and more defensive: rather than reallocating budget to new channels, the company’s primary actions were to permanently suspend the specific paid-install campaigns found to be compromised and to pursue legal remedies against the ad agencies and networks it identified as responsible for the fraudulent attribution.

A separate, later episode illustrates the structural dependency of digital platforms on attribution data. When Apple introduced App Tracking Transparency in iOS 14.5 in April 2021, requiring apps to obtain explicit user permission before tracking activity across other companies’ apps and websites, a large share of iPhone users opted out of tracking. During its fourth-quarter 2021 earnings call in February 2022, Meta’s then-CFO Dave Wehner stated that the cumulative impact of Apple’s iOS changes would represent a headwind “on the order of $10 billion” to Meta’s 2022 revenue, directly attributing the impact to reduced ability to measure and optimize ad performance at the device level. Meta’s shares fell approximately 26 percent the following day, erasing roughly $200 billion in market value in a single session among the largest one-day losses in stock market history at that time. This episode did not involve fraud, but it demonstrated, at platform scale, how directly advertising revenue is tied to the availability and granularity of attribution data, and how quickly that revenue can be exposed when the underlying measurement infrastructure changes.

Business & Brand Outcomes

The documented outcomes in each case are narrow but significant. At Uber, the central, repeatedly reported outcome was that cutting $100 million of annual ad spend — two-thirds of the paid-acquisition budget — produced no material drop in rider app installations, strongly indicating that the ad spend being eliminated had not been driving incremental growth. At P&G, the company reported no reported negative impact on sales following its $200 million digital ad reduction, and media reach reportedly increased by approximately 10 percent due to reallocation toward other formats. P&G also reported that its transparency campaign contributed to behavioral change among its largest digital partners: Pritchard stated that both YouTube and Facebook demonstrated what he described as “a significant change in attitude” following P&G’s public pressure, including YouTube’s role in curbing the spread of harmful viral content such as the “Tide Pod Challenge” videos, and Pritchard noted that the major tech platforms had met approximately 90 percent of P&G’s transparency demands by early 2018.

At Meta, the outcome was the inverse: rather than demonstrating that attribution-dependent spend was wasteful, the ATT episode demonstrated how much legitimate commercial value had depended on granular, cross-app tracking, and how a single external change to that infrastructure could remove a double-digit-billion-dollar share of a platform’s annual revenue virtually overnight.

No verified public information is available detailing Uber’s post-fraud-discovery performance marketing budget allocation in subsequent years, nor is there verified public disclosure quantifying the exact proportion of P&G’s $200 million digital reduction that was fraud-related versus simply low-viewability, non-fraudulent waste. These distinctions were not separately broken out in the public record.

Strategic Implications

These three episodes, taken together, offer a composite lesson for marketing organizations rather than a single prescriptive formula. The Uber case demonstrates that attribution systems built on self-reported, vendor-supplied data create a structural conflict of interest: networks and agencies compensated for conversions have a direct financial incentive to over-claim credit, and only a deliberate, controlled reduction in spend effectively a real-world incrementality test exposed the gap between reported and actual causal impact. The P&G case demonstrates that viewability and verified reach are necessary but distinct prerequisites to attribution; an impression cannot be meaningfully attributed to an outcome if it was never genuinely seen, and transparency demands directed at platform partners can shift industry behavior when backed by an advertiser with sufficient scale and public visibility. The Meta case demonstrates the reverse risk: that attribution infrastructure, once built into a platform’s commercial model, becomes a source of fragility when regulatory or device-level privacy changes restrict the data it depends on, with consequences that can be measured in the tens of billions of dollars and can register immediately in public equity markets.

For practicing marketers, the throughline is that attribution should be treated as a testable hypothesis rather than an accepted reporting output. Incrementality testing deliberately withholding spend from a channel or audience segment to observe whether outcomes change is the method both Uber and, more structurally, P&G effectively applied, even though neither firm publicly used that specific academic terminology in the sources available. The broader implication for the industry is that as device-level tracking continues to be restricted by regulation and platform policy, the methods capable of replacing granular last-touch attribution such as aggregated measurement, data clean rooms, and statistical modeling approaches that do not depend on individual user-level tracking — are likely to grow in relative importance, though the specific competitive outcomes of that shift for any individual company are not something that can be stated as established public fact at this time.