The advertising industry told itself a story about 2020. It went like this: consumers were displaced. Behaviors were suspended. When the emergency lifted, the old machinery would resume. Brands would reassert their equity. Media plans would return to their pre-pandemic architecture. The interruption was an interruption.
Five years later, the machinery has not resumed. The behaviors that were supposed to be temporary turned out to be the permanent record. And the industry's response — the way it reorganized its budgets, restructured its agencies, and redefined what measurement means — has become a structural condition it cannot reverse without dismantling the system it built in the emergency.
This is the recalibration that keeps giving. Not because it solved anything, but because the solution became the new problem.
In March 2020, Omnicom Media Group measured brand loyalty at 65%. By November of that year, it had fallen to 49%. Sixteen percentage points, evaporated in eight months.
The industry expected a recovery. It did not come.
McKinsey's consumer surveys found that 75% of U.S. consumers tried new shopping behaviors during the pandemic. Of those who switched brands — whether due to availability, value, or alignment with social positions — 90% planned to continue the new behavior in 2022. The Ketchum Brand Reckoning 2020 found 45% of Americans had changed brand preferences, with 62% expecting those changes to be permanent.
They were right. As of late 2020, the loyalty score had not recovered, and no publicly available tracking survey has documented a return to Omnicom's pre-pandemic 65% baseline. The industry learned to operate in the new territory rather than exit it.
This matters for advertising because loyalty is not merely a brand metric. It is the mechanism by which brands amortize the cost of acquisition over time. A loyal customer reduces the per-conversion cost of every downstream marketing investment. When loyalty erodes, the cost of acquisition compounds. Every new customer must be won again. The advertising system, built on the assumption of a returning baseline, found itself spending into a hole that wouldn't fill.
The academic literature on habit formation offers a partial explanation. The Habit Discontinuity Hypothesis — documented by Verplanken and Wood in 2006, and cited extensively in COVID-era behavioral research — holds that context changes can "unfreeze" established habits, creating windows for new preference formation. The pandemic was the largest context change in modern economic history. It did not just disrupt behavior; it opened a period in which existing brand relationships were renegotiable.
What Salon et al. found in PNAS (2021) was that more than 70% of respondents wanted to maintain some pandemic-era behavior. The doubling of telecommuting — from 13% to 26% — was expected to persist. The shift toward digital grocery, while partially reverting in pure transaction volume, left infrastructure and expectation changes that did not revert.
The loyalty collapse was not temporary confusion. It was the permanent re-opening of a negotiation that brands thought had closed. (Not all pandemic effects proved durable: Inoue and Todo, using Yahoo! Shopping transaction data through 2021, found that Japan's online purchasing behavior returned to baseline trend after emergency declarations lifted, suggesting that pure e-commerce volume may be more revert-prone than the structural infrastructure changes around it.)
The industry did not respond to the loyalty collapse by rebuilding loyalty. It responded by optimizing for the measurement that loyalty makes unnecessary: last-click attribution on direct response.
The numbers are precise and they are damning. The CMO Survey (Fall 2024) found actual marketing spend running 68.8% short-term performance advertising and 31.2% long-term brand building — not the 60/40 brand-to-activation split that WARC's Binet & Field research recommends as optimal, and not even close to the stated ideal of practitioners who know better.
This is not ignorance. The WARC research program, across multiple studies by Peter Field and others, has extensively documented the "doom loop" dynamic: when brands over-invest in attribution-measured performance advertising, they sacrifice long-term brand equity, which reduces organic demand, which requires more performance spend to compensate, which further erodes brand equity. The loop is self-reinforcing and it is well-understood. Brands keep running it anyway.
The pandemic provided institutional cover. When loyalty collapsed and the measurement infrastructure could not explain why, the obvious response — and the response that the performance advertising platforms were structurally optimized to enable — was to spend more on what could be measured. That meant paid social, search, and retail media. That meant programmatic. That meant optimizing for the click, the add-to-cart, and the completed purchase, because those were the signals the system could see.
The 2021 surge in digital advertising spending — 37.5% growth according to GroupM, the strongest in four decades — looked like recovery. It was not. It was the acceleration of a structural dependency. Brian Wieser of GroupM has noted that the correlation between advertising spending and economic activity is weak in many markets, and that the pandemic's long-term impact on advertising was probably overstated. The surge was partly pent-up demand — brands that had gone quiet during the crisis returning to the air — and partly the locking-in of the performance orientation that the crisis had demanded.
The WARC analysts cut $90 billion from subsequent forecasts. The recovery moderated. The structural condition remained.
One thing the crisis did produce, with some permanence, was the validation of retail media as an advertising channel.
Retail media networks — the sponsored product placements, search ads, and display inventory operated by retailers like Walmart, Kroger, Target, and Amazon — were not invented during the pandemic. Amazon's advertising business was already generating billions. But the pandemic made the channel's value proposition legible: when consumers are shopping online and increasingly starting their product searches on retail platforms rather than search engines, the retailer controls the moment of consideration that advertising has always wanted to own.
The numbers are now large and they are growing on a different curve than the rest of advertising. Insider Intelligence estimated U.S. retail media at $41.37 billion in 2022, projecting $109.40 billion by 2027 — more than doubling, accounting for more than a quarter of U.S. digital ad spend. (A separate "commerce media" figure that includes broader off-site and closed-loop formats runs higher; the U.S. retail media network figure alone is the $109B figure.) The Channel Mix — retail media networks capturing share from Google search — has become a documented concern in media plans. Amazon's ad revenue grew from $31.2 billion in 2021 to $56.2 billion in 2024. Walmart Connect grew 61.9% from 2021 to 2023.
What makes this structurally significant is not the size but the position. Retail media operates at the point of purchase intent. It is the one channel where advertising and commerce are not adjacent — they are the same moment. This changes what measurement can see. When an ad on a retail platform is followed by a purchase on the same platform, the attribution problem largely dissolves. The sale is the measurement.
This is precisely why it is addictive to advertisers in a low-loyalty, attribution-opaque environment. When you cannot trust your brand equity to carry customers to the finish line, you pay for the customer at the finish line. Retail media lets you do that. The channel's growth is partly a rational response to the measurement failure elsewhere in the system.
The most recent data confirms the trajectory. Q1 2026 retail media spend rose 27% year over year, with clicks up 38% and CPCs declining across every category — the first broad-based efficiency gain on record (Skai Q1 2026 Quarterly Digital Trends Report, April 29, 2026). IAB's 2026 Outlook Study, released January 28, 2026, projects U.S. commerce media advertising to grow 12.1% in 2026. Meanwhile, the IAB found that customer acquisition as a primary objective has declined 10 percentage points year-over-year, while repeat purchase focus has nearly doubled since 2024 — the structural shift toward retention that the doom loop was supposed to solve is now being explicitly named as a growth lever, not just a defensive posture.
The reason performance advertising grew to dominate is that it appeared to offer something the brand side could not: accountability. If the brand cannot be measured, and the sale can be, you optimize for the sale.
The problem is that the sale is not the brand. And the measurement infrastructure that made the sale legible has been systematically dismantled.
Apple's App Tracking Transparency (iOS 14.5, April 2021) eliminated the IDFA identifier for mobile app tracking, collapsing mobile attribution almost overnight. Google replaced Universal Analytics with GA4 (July 2023), a different measurement architecture that no longer provides the same cross-session visibility. Third-party cookie deprecation, repeatedly delayed, was finally suspended indefinitely in 2024 — but the industry had already restructured around its anticipated arrival.
The result is that the measurement stack that made last-click attribution look reliable has been hollowed out. Multi-touch attribution models, which require tracking signals across sessions and devices, now operate on dramatically reduced signal. Marketing Mix Modeling — aggregate, time-series regression analysis that is privacy-immune — has experienced what Deloitte calls a "renaissance," as practitioners seek frameworks that do not require individual-level tracking.
Incrementality testing — randomized control experiments comparing exposed and unexposed groups — has emerged as the most defensible way to measure advertising causality. But it is expensive, slow, and operates at the campaign level, not the optimization level. The platforms that sell advertising cannot run incrementality tests on their own auctions without undermining the real-time bidding logic those auctions depend on.
What this means, stated plainly: the industry organized itself around a form of measurement that is no longer available, and it has not yet found a replacement that preserves the strategic architecture the old measurement enabled.
The industry told itself it was waiting for things to return to normal. The cookies would be replaced by something equivalent. The attribution models would recover. The behaviors would revert.
The cookies were not replaced. The behaviors did not revert. The attribution models did not recover.
What happened instead was the construction of a new system — one optimized for a world of low loyalty, opaque attribution, platform-controlled measurement, and retail media as the only channel with a clean signal. The system is not irrational given its premises. It is rational given a set of premises that happened to be set during an emergency and then locked in.
The recalibration is not an event on a timeline. It is the permanent record of what the industry decided to do when it thought it had no choice. It has the choice now. It is not taking it.