441. Does Advertising Actually Work? (Part 2: Digital)

Episode Summary

Executive Summary: The episode argues that both TV and digital advertising are far less effective than widely believed, using academic and real-world experiments—especially at eBay—to show that many ads drive little or no incremental sales. It explores how opaque incentives, data limitations, and platform economics sustain a massive ad ecosystem despite weak evidence of return, raising concerns about a possible digital ad bubble and the need for independent scrutiny.

Main Topics: TV advertising’s weak sales impact (Priority: 5/5): The episode revisits earlier findings that TV ads are far less effective at driving incremental sales than marketers assume, despite enormous spending and industry objections about brand building and competitive effects. eBay experiments and causal measurement (Priority: 5/5): Steve Tadelis describes how eBay’s paid search ads were tested through natural experiments and randomized-like interventions, revealing that brand keyword ads were mostly wasted and non-brand ads produced near-zero average sales lift. How digital ad auctions and keyword buying work (Priority: 4/5): The transcript explains search advertising as an instant auction for keywords, with paid placements appearing above organic results and advertisers bidding on branded and non-branded terms to capture demand. Opacity, incentives, and weak accountability in ad tech (Priority: 5/5): The industry’s structure—platforms, agencies, publishers, and analytics firms—creates little incentive for transparency or rigorous testing, making it easy for ineffective practices to persist. Digital advertising as a possible bubble (Priority: 5/5): Tim Huang argues that online advertising may be overvalued like past financial bubbles because its true effectiveness is hard to measure, many ads are never seen, and industry players have incentives to overstate performance. Brand safety, bots, and poor ad quality (Priority: 4/5): The episode highlights chronic concerns about ads appearing next to harmful content, the prevalence of bot traffic, and studies suggesting a large share of digital ads go unseen or add little value. Implications for the internet economy (Priority: 4/5): Because ads subsidize major platforms, media, podcasts, and even AI research, any broad correction in ad spending could ripple across the digital economy and force a search for more diverse business models.

Key Arguments: TV advertising is much less effective than conventional wisdom suggests; doubling TV ad spend yields only about a 1% sales increase. Brand keyword search ads, such as bidding on your own company name, often merely replace free organic clicks rather than create new demand. Non-branded search ads also had very limited average impact at eBay; turning off paid search in a third of DMAs reduced sales by only about 0.5%, not statistically different from zero. eBay’s internal belief that paid search drove 5% of sales was far above the measured effect, and the company concluded it was losing more than 60 cents on every dollar spent. Advertisers, agencies, platforms, and analytics vendors all have incentives that discourage frank evaluation of ad effectiveness. Digital advertising is opaque: many ads are never seen, cookies may add little marginal revenue, and automated auctions obscure how ads are priced and placed. Tim Huang argues the industry resembles a bubble because asset value is overestimated while empirical evidence of return is weak and declining attention lowers effectiveness. A healthier internet would have more diverse revenue models instead of relying so heavily on advertising to fund dominant platforms and media. Independent, third-party research is needed because industry-sponsored evidence is often biased by self-interest. Companies often keep spending because prior beliefs, organizational identity, and industry norms are hard to overcome even when data suggest waste.

Data Points: Annual U.S. advertising spend: More than $250 billion - Referenced at the start as the scale of U.S. ad spending Annual global advertising spend: Nearly $600 billion - Used to emphasize the size of the worldwide ad market TV share of U.S. ad spending: Roughly one-third - Even after digital growth, TV remains a major ad channel Super Bowl ad revenue: More than $300 million - Example of the continuing value of TV advertising Doubling TV advertising effect on sales: About 1% increase in sales - From Anna Tuchman’s study on TV ad efficacy Relative inefficiency of TV ads: 15 to 20 times less effective - Compared with conventional wisdom about TV advertising U.S. internet ad spend last year: $123 billion - Digital advertising’s scale in the U.S. Google revenue from advertising: More than 80% - Shows dependence of Alphabet on ads Facebook revenue from advertising: More than 98% - Shows dependence of Facebook on ads eBay annual marketing spend: About $1 billion - At the time Steve Tadelis worked there eBay paid search budget cut: $100 million per year - After the findings showed weak returns eBay’s belief about paid search-driven sales: About 5% of sales - Company estimate before the experiment Measured sales impact from turning off paid search: About 0.5% drop in sales - Actual observed effect in the experiment Measured return on ad spend at eBay: Losing more than 60 cents per dollar - Calculated after comparing costs and incremental sales Estimated share of internet ads never seen: Close to 60% - Cited from a Google study discussed by Tim Huang Revenue drop without cookies: About 4% - 2019 study measuring the marginal value of cookie-based targeting Banner ad click-through rate at launch: About 50% - Historical example of early novelty on the internet Banner ad click-through rate today: About 0.01 to 0.03% - Used to illustrate attention collapse over time Procter & Gamble digital ad cut: $200 million - Company test of reducing digital ad spend P&G bottom-line impact: No noticeable impact - Reported result after the spend cut Number of DMAs in the U.S.: 210 - The geographic market units used for the search ad experiment

Pivotal Quotes: "Doubling the amount of advertising would lead to about a 1% increase in sales." — Anna Tuchman: Summary of her study on TV ad effectiveness "Steve, are you driving now? Because I can't hear you, you're breaking up." — Consulting firm head: Tadelis recounts realizing the consultant was using jargon and then getting deflected "What we showed is that on average, they're losing more than 60 cents on every dollar." — Steve Tadelis: Conclusion from eBay’s paid search analysis

Implications: Marketers should demand rigorous causal evidence before spending heavily on ads. The findings suggest major inefficiencies, possible market overvaluation, and a need for more transparent measurement and diversified internet business models.