Article: Long-Term Persuasion
Reviewed: November 2026 · Method version 2 · How these reviews work
Each factual claim in the article was checked against the source it cites, then challenged by reviewers whose job was to find the strongest evidence against it: replication attempts, meta-analyses, later papers that cite the original work, and research from other fields that study the same mechanism. A separate reviewer weighed both sides, and a script re-checked every quoted source. Where a claim was overstated or misattributed, the article was corrected. The research below supports the article as it now reads.
In October 2019, Adidas’s global media director, Simon Peel, told the EffWeek effectiveness conference that the company had been measuring the wrong things. Four years earlier its attribution was last-click, it had no econometric model and no brand tracking, and it split advertising 23% to brand building and 77% to performance.
Marketing Week dates Adidas’s last-click, no-econometrics, no-brand-tracking setup to about four years before Peel’s October 2019 talk; the article now says so.
Once Adidas built an econometric model, the picture turned over. Brand activity was driving 65% of sales across wholesale, retail and ecommerce, though performance advertising also drove wholesale and retail sales, and 60% of revenue came from first-time buyers rather than the loyal customers its CRM spending targeted.
Marketing Week’s account adds that performance advertising also drove Adidas’s wholesale and retail sales; the article now includes that. The figures come from Adidas’s own econometric model, which has not been published.
Last-click reporting sees what arrives quickly and attaches to a click. Persuasion that builds over months mostly leaves nothing for it to attach to.
Large advertising field experiments (Lewis & Rao 2015; Gordon et al. 2019) show that observational measurement, including click-based attribution, often misreads what advertising does.
Letting last click settle budgets. It tends to credit the channel that caught demand, not the one that created it, which is what Adidas found.
Field experiments at eBay (Blake, Nosko & Tadelis 2015) and Facebook (Gordon et al. 2019) show that click-based and other observational attribution overstates the returns of channels that reach people already likely to buy.
Refreshing brand cues every season. Frequent changes can throw away part of a memory that took years to build, and loyal customers notice most. Vary the executions, not the cues.
Repetition research (Schmidt & Eisend 2015) supports keeping cues consistent so exposures add up. A 2022 study of TV creative found consistent templates and varied executions both raised advertising response, and logo-redesign research (Walsh, Winterich & Mittal 2010) found changes upset committed customers more than others.
The average short-term elasticity was 0.12; the average long-term elasticity, drawn from a partly different set of studies, was 0.24, twice as large. A later study of 288 brands found TV advertising effects smaller still, so treat these as upper-end averages.
Shapiro, Hitsch & Tuchman (2021) estimated TV advertising effects for 288 brands and found them substantially smaller than published averages, with many not significant.
Put plainly, a measurement window that closes with the campaign sees, on average, about half of what the advertising did. In-market TV tests found the same pattern: when extra advertising worked in its first year, its effect over three years was about double, but when it showed nothing in year one, nothing showed up later either.
Lodish et al. (1995), summarizing 55 in-market TV experiments, found long-term effects roughly doubled first-year effects when there was a first-year effect, and found no later effect when there was none in year one.
Quoting 60:40 as a law. It comes from award entries, not experiments, and Binet and Field’s later work found the right balance differs by category. Use it as a starting question for your own mix.
Binet and Field’s 2018 report Effectiveness in Context set different brand-building and activation balances for different categories rather than a single 60:40.
In Binet and Field’s cases, large reductions in price sensitivity went with brand-building effects; a voice built on offers teaches people to wait for the next one.
Long-run panel data (Mela, Gupta & Lehmann 1997) and experiments on price expectations (Kalwani & Yim 1992) support the point that frequent promotions make customers more price- and deal-sensitive.
What the case shows is narrower and still useful: by Airbnb’s own account, the strength of its brand let it rely less on traffic bought one click at a time.
The Airbnb conclusion is now framed as the company’s own account, since the 10-K credits brand and communications strategy and a pandemic recovery cannot isolate brand effects.
These edits were made on November 12, 2026 after the review.
Claim reviewed: In October 2019, Adidas’s global media director Simon Peel told a London effectiveness conference that Adidas’s attribution was last-click, it had no econometric model and no brand tracking.
Now reads: In October 2019, Adidas’s global media director, Simon Peel, told the EffWeek effectiveness conference that the company had been measuring the wrong things. Four years earlier its attribution was last-click, it had no econometric model and no brand tracking, and it split advertising 23% to brand building and 77% to performance.
Why: Marketing Week confirms Peel spoke at EffWeek in October 2019 and describes last-click attribution, no econometrics and no brand tracking. But the source places that setup four years before the talk, and says Adidas had been fixing it since. The article reads as if this was still the setup in 2019, and the source does not say London.
Claim reviewed: Once Adidas built an econometric model, it found brand activity was driving 65% of sales across wholesale, retail and ecommerce.
Now reads: Once Adidas built an econometric model, the picture turned over. Brand activity was driving 65% of sales across wholesale, retail and ecommerce, though performance advertising also drove wholesale and retail sales, and 60% of revenue came from first-time buyers rather than the loyal customers its CRM spending targeted.
Why: The 65% figure is in the source. The article leaves out the same sentence’s point that performance advertising also drove wholesale and retail sales, so the reversal looks cleaner than it was. The figure comes from Adidas’s own unpublished econometric model, and field experiments show that model-based ad measurement can be badly off.
Claim reviewed: At Adidas, performance channels had been taking credit for demand that brand work created; last click credits the channel that caught demand, not the one that created it.
Now reads: Letting last click settle budgets. It tends to credit the channel that caught demand, not the one that created it, which is what Adidas found.
Why: Marketing Week supports the idea that Adidas’s attribution over-credited digital and performance channels, including a Latin America episode where paid search stopped and revenue did not dip. Large field experiments at eBay and Facebook show the same bias in general. The Avoid-list line turns one company’s finding into a flat rule; ‘tends to’ is closer to the evidence. Now cites Blake, Nosko & Tadelis, Consumer Heterogeneity and Paid Search Effectiveness: A Large-Scale Field Experiment, Econometrica (2015).
Claim reviewed: Last-click reporting sees what arrives quickly and attaches to a click; persuasion that builds over months leaves nothing to attach to.
Now reads: Last-click reporting sees what arrives quickly and attaches to a click. Persuasion that builds over months mostly leaves nothing for it to attach to.
Why: Experiments show that click-based and other observational measurement misreads advertising effects, which supports the first sentence. But studies of advertising carryover put most lasting effects at months, not years, and in-market TV tests found that long-term effects rarely appear without some short-term effect. ‘Leaves nothing to attach to’ is a little stronger than the evidence. Now cites Gordon, Zettelmeyer, Bhargava & Chapsky, A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook, Marketing Science (2019).
Claim reviewed: Sharp’s later work calls the same idea mental availability.
Now reads: Sharp’s later work builds on the same idea under the name mental availability.
Why: How Brands Grow does make mental availability a central idea, but that comes from the reviewer’s knowledge of the book; the cited catalog link could not show the text. Ehrenberg-Bass later measures mental availability more broadly than salience, through the many buying situations a brand is linked to, so ‘the same idea’ simplifies a little.
Claim reviewed: Every exposure, including product touchpoints, adds to brand memory, and whether exposures add up to a memory depends partly on whether they look, sound and behave like the same brand; refreshing brand cues every season discards part of that memory.
Now reads: Refreshing brand cues every season. Frequent changes can throw away part of a memory that took years to build, and loyal customers notice most. Vary the executions, not the cues.
Why: Repetition research supports the core idea that repeated, recognisable exposures build memory. But a study of 91 brands’ TV ads found that varying creative over time also raised advertising response, and logo-redesign experiments found changes hurt committed customers while helping weakly committed ones. ‘Each change discards part of a memory’ states as fact something the evidence shows only for some people and some changes. Now cites Schmidt & Eisend, Advertising Repetition: A Meta-Analysis on Effective Frequency in Advertising, Journal of Advertising (2015).
Claim reviewed: The average short-term advertising elasticity was 0.12 and the average long-term elasticity was 0.24, twice as large.
Now reads: The average short-term elasticity was 0.12; the average long-term elasticity, drawn from a partly different set of studies, was 0.24, twice as large. A later study of 288 brands found TV advertising effects smaller still, so treat these as upper-end averages.
Why: Both averages are in the abstract. The two means come from partly different sets of studies (751 and 402 estimates), so ‘twice as large’ compares averages rather than paired results. A later study of 288 brands found TV advertising effects much smaller than the published literature, which suggests these averages may run high. Now cites Shapiro, Hitsch & Tuchman, TV Advertising Effectiveness and Profitability: Generalizable Results From 288 Brands, Econometrica (2021).
Claim reviewed: A measurement window that closes with the campaign sees, on average, about half of what the advertising did.
Now reads: Put plainly, a measurement window that closes with the campaign sees, on average, about half of what the advertising did. In-market TV tests found the same pattern: when extra advertising worked in its first year, its effect over three years was about double, but when it showed nothing in year one, nothing showed up later either.
Why: ‘About half’ is the article’s inference from 0.12 and 0.24, not a statement in the paper, and it holds only if the long-term figure includes the short-term one. In-market TV experiments back it up: when extra advertising worked in its first year, the total over three years was about double. Those same tests found no later effect when there was no first-year effect, and ‘short-term’ in these models usually means one data period, not a campaign. Now cites Lodish, Abraham, Livelsberger et al., A Summary of Fifty-Five In-Market Experimental Estimates of the Long-Term Effect of TV Advertising, Marketing Science (1995).
Claim reviewed: Elasticities were higher early in a product’s life cycle than in maturity, and they have declined over the decades.
Now reads: The same paper found short-term elasticities were higher early in a product’s life cycle than in maturity, and that they have declined over the decades.
Why: The abstract reports both findings for short-term advertising elasticity. It adds that many, but not all, short-term results also hold for long-term elasticity, and the abstract does not say which. The article states them for all elasticities.
Claim reviewed: The IPA analysis reports the most very large effects when about 40% of the budget went to activation and 60% to brand building; the ratio is a heuristic from self-selected award entries, not a law.
Now reads: Quoting 60:40 as a law. It comes from award entries, not experiments, and Binet and Field’s later work found the right balance differs by category. Use it as a starting question for your own mix.
Why: The 60:40 chart and its roughly 40% activation optimum are in the presentation, and the article already calls the ratio a heuristic from self-selected cases. Binet and Field’s own later work set different balances by category and buying context, which is worth saying so readers do not treat 60:40 as universal. Now cites Binet & Field, Effectiveness in Context, IPA (2018), as summarized by System1.
Claim reviewed: Years of brand memory gave Airbnb traffic it no longer had to buy one click at a time.
Now reads: What the case shows is narrower and still useful: by Airbnb’s own account, the strength of its brand let it rely less on traffic bought one click at a time.
Why: Airbnb’s filing does say brand strength let it rely less on performance marketing, and paid-search experiments show strong brands get much of their traffic without paid clicks. But the filing credits its brand and communications strategy, not ‘years of brand memory’, pandemic travel shifts muddy the case, and the filing itself warns the cut may not hold. The conclusion should be stated as Airbnb’s own claim.