A mattress company’s advertising aimed at prospective new customers produced stronger returns than ads aimed at previous website visitors in a geography-based field experiment across Google and Facebook.
The Journal of Interactive Marketing study by Jeffrey Larson and Jeffrey Dotson challenges a familiar assumption in digital advertising: that people who have already visited a website are necessarily the best audience for the next advertising dollar.
The experiment covered a single bed-in-a-box seller. Its results should not be treated as a budget formula for every retailer, but they give advertisers a reason to test whether retargeting generates purchases that would not otherwise have happened.
Two ways to reach a customer
Retargeting shows ads to people who have previously interacted with a business, such as by visiting a product page. Prospecting seeks potential customers beyond that pool. Both can appear in the same customer’s buying process, making their individual contributions difficult to separate.
The researchers compared the strategies using a geographic field experiment involving advertising on Google and Facebook. The American Marketing Association’s research summary says the work involved more than $5.7 million in advertising spending.
Prospecting produced large and consistent returns across both networks in the experiment, while returns from retargeting were substantially lower. The study appeared online in January 2026 and in the journal’s August issue.
Attribution is not the same as an extra sale
Advertising platforms assign conversions to interactions according to their attribution rules. A customer might click an initial ad, return days later through a reminder and then buy. The last click may receive credit even if the person was already likely to purchase.
A holdout comparison asks a different question: how many additional sales occurred because advertising was shown? Businesses can compare otherwise similar groups or locations exposed to different levels of advertising, while accounting for seasonality, demand and other differences.
Such tests need enough observations and careful design. Small samples can produce uncertain results, and local promotions or inventory problems can distort comparisons. Our earlier reporting on customer review requests similarly showed why measuring downstream behavior can change an assessment of a marketing activity.
One experiment is not a universal spending split
The research concerns a durable product that customers may consider for some time before buying. A low-priced repeat purchase, a subscription or a business-to-business sale may follow a different path. Platforms, tracking rules and campaign settings can change as well.
For a marketing team, the starting point is to compare reported conversions with an estimate of incremental revenue and the cost of reaching each audience. Retargeting can still help specific customers, but a high platform-reported return alone does not establish how many sales the reminder created.
