Incrementality Gap

The Incrementality Gap describes the massive discrepancy between 'attributed' sales (what your dashboard claims) and 'incremental' sales (what actually wouldn't have happened without the ad). In digital marketing, selection bias often leads algorithms to target 'low-hanging fruit'—consumers who are already at the bottom of the funnel and highly likely to convert regardless of seeing an ad. This creates an illusion of high efficiency while masking the reality that marketing spend is often subsidizing existing behavior rather than changing it. To overcome this, marketers must move away from correlational metrics like ROAS and toward experimental designs (A/B testing, geo-holdouts) that isolate the causal impact of advertising. Without measuring incrementality, brands risk over-investing in retargeting and branded search while neglecting the broad-reach brand building that creates new demand.

The scientific scrutiny of the Incrementality Gap gained significant momentum in the mid-2010s as digital attribution models began to collapse under the weight of their own complexity. While marketers have long suspected that 'half my advertising is wasted,' the formalization of the 'Gap' was driven by economists and data scientists at major tech firms. Key foundational work was conducted by Thomas Blake, Chris Nosko, and Steven Tadelis, who published their landmark study on eBay's search advertising in 2015. Their research challenged the industry's reliance on 'Last-Click' and 'Multi-Touch' attribution by demonstrating that for a well-known brand, paid search ads often had zero or even negative incremental value. This was followed by similar revelations from Airbnb and Facebook’s internal research teams, who began advocating for 'Lift' studies over 'Attribution' models. The concept is rooted in the statistical principle of Selection Bias: the very people most likely to see and click an ad are the people already predisposed to buy the product, making the ad a symptom of the sale rather than the cause.

The Incrementality Gap functions through three primary structural distortions: Selection Bias, The Path-to-Purchase Fallacy, and Signal Feedback Loops.

1. Selection Bias: This is the 'Pizza Delivery' problem. If you pay a guy to hand out flyers for your pizza shop, and he stands right outside your front door giving flyers to people already in line, his 'conversion rate' will be 100%. Digital algorithms are essentially that guy. They are trained to find people with high 'intent.' However, high intent usually means the consumer has already decided to buy. When you target 'high-intent' keywords or 'warm' retargeting audiences, you are often just intercepting a pre-determined transaction.

2. The Path-to-Purchase Fallacy: Standard attribution assumes a linear progression where the last ad seen 'caused' the purchase. In reality, the purchase is the result of long-term mental availability built over months or years. The 'Incrementality Gap' widens when marketers ignore the 'Base Sales'—the volume of sales the brand would achieve with zero advertising. For established brands, base sales can account for 60% to 90% of total volume. Attribution models rarely account for this baseline, crediting the ad for the entire transaction value.

3. Signal Feedback Loops: Modern ad platforms (Google, Meta) use machine learning to optimize for conversions. If the algorithm sees that people who search for 'Nike shoes' buy Nike shoes, it will spend your entire budget on the keyword 'Nike shoes.' This creates a beautiful ROAS on your dashboard, but zero incremental growth because those people were already looking for you. The algorithm is essentially 'gaming' the attribution model to prove its own worth, creating a feedback loop that prioritizes efficiency over effectiveness.

Mathematically, the gap is expressed as: (Attributed Sales - Incremental Sales) / Attributed Sales. In many retargeting campaigns, this gap can exceed 90%, meaning 9 out of 10 'conversions' would have happened anyway.