Brain connected to multiple ad screens, symbolizing data-driven decisions in advertising.

Digital Ad Spend Under the Microscope: Can Multi-Cell Experiments Deliver Better Results?

Reese Marlowe in Business & Economy June 2025 • 5 min read.

"Rethinking digital advertising effectiveness with advanced experimental designs and data-driven decisions."

In the dynamic world of digital advertising, where marketing budgets are continuously scrutinized, accurately measuring the impact of online campaigns is essential. Traditional methods often fall short, especially when individuals have the option not to engage with an ad, leading to skewed results and ineffective strategies. This challenge calls for more sophisticated approaches to experiment design and data analysis.

Randomized experiments, where treatment and control groups are used, offer a reliable method to measure the impact of interventions. However, in digital advertising, where users can choose to ignore ads, standard experimental designs can fail to provide decision-makers with the detailed insights needed. The issue? Existing empirical methods often don't properly address the intensive margin—the ability to determine how many consumers should be targeted or how much should be spent.

To address this gap, a new method combines a multi-cell experimental design with modern estimation techniques. This innovative approach helps decision-makers gather enough information to tackle complex problems related to the intensive margin, offering a straightforward solution that doesn't require additional budget. By calibrating simulations using data from an actual Facebook advertising experiment, this method outperforms standard techniques, leading to better-informed decisions.

The Multi-Cell Advantage: A Clearer View of Ad Impact

Brain connected to multiple ad screens, symbolizing data-driven decisions in advertising.

The problem with current experimental routes is that firms want to measure the effectiveness of their digital ad campaigns, but they cannot directly randomize advertising exposure. Instead, they randomly assign consumers to be eligible or ineligible to be exposed to ads. Randomizing eligibility for ad exposure results in one-sided non-compliance because users in the eligible group may or may not be exposed. The ineligible group, in contrast, cannot view the ads at all. This experimental design is not only popular for measuring online advertising effects but it is also used in economics, political science, and medicine.

The multi-cell experimental design offers a robust solution to these challenges by providing a more nuanced understanding of ad performance. Unlike traditional methods, this design incorporates multiple experimental cells, each with different treatment conditions. This allows for the collection of more granular data and a more accurate estimation of treatment effects. The experimenter has full control over the test/control split within each cell, and so can guarantee that Pr(Zc = z|C = c) is always strictly between 0 and 1. Second, consider cases in which the probability of treatment conditional on eligibility is either 0 or 1. If v(Zc = 1) = 1, the endogeneity problem is resolved because eligibility to receive treatment becomes equivalent to exposure to treatment itself. In turn, if v(Zc = 1) = 0, this exercise becomes meaningless because it implies that it is impossible for units to receive the treatment under consideration.

Here’s how the multi-cell design works:

Multiple Experimental Groups: Units are randomly assigned across multiple cells, each representing a unique experimental condition.
Test and Control Within Each Cell: Within each cell, units are further divided into test and control groups, allowing for direct comparison of treatment effects.
One-Sided Noncompliance: Each cell features an experiment with one-sided noncompliance, meaning that while some participants are eligible for treatment (ad exposure), not all will receive it.
Propensity Score Variation: Intentionally varying the probability of treatment across cells (e.g., by adjusting budget per user) to generate a range of propensity scores.

By combining this design with modern estimation techniques, decision-makers can recover the marginal treatment effect (MTE) function. This function is what we want to inform these decisions and to recover the most common treatment effect parameters of interest, including the ATT. This makes it possible to see how different ads perform across audience groups.

Data-Driven Decisions: Optimizing Ad Spend

The rise of sophisticated experimental designs marks a turning point in digital advertising. By addressing the limitations of traditional methods and providing deeper, more actionable insights, these techniques empower advertisers to make informed decisions and optimize their ad spend for maximum impact. In the competitive landscape of online advertising, embracing these innovative approaches is the key to unlocking better results and achieving sustainable growth.

About this Article -

This article was crafted using a human-AI hybrid and collaborative approach. AI assisted our team with initial drafting, research insights, identifying key questions, and image generation. Our human editors guided topic selection, defined the angle, structured the content, ensured factual accuracy and relevance, refined the tone, and conducted thorough editing to deliver helpful, high-quality information.See our About page for more information.

This article is based on research published under:

DOI-LINK: https://doi.org/10.48550/arXiv.2302.13857,

Title: Multi-Cell Experiments For Marginal Treatment Effect Estimation Of Digital Ads

Subject: econ.em

Authors: Caio Waisman, Brett R. Gordon

Published: 27-02-2023

Everything You Need To Know

What is the main challenge with traditional methods for measuring digital ad campaign effectiveness?

Traditional methods often struggle because they cannot account for the fact that individuals can choose not to engage with an ad. This leads to skewed results and ineffective strategies in the dynamic world of digital advertising, making it difficult to accurately measure the impact of online campaigns and allocate budgets efficiently. These methods are designed in a way that makes it hard to see how much the ad campaign is working.

How does a multi-cell experimental design improve upon traditional experimental methods in digital advertising?

The multi-cell experimental design offers a robust solution by providing a more nuanced understanding of ad performance. Unlike traditional methods, this design incorporates multiple experimental cells, each with different treatment conditions. This allows for the collection of more granular data and a more accurate estimation of treatment effects. Also, by intentionally varying the probability of treatment across cells, decision-makers can recover the marginal treatment effect (MTE) function, which is crucial for optimizing ad spend and making informed decisions.

What is one-sided non-compliance and why is it relevant in the context of digital advertising experiments?

One-sided non-compliance refers to a situation where, within an experimental group eligible for a treatment (like ad exposure), not all participants actually receive the treatment. In digital advertising, this is common because users can choose to ignore or not see an ad even if they are eligible. Understanding one-sided non-compliance is crucial because it can skew results if not accounted for, making the multi-cell design particularly effective.

Explain the components of the multi-cell experimental design and how they contribute to a clearer view of ad impact.

The multi-cell design comprises multiple experimental groups, each representing a unique condition. Within each cell, units are divided into test and control groups for direct comparison. Each cell also features one-sided noncompliance. By intentionally varying the probability of treatment across cells (e.g., by adjusting budget per user), this design allows for a more detailed understanding of how different ads perform across various audience segments, providing deeper insights into the effects of each ad.

How can understanding the marginal treatment effect (MTE) function and the ATT from a multi-cell experiment lead to better results in digital advertising?

Recovering the marginal treatment effect (MTE) function allows decision-makers to see how different ads perform across audience groups. This function is used to inform decisions and to recover the most common treatment effect parameters of interest, including the ATT, or Average Treatment Effect on the Treated. By using these parameters, advertisers are empowered to make informed decisions and optimize their ad spend for maximum impact, leading to better results and sustainable growth in the competitive landscape of online advertising.