Research & Data Quality

Should Experiments Use LLMs as Human Surrogates? This Study Gives a Resounding No.

This study evaluates the reasoning depth of large language models (LLMs) using the 11-20 Money Request Game, an experimental game designed to test level-k reasoning. Level-k reasoning is a theoretical framework in game theory where individuals operate at varying levels of strategic thinking. The findings of the study reveal significant differences between the responses of LLMs and human participants, highlighting the limitations of using LLMs as human surrogates in behavioral experiments. This research emphasizes the need for caution when interpreting LLM behavior as human-like, as the models often exhibit inconsistent and non-human-like reasoning patterns. The study suggests that while LLMs can provide valuable insights, they should not be relied upon as accurate simulations of human behavior.

Member Only Access

The Importance of Incrementality in Retail Media Measurement

  • INSIGHTS STUDIOS

Despite massive growth driven by significant investments, retail media performance measurement still falls short in many areas. On October 15, OptiMine and Best Buy dove deep into the use of incrementality measurement for retail media, how it works and why it is so unique in the RMN space. Attendees explored why (and how) some of the world’s largest brands have embraced it for improved success.

Member Only Access

ARF Attention Measurement Validation Initiative: Phase 2 Report (2nd Edition)

  • ARF ORIGINAL RESEARCH

Explore the latest findings from the ARF Attention Measurement Validation Initiative. The phase two report is a comprehensive examination of various attention measurement methods used in creative testing. It concludes with reflections on the challenges of attention measurement, as well as some suggestions for advertisers on how to choose and evaluate attention measurement providers.

Member Only Access

AI for Everyday Use

On November 20, the ARF held a workshop exploring prompts, personas and how to use AI responsibly. This dynamic event, designed for advertising and marketing professionals looking to explore the evolving landscape of AI-powered research, provided insights into prompt crafting. Participants also gained a deeper understanding of the promises and pitfalls of using personas in AI-powered advertising and marketing research.

Member Only Access

LOLA: Revolutionizing Content Experiments with LLM-Assisted Online Learning

In the rapidly evolving digital content landscape, media firms and news publishers require automated and efficient methods to enhance user engagement. This study introduces the LLM-Assisted Online Learning Algorithm (LOLA), a novel framework that integrates Large Language Models (LLMs) with adaptive experimentation to optimize content delivery. Leveraging a large-scale dataset from Upworthy, which includes 17,681 headline A/B tests, the study investigates three pure-LLM approaches and finds that prompt-based methods perform poorly, while embedding-based classification models and fine-tuned open-source LLMs achieve higher accuracy.


LOLA combines the best pure-LLM approach with the Upper Confidence Bound (UCB) algorithm to allocate traffic and maximize clicks adaptively. Numerical experiments on data from the website Upworthy show that LOLA outperforms the standard A/B test method, pure bandit algorithms and pure-LLM approaches, particularly in scenarios with limited experimental traffic. This scalable approach is applicable to content experiments across various settings where firms seek to optimize user engagement, including digital advertising and social media recommendations.

Member Only Access

Marketing Analytics Accelerator: 2024

The Marketing Analytics Accelerator – the only event focused exclusively on attribution, marketing mix models and the science of marketing performance measurement – returned for its ninth year on November 13. The industry’s boldest and brightest minds joined us in NYC to share their latest innovations and case studies that will improve your business outcomes.

Member Only Access

Improve Marketing Mix Model (MMM) Accuracy by Identifying these Effects

  • MSI

This study explores the identification of nonlinear and time-varying effects in marketing mix models (MMM). It highlights the challenges of conflation in model selection and proposes a framework for simulating and estimating these effects using Gaussian processes. The study emphasizes the importance of accurately identifying the underlying response to optimize marketing spending.

The research provides insights into the complexities of marketing effectiveness and offers practical solutions for improving model accuracy. By addressing the issue of conflation, the study aims to enhance the decision-making process in marketing strategies.

Member Only Access