Research

Quantitative research from people
who've built careers on this

The research and white papers published here are the product of a team with decades of combined experience across data science, statistics, and performance marketing.

Featured

OMEN: Identifiability, Inductive Bias, and Mechanistic Modeling in Marketing Mix Models

Marketing Mix Modeling (MMM) is widely used to guide large-scale marketing investment decisions, yet practitioners routinely observe instability, implausible elasticities, and contradictory budget recommendations across model implementations. We argue that the primary explanation is structural: the separable, additive factorization used by most MMMs conflicts with actual marketing system dynamics, leading to non-identifiability under realistic conditions. We compare OMEN, a mechanistic and nonlinear MMM, against two widely used open-source baselines using a synthetic, agent-based marketing environment with known causal ground truth. Across ten independently generated datasets, OMEN exhibits substantially lower error in recovering true incremental contributions, sharply reduced baseline leakage into media attribution, higher peak-lift fidelity during high-demand periods, and markedly lower optimization regret under counterfactual budget shifts.

June 1, 2025
Cody Greco Philip Hofmeister
Cody Greco, Philip Hofmeister

About our research

Deep expertise

The research and white papers published here are the product of a team with decades of combined experience across data science, statistics, and performance marketing. Our researchers hold advanced degrees in quantitative disciplines and have spent their careers studying the problems that make marketing measurement unreliable at scale.

Rigorous standards

That depth of expertise shapes everything from how we frame a research question to the standards we hold ourselves to before anything gets published. Every research paper and white paper goes through rigorous internal review to make sure the methodology is sound, the claims are defensible, and the findings are genuinely useful to the marketing community.

Research-first

At Prescient AI, we operate as a research organization first. Our work isn't produced to support a marketing agenda or validate a predetermined conclusion. It's designed to advance the field's understanding of marketing mix modeling, attribution, and media measurement in ways that hold up under scrutiny.

All research

Halo Effects and CTV Attribution

Connected TV advertising drives measurable lift in direct-to-consumer channels that traditional last-click attribution systematically misses. Using a panel of 47 omnichannel brands running CTV campaigns alongside Meta and Google, we quantify the cross-channel halo effect and propose a correction factor for existing attribution models. Brands with CTV in their media mix show 23% higher baseline conversion rates on owned channels during campaign flight windows, with effects persisting 14 days post-exposure.

Prescient AI Research

Channel Interaction Effects in Omnichannel Media Mix

Additive media mix models assume channel effects are independent and separable. We demonstrate that this assumption fails systematically for omnichannel brands: channel interactions account for 18–34% of total media-driven revenue in our sample. TikTok-Meta combinations show the strongest synergy effects, with simultaneous exposure producing 2.1× the revenue of either channel alone. Ignoring interaction terms leads to budget recommendations that leave 12–19% of accessible revenue on the table.

Prescient AI Research

Media Saturation Curves in Ecommerce

Every media channel has a saturation point — a spend level beyond which incremental returns decline sharply. Yet most brands discover their saturation points reactively, through wasted spend rather than proactive measurement. We characterize saturation curves for five major ecommerce channels using data from 200+ brands, identify early-warning signals that precede saturation, and propose a continuous monitoring framework that identifies diminishing returns 3–6 weeks before they become apparent in ROAS reports.

Philip Hofmeister

Budget Reallocation Under Uncertainty

Optimal media budget allocation requires quantifying not just expected returns but the full uncertainty distribution around those returns. We develop a Bayesian decision framework for media planning that explicitly models parameter uncertainty, produces credible intervals on channel-level ROAS, and generates budget recommendations that are robust to model uncertainty. Applied to 85 brands over 18 months, the framework reduces realized ROAS variance by 31% compared to point-estimate-based allocation, with no significant reduction in mean ROAS.

Cody Greco

Retail Media Network Measurement

Retail media networks report strong ROAS figures, but a significant fraction of measured sales would have occurred without the ad exposure. We conducted geo-holdout experiments across 22 brands running retail media campaigns on Amazon, Walmart Connect, and Target Roundel, measuring true incremental lift against network-reported attribution. True incrementality averaged 41% of reported ROAS, with substantial variation by category, brand awareness tier, and campaign objective. We propose a correction methodology and a continuous incrementality monitoring framework for brands running sustained retail media investment.

Prescient AI Research

Why Attribution Fails

Attribution model failures are not random — they follow predictable patterns rooted in specific structural assumptions that misalign with real marketing system behavior. We identify and characterize six distinct failure modes: last-touch bias, same-session inflation, dark funnel blindness, baseline contamination, creative saturation miscounting, and platform self-attribution. For each, we quantify the expected direction and magnitude of the resulting measurement error and propose diagnostic tests that practitioners can apply to their own attribution data.

Cody Greco, Philip Hofmeister
More papers coming soon

The platform

A platform powered by cutting-edge research

Our suite of features is built on top of this body of research. Discover what the platform can reveal about your campaigns.

Book a demo