HOW IT WORKS

How we measure commerce media incrementality

Direct integrations, a commerce graph, an ensemble causal model and a daily learning loop. All feeding signals back into the platforms that run your media.

WHAT IT DOES BEST

Three principles run through every step.

/ 01

Causal Inference

Multiple causal techniques link investment to outcome, not the click that came last.

/ 02

Continuous Learning

Daily retraining keeps insights fresh, not stuck on last quarter's reality.

/ 03

Speed & Granularity

Reporting down to the campaign and line item, ready to drill down or roll up.

01

INTEGRATIONS

Direct connections to retailers, media, and the digital shelf.

A few clicks connects you to retailer and platform data with out-of-the-box connectors, or bring your own data.

Incremental partners including TikTok, Google, Walmart, Amazon, Shopify, Target, Ulta, Meta, and others.

02

CLEANSING & NORMALIZATION

A graph that maps every product, campaign, promo, and shelf signal.

The commerce graph is the foundation. It links what each campaign targeted to where products were on promotion and how they appeared on the digital shelf. AKA, the context generic measurement skips.

03

UNTANGLING INCREMENTALITY

An ensemble causal model decomposes sales into what caused them.

Econometric techniques combined with designed and synthetic experiments separate media impact from other factors by retailer and channel, down to the line item.

Diagram illustrating different types of experiments and metrics in data analysis, including passive and active experimentation, econometrics, and model refinement.

04

CONTINUOUS LEARNING

Daily retraining keeps the model in sync with the business.

Short-term predictions are compared to actuals on a daily basis, and the delta becomes feedback. Recency bias is intentional: smart decisions require fresh data.

Diagram illustrating data analysis and experimentation methods with central AI or machine learning icon. Four sections include: top left, "Econometrics" showing forecasting techniques with steps like grouping, categorizing, holdout, calibration, validation; top right, "Active Experimentation" with bar charts labeled geo-based and randomized control; bottom left, "Econometrics" displaying in-market forecasts with a timeline from today to tomorrow; bottom right, "Passive Experimentation" with disclaimer "Quasi Experiments" and checklists for sneaker feature comparisons including price, shelf, media, and competition.

05

DATA SHARING

Signals flow into the platforms that already run your media.

Direct API integrations with omnichannel media buying platforms to enable incrementality signals and recommendations to flow seamlessly into campaign management workflows. We can also drop it into your internal data lake, or just send you a daily email.