Introduction

Market mix modelling (MMM) has long been considered to be the gold standard of media performance measurement. It applies a robust linear regression technique to time series data to isolate the impact of media investment at a channel or total level to a KPI; outputting critical measurement criteria such as ROI, media contribution and media KPI response curves.

Traditionally, MMM has been an expensive and laborious task, often taking several weeks to complete and costing clients upwards of £30,000 a model. An in-depth data collection and review process is required, with a further initial exploratory data analysis process (EDA) to test for data issues and initial correlations. A detailed feasibility study may even reveal that the data is insufficient to proceed with modelling, a loss to both client and agency. Modelling itself usually requires a small team of experienced econometric modelers, and is generally considered as much an art as a science to create the perfect model. The modelling output is then subject to much scrutiny and client follow up, often well beyond the expected timeframes, before the media response curves, curves that show how an increase in spend for total media or a channel affects the KPI, are handed over to the planning teams for budget setting or allocation.