Marketing science

Marketing science applies statistical models and experiments to marketing decisions: what did each channel add to sales, what would the next euro add, and how sure are we? The methods covered on this site include media mix models, incrementality experiments, response curves and audited benchmarks, and its discipline is to say, for every figure, whether it was measured, projected or borrowed.

It is not the same as marketing analytics or reporting. A report describes what happened, usually as the platforms attributed it. Marketing science looks for the counterfactual, what would have happened without the advertising, and builds it honestly, with an interval. It is also older than digital advertising: marketing mix models were the packaged-goods industry’s way of measuring television and promotions from aggregate sales, and they are back because click tracking has become less reliable as a way to measure advertising.

The role that does this work is the marketing scientist, or MMM data scientist: a large part of it goes into preparing the data, reconciling platform exports with the warehouse, documenting the promotions and price changes nobody logged, and deciding what counts as a channel. Marketing science is still often delivered as a bespoke project, where a single model can take months and the answer arrives after the next plan is live. Kuwalyst is a marketing-science company based in Paris: it brings that work into software, the Media Planner, and keeps marketing-science expertise for the decisions that need it. Hajime Takeda, co-founder and Head of Marketing Science, wrote the method sheets on this site and the book Marketing Science in Python. More on the company page.