What are incremental sales?

A customer who clicks a branded search ad on the way to buying something they had already decided to buy is a touched sale, not an incremental one. Incremental sales are the ones that would not have happened without the marketing: the extra orders a campaign caused above what the week would have sold anyway. Everything else is baseline that happened to pass through an ad.

Two methods measure incrementality. A marketing mix model does it for every channel at once, from aggregated data: it estimates the baseline and credits each channel with the sales above it, with carryover and diminishing returns taken into account, offline channels included. An incrementality test (a geo experiment or a conversion lift study) does it for one channel at a time by comparing exposed and unexposed groups, and gives a causal reading of that channel in that period.

The two are complementary. A test result and its uncertainty go into the model as a prior on that channel's return, so the model's incremental estimate starts from causal evidence where it exists.

Platform conversions count sales the platform touched, and every platform is built to claim as much credit as it can. Channels that sit at the end of the journey, branded search and retargeting above all, claim credit for sales that would have happened anyway; channels that create demand early, TV and social, rarely get the last touch and are undercounted. The platform's tracking is still the best tool for optimizing inside that platform. It just is not a currency for comparing channels or sizing a budget.

In Odins every recommendation is about incremental effect: the marginal return of the next unit, measured above the baseline, with a range.

• Baseline (base sales): The sales that would have happened without paid marketing. Everything the model credits to a channel is measured above it.
• Channel contribution: What each channel added to sales above the baseline over a period. The decomposition behind every ROAS figure.
• Incrementality test: A controlled experiment that measures what one channel causes, by comparing exposed and unexposed groups or periods.
• ROAS: Return on ad spend: revenue per unit of spend. Useful, as long as you know which ROAS you are looking at.
• Attribution: Crediting conversions to the touchpoints that preceded them. Great inside a platform, not a currency across them.

Go one level deeper: Triangulation: MMM, attribution and experiments. Or go back to all terms in the glossary.