In short: You should always use sell-out data.
Introduction
Choosing between sell-in data and sell-out data for Marketing Mix Modeling is especially relevant to FMCG companies, manufacturers and brands who distribute their products primarily through retailers or other 3rd parties.
What is sell-out sales data?
Sell-out sales data, also known as retailer sales data, refers to the sales of products directly consumers.
Sell-out sales data reflects consumers' buying behavior well, because a product is recorded to sell-out data as 'sold' at the time of consumers' actual purchase. For this reason, one should always use sell-out data in Marketing Mix Modeling.
Highest quality sell-out data comes directly from a retailer's system that directly record cash register purchases or ecom store sales. However, many Marketing Mix Models for FMCG companies leverage sell-out data from third party data companies, who consolidate sell-out data across multiple retailers, offering some incentives for the retailers to share the data.
What is sell-in sales data?
Sell-in sales data, also known as wholesale data, refers to sales of products from manufacturer to retailers.
Because of this, sell-in sales data is "lumpy" over time. For example, a manufacturer producing high-quality products that have a long shelf-life might get monthly or even less frequent orders from retailers. Since Marketing Mix Models leverage variation in sales over time, sell-in sales data is typically very difficult or impossible to leverage in Marketing Mix Modeling.