Making money selling “maybe” - the pricing of predictions : a literature review of pricing models of goods and services
MetadataShow full item record
- Master Thesis 
Advances in technology is a game changer for business. Today we can predict faster, cheaper and better than ever before (McKinsey, 2018), which enables humans to work smarter and faster. The technological development changes the way the world works and how businesses create, capture and deliver value. Apple transformed the music industry when they in 2003 introduced the iTunes Music Store (Apple, 2003), distributing songs separately online and sidelining the traditional CD. Using technology, they found a new way to deliver their product. When Spotify later launched in 2008, also they made individual songs available (Spotify, 2019). Changing the game was the way they charged their customers. Instead of charging for each individual song, Spotify charged a monthly fee in exchange for access to all available music files. A part of businesses maximizing their benefits from new technological opportunities lies in their pricing scheme. As technology advances and machines can do what humans do, predictions will become both better and cheaper. As a result, the use in businesses will accelerate in the time to come. The main objective of this thesis is to find out how pricing models of goods and services can be used in the pricing of AI-based predictions. Through a literature review of pricing, we identify pricing models guiding the seller in how to charge the buyer. Going through 1,745 articles we identify three broader categories; unit-based, subscription-based and output-dependent pricing. Reviewing 60 articles in detail we placed subcategories of pricing within these categories, forming a picture of the pricing literature from 2000 until today. Combining pricing models found in the literature review with characteristics of predictions we create a model for decision making. Dependent on willingness to pay and degree of judgment needed for the given prediction, we suggest a suitable pricing model. With this we aim to help the decision maker make better and more substantiated choices. In the case of low willingness to pay we suggest subscription-based pricing regardless of the degree of judgment needed. As for high willingness to pay we recommend prediction-sellers to use output-dependent pricing in the state of a low degree of judgment needed and unit-based in the state of high.