Thinking About the Proprietary Data Edge - Creating, Manipulating, and Applying

How do managers create an edge? Many will say it is associated with skill; however, that does not provide any insight. A deeper explanation is that an edge is created from proprietary data, yet there are no clear definitions or framework for how we should we think about proprietary data? 

There are three broad means of creating and using proprietary data:

1. New data / unused data - Investors will often mean alternative, new, or unused data. If there is data that others don't have, there can be an edge. If previous unavailable data is used to answer a question more quickly, it can be viewed as alternative data. If there is data that is available but not well-known, it can be an alternative data source. Of course, the edge issue is whether the alternative data is actually correlated with future prices.   

2. Manipulation of existing data - A second set of proprietary data is existing information that is manipulated in new and different ways. It could be z-scoring data or creating a ratio between two data sets. There can be value in looking at the same data differently than convention.

3. The application of data to markets - Once found, data have to be mapped or linked to market returns. There is an edge with the techniques used to relate data to a forecast. For example, the link between macro data and bond returns can be measured through linear regression or through some form of machine learning. A non-linear technique may find a relationship that may not be displayed with a linear model.

4. The use of data to make decision - Finally, there is an edge created when the manager acts on the data or converts information to action. For example, there can be two trend-followers who use similar models but generate different returns because the map between data, forecast, and action is different.

Managers should walk through their use of data to determine whether there is an information edge. Investors who select managers should try and determine what it is that the manager does with data that creates an edge.