The Book of Alternative Data

CHAPTER 5The Processes behind the Use of Alternative Data

5.1. INTRODUCTION

As explained in a lot of detail in the previous chapters, there are several potential pitfalls in the implementation of an alternative data strategy. In this chapter, we will discuss how to organize these implementation efforts in order to deliver a successful strategy. We must emphasize that the key to this success is to have the right processes, systems, and people in place. There are, of course, external constraints, such as the availability of talent in the marketplace, or internal ones, like budgeting and legacy systems. These will also determine how successful one can be in implementing an alternative data strategy.

We must also note that once a given strategy is deployed in production, that does not mark an end of the work. In fact, the signals generated from alternative data may begin to degrade. This means that the accuracy and performance achieved in model development and as measured through backtesting declines in time. Reasons for this can range from non-stationarity through technical implementation problems. While we cannot solve the problem of non-stationarity except under very special circumstances, we can address most other performance degradation issues by establishing and acting upon a suitable monitoring process. We will discuss this matter further.

We will divide the process of developing an alternative data process into a number of steps as follows:

• Set up the vision and strategy. ...