Analytics the Right Way

Chapter 8Descriptive Evidence: Pitfalls and Solutions

Descriptive evidence typically comes from the analysis of historical data. We find data sets ripe for analysis in our enterprise data lakes, in our CRM and point-of-sale and ERP and HR systems, in the Excel spreadsheets that we have created by organizing the information we gather, and in publicly accessible data sets that we can download from the government and other sources online. We use a subset of this data for performance measurement, as we discussed in Chapter 5, but the totality of collected and managed data is both broader and deeper than the aggregated rollups that we use for tracking KPIs against their targets, which means this data can be a mechanism for validating hypotheses on short notice.

In fact, the descriptive analysis of already-collected data is the workhorse of most analytics teams, which are under pressure to produce insights for urgent priorities and without much resourcing. In a business context, the main benefits of historical data analysis are twofold:

• The data have already been collected—by definition, there is no need to plan or instrument any additional data collection because the data already exists!
• The hypothesis validation work can be done immediately—since the data already exist, they can be analyzed at any point; there is no “wait for the experiment to run and then analyze the resulting ...