Changes in Revenue
Prev Next

Changes in Revenue

For new projects finding the revenue base line could be very challenging. One of the most common mistakes is to assume that without an explicit Data Analytics based decision support the entire previous decision system was missing and the associated revenue ignored. In reality leaders have always made decisions more or less correct and the Data Analytic project may only eventually improve the results.

Due to their nature different types of analytics demand specialized approaches for finding the economic value under the new project. For example:
  • Predictive Analytics: represent a forward looking approach thus results may be estimated statistically, using simulations, or extrapolated from historic data;
  • Descriptive Analytics: represent a backward looking approach thus results may be estimated from historic data as cost reduction versus alternative solution.

A very interesting and extreme case is when the insight gained from the analytics forms the basis of a change in strategy. The strategic advantage thus achieved may bring economic benefits lasting many years if it can be protected by a patent or mere hours when the change is public and can be easily copied as in the on-line fight for search term supremacy.

Predictive Analytics

Predictive Analytics can support quasi continuous or discrete decisions. A discrete decision is for example to either invest or not in a new analytic system. It is called discrete because it selects from a limited number of options. A continuous decisions is for example to invest a certain amount in a new security system. It is called continuous because the amount to be invested can have almost any value.

  • Continuous predictions: the relative change in revenue is described by the distance on the Probability Density Function between the prediction and the optimal value;
  • Unique binary decision: the theoretical baseline of a skew-less purely random decision is 50% (1/2), however intuition of trained individuals commonly provides accuracy better than 65%.
  • Chained binary decisions: the theoretical global baseline of n step skew-less purely random decision is 1/2n, human intuition could be substantially better than 50%.

The following pages show valuation examples of a Data Analytic project intended to improve company's market share. In this case the missed market share is the error of the predictive analytic project. Although the predictive algorithm may be able to achieve a lower error what really matters is what the company can do with the prediction in its current corporate structure and business environment.