A new approach was employed to improve Army software sustainment cost estimation: Causal Data Analysis. The Army has conducted a study over the past seven years to improve the estimation accuracy of software sustainment systems cost using data from 192 Army software systems. This past year, the focus was on first analyzing the causal relationships in the data before modeling.

Causal analysis examines the cause/effect relationships in data. The analysis can be structured to look for causes of a desired output in a Cost Estimating Relationship (CER). The discovery of causal relationships identifies inputs for a CER and prevents incorporating spurious correlations into a CER, e.g. shark attacks increase with higher consumptions of ice cream. The search for causal relationships has also provided additional insight into non-CER cause and effect relationships in the data.

This presentation will show the results of Army Software Sustainment causal analysis and the resulting CERs.

Authors:
Cheryl Jones, US Army Futures Command
Paul Janusz, US Army Futures Command Command
Bradford Clark, Software Metrics Inc.
Christian Smart, Galorath Federal, Inc.
Kimberly Roye, Galorath Federal, Inc.