Date of Award

3-2022

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Systems Engineering and Management

First Advisor

Scott T. Drylie, PhD

Abstract

The DoD and the software development industry as a whole has long dwelt over the idea of requirements volatility (RV). The DoD has toiled with this concept so much so that its guidance was modified four times over a 13 year span. In these changes, its policy completely transformed the what, how, and when regarding RV information. As a result, the volatility data it has received is quite varied and seemingly useless for anything more than anecdotal analysis. This study takes several approaches to salvage value from this data. It begins with a survey of the uncertain concept of volatility, and provides an array of descriptive statistics to make clear what DoD currently has available for analysis. It then places volatility in its intended place as a mediator between problem characteristics and problem outcomes. It does this in two steps. First, it evaluates various volatility measurement schemes against an array of possible measures of growth where the impact of volatility may occur. This work is exploratory to determine which scheme may be most predictive and how so. Second, it identifies relationships between these various measures of volatility and the program attributes which program managers may be contemplating when they try to portray volatility. Both initial and final relationships are tested, capturing the ability of managers to assess volatility in any meaningful way at the start of the program and as a retrospective, or post-mortem. All tests are completed utilizing Contingency Analysis and the respective Odds Ratio to determine the strength of the relationships identified.

AFIT Designator

AFIT-ENV-MS-22-M-270

DTIC Accession Number

AD1174735

Share

COinS