Our Blog Excerpts Savings Contact

logo

Dorset House Publishing
High-Quality Books on Software Engineering and Management.  Since 1984.
dorsethouse.com > features
Features       Excerpts       Interviews

 

iDH Sign-Up


Get Our e-News
Delivered by FeedBurner

Ten Great Truths

by Lawrence H. Putnam and Ware Myers

Adapted from Five Core Metrics. Reprinted by permission. All rights reserved. See below for copyright notice.

The truth is, everyone has had trouble with software development. The sad truth behind this first truth is that we, as individuals, as managers, as organizations, have had trouble facing up to this truth. In recent years, however, many organizations are overcoming these troubles. Others are not yet so successful. Successful organizations have found that it is the intelligence behind successful software management that makes the difference. The first step toward identifying that intelligence rests on the effective application of simple metrics to manage software development.


The Five Great Truths About Metrics


1 There must be an accurate way to represent in metric terms what goes on in software development:

  • Establish a measure of functionality beforehand.
  • Establish a measure of the productivity of your software process.
  • From this functionality, given this productivity, estimate time, effort, and defect rate.

The feature that distinguishes the accurate way from other ways is the realization that time and effort are multiplicatively intertwined; they are interdependent; they cannot, with accuracy, be planned independently of each other. Every estimate is a time-effort pair, an effort with an associated time.

2 We need this accurate way to underpin estimating, bidding, project control, and the relations between client and developer or client and outsourcer.

3 We need this accurate way, not only for these business reasons, but because it is the only way to provide development people in appropriate numbers (effort) with enough time (schedule) to do the work at the quality and reliability levels needed by the client's application.

4 We need this accurate way to provide a satisfactory working situation. The accurate way is the only way to keep staff around long enough to grow a superior organization.

5 We need this accurate way to measure a software organization's competitive standing. This measure—process productivity—tells management and staff not only that they are on the right track for improving their organization, but also that they can endure for the substantial time it takes to grow a superior organization.


The Five Great Truths About Software Development


Metric truths are of great importance, but their further consequence is to make possible the realization of the five great truths about the field of software development as a whole.

1 There must be an activity to which to apply the five core metrics. That activity is the process. A software organization must have a way of doing its work. That way may be informal or it may be as concrete as the Unified Process. At the very least, it has to be repeatable, if estimates of time, effort, and process productivity are to have any meaning.

2 There must be some standards, and there have been—a lot of them, from programming languages to text editors. In recent years, the standards idea has moved on to the Unified Process and the Unified Modeling Language.

3 There must be reuse of previously developed software to reduce the effort and time needed to develop the new product. There is considerable reuse already, under the general name component-based development. The preexistence of process and standards is a prerequisite to more extensive reuse.

4 There must be software tools to take over the routine tasks of software development. For example, there is no reason to make a developer hand-draw the rectangles, ovals, lines, arrowheads, and other features of UML when he or she can command a software tool to do so. Note, however, that first the modeling features have to be standardized before developing and marketing the tool becomes economical. As Jacobson, Booch, and Rumbaugh have noted, "Successful development of process automation (tools) cannot be achieved without the parallel development of the process framework in which the tools are to function."[1]

5 There must be a means of bringing to the developer the procedural knowledge he or she needs to accomplish the task immediately at hand. This need arises out of the reality that the volume of knowledge applicable to software development is now overwhelming, with one example alone amounting to more than 20,000 pages. Mere humans need an automated tool to guide them to what they need when they need it.


[1] Ivar Jacobson, Grady Booch, and James Rumbaugh, The Unified Software Development Process (Reading, Mass." Addison-Wesley, 1999), p. 29.

AddThis Social Bookmark Button

 

 

 


COPYRIGHT NOTICE: This excerpt from Five Core Metrics [0-932633-55-2] appears by permission of Dorset House Publishing. Copyright © 2003 by Lawrence H. Putnam and Ware Myers. All rights reserved. See http://www.dorsethouse.com/books/fcm.html. The material contained in this file may be shared for noncommercial purposes only, nonexclusively, provided that this Copyright Notice always appears with it. This material may not be combined with advertisements, online or in print, without explicit permission from Dorset House Publishing. For copies of the printed book or for permissions, contact Dorset House Publishing, 1-800-342-6657, 212-620-4053, http://www.dorsethouse.com, info@dorsethouse.com, New: 3143 Broadway, Suite 2B, New York, NY 10027 USA. Additional rights limitations apply, as presented in the Legal Disclaimer posted at http://www.dorsethouse.com/legal.html.

 

 

 
  DORSET HOUSE PUBLISHING CO., INC.
New:3143 Broadway, Suite 2B  New York, New York 10027  USA
1-800-DH-BOOKS or 212-620-4053, fax 212-727-1044
Copyright © 1996-2008 by Dorset House Publishing Co., Inc. All rights reserved.
Home | Blog | Savings | Stores | Features | Titles | Authors | Subjects | Orders | About | Contact | Legal