Why SEER Got Started

Author: · August 18, 2008 · Filed Under Estimating, General, Thoughts  - 0 Comment(s)

Many years ago a project was never developed because of my realistic estimate.  I wondered if I had done the right thing, which was try to provide an achievable project plan and a realistic estimate, even though that estimate was longer than the company desired. This experience made me wonder if I had failed as a manager. However, some years later I tried to reproduce this estimate using SEER-SEM and discovered that I had significantly underestimated the project. SEER-SEM enabled me to understand that I had not failed and that my refusal to give in to the division head’s pressure had been the best thing for that company. 

 

This experience was the beginning of my quest to understand how long software development should take and how many people were needed. A few years later Don Reifer and I were tasked with developing a concept for software estimation within NASA JPL. The results of this study prompted the development of the JPL Softcost software estimation model.

 

In 1981, when Dr. Barry Boehm’s classic book, Software Engineering Economics, was published I was operating a consulting business that remediated failing projects. Although Boehm’s book was extremely useful, I still used the early Softcost model which was automated and performed risk and trade off analysis.. In 1984 I began consulting for Computer Economics, Inc., where I met Dr. Randall Jensen and was introduced to the “Jensen model”.  This model was originally known as the Modified Putnam model and had been  implemented as the JS-2. I recognized its strengths, which were its ability to conduct uncertainty estimation and minimum time estimation, but it needed better usability. After redesign, it was released as the “CEI System-3,” which was relatively successful in helping people answer the difficult questions surrounding software development.

 

In 1988 Galorath Incorporated began developing SEER-SEM, deciding to implement it under Microsoft Windows version 2. The initial product, which was approximately 22,000 lines of code, relied heavily on the mathematics that served as the foundation of Jensen’s public domain model. Dr. Jensen, Judy Crockett, Dave Lawrence, and myself (Dan Galorath) began the task. We shipped a run-time version of Windows on thirteen 5.25″ floppies. DOS was pervasive at this time, and some people complained about being required to use Microsoft Windows, but this decision was critical to SEER-SEM’s early success. At one time there were Macintosh and Sun Solaris versions of SEER-SEM as well, but both were based on the Windows code baseline.

 

As the state of software development progressed, and as user demand grew more sophisticated, SEER-SEM required continually more robust and advanced mathematics and refined knowledge bases to handle the situations our users found themselves confronting. User’s wanted to know the answers to such questions as: “How do I plan with my real-world staffing constraints”; “How do I estimate my COTS software?”, ”Do risk and uncertainty affect schedule independently of effort?; “How can I calibrate for special situations or to my particular environment?”; and a host of other situations. SEER-SEM has also evolved from using lines of code as the only size input to function points, developing its own function based sizing and the ability to accommodate: objects, web pages, use cases, classes, and a host of other sizing methods.

 

Most software estimation models have common ancestors, and SEER-SEM is no different in this regard. SEER-SEM, which was based on the Jensen model developed at Hughes aircraft, diverged significantly in the early 1990s. Earlier work by Doty Associates introduced the idea of factoring in development environment influences via parameters. Barry Boehm’s COCOMO work contributed to the original Jensen model technology parameters. Don Reifer and Dan Galorath’s research that stimulated work on the NASA Softcost model also influenced SEER-SEM, as did Halstead’s software science metrics and McCabe’s complexity metrics. Today, through the process of Galorath’s “continuous product improvement “, the SEER-SEM suite has grown to over 200,000 lines of code. Using this process, Galorath engineers analyze and begin to address software industry trends even before they become visible in the mainstream. The SEER models continue to evolve with data collection, application, research and development being conducted continuously.   Today there are four major SEER domains:

 

  • SEER for Software
  • SEER for Hardware
  • SEER for IT
  • SEER for Manufacturing

 

Thank you for reading “Dan on Estimating”, if you would like more information about Galorath’s estimation models, please visit our contact page, call us at +1 310 414-3222 or click a button below to ask sales questions, sign up for our free library or schedule a demo.

Comments