This session reintroduced CASE methods as "software engineering with the aid of computer-based tools and methods". Motherhood and goodness abounded, and praise was lavished on those who embark on serious CASE methodologies. Save money! Beat the competition! I suppose it was only incidental that the speaker is a consultant who stands to gain by a headlong rush to CASE...

Software engineering discipline becomes more critical as the size of your operation increases. There are industries wherein Information Systems are the lifeblood of the business; airlines and railroads immediately come to mind. As a program nears release into production, the cost associated with extinguishing a bug increases geometrically. According to the government General Accounting Office, the cost of each code defect is:

$7	- if found by the programmer
$70	- if discovered during unit testing
$700	- if discovered during system integration
$7000	- if found during release QA
$70000	- if discovered in the field

There are horrible examples. USA Today reported that Federal Express spent between ten and fifteen million dollars on a single bug.

These aren't the only figures the GAO has to offer. The GAO estimates that each line of code costs about $300 to develop and fully document. Your decision to buy or build a new system should probably take numbers like these into account. Look at these national averages:

• 6 defects per 100 lines of code written
• 3 defects per 1000 lines of code delivered to the user
• 100-1000 lines of code produced by each programmer per month (yes, that is per month)
• 30 words of English documentation for each line of COBOL code
• 200 words of English documentation for each line of COBOL code for a military installation (!)

Ah ha, the session finally rolls around to the topic at hand, viz. "Software Metrics". You can measure lots of things: productivity, quality, complexity, reliability or sheer volume. The session stopped here without talking much about the mechanics of doing any of these. Too bad.