The president had hired a well-known programming guru I'll call "Wiley" to lead development. I think the plan was to groom this guy as an in-house management consultant whose services could be marketed to other companies, along with the software I would be developing.
Wiley had a reputation for being on top of the very latest trends in IT. I was looking forward to working with him. But on my first day at the job, I discovered we were deep in the doo-doo. The president was talking about a fairly complex system -- one that would handle sales, warehousing, project management, customer access to production progress, even vacation scheduling -- and nothing was documented. I mean there was no scope, no requirements, no project plan, nothing.
To make matters worse, Wiley insisted on our doing software development in an obscure 3GL he had helped create a few years earlier -- one that I'd heard was moderately unstable. And although we were several years away from going live, Wiley had already spent a fortune on several refrigerator-size' computers and storage arrays to run the system. Meanwhile, were there platforms for development? Servers we could use for testing and production? Of course not.
When I expressed my concern about the lack of requirements, Wiley assured me he had always been successful working this way. Being able to change requirements during development, he explained, would ultimately save lots of time and produce better code than attempting to define all requirements at the beginning. And of course Wiley himself would be providing all the initial requirements.
We started on the warehouse management system. Wiley had already built and installed a component of this system at the warehouse site, something involving scanners and RF receivers, and I decided to drive out and see how it was working. When I introduced myself, I encountered a wave of hostility aimed directly at me. Everyone I spoke to, from secretaries to the warehouse manager, told me that the new system was a disaster. Wiley's scanner technology was frequently off-line, and even when it worked, it was far less efficient than the paper-based system they'd been using before.