Limit the Moving Parts

It was a crushing defeat. One of the Big Three consulting companies had swooped in and won the contract out from under our noses, promising “economies of scale” and “years of experience”. Ten months later, however, the client was back, asking us to take over the project.

It was a Sales Force Automation system for one of the biggest consumer appliance manufacturers in the world. We priced out a system using Windows-based laptops, and a 4GL/database system to build the app with.

The night before the contract was signed, a sales rep from a different hardware platform ... let us call it Granny Smith .... took the client SVP in charge of the project out to dinner and convinced him that the whole thing could be done on their soon to be released first foray into the laptop hardware / operating system space. VP happily agreed, so the deal was struck.

The OOPS!

Me? I got to develop a system on beta version of the hardware, using a beta version of the OS, with beta versions of the database and the 4GL programming system.

It went about as well as you would expect.

For the 4GL, I got to the point where I built a spreadsheet with bugs, and once a week I would send it in. Eventually, about the time I filed my 600th bug, the company that made the 4GL assigned a senior QA resource just to me. All she did was verify and replicate the issues I was finding.

At one point, you could ask the database to "Select A, B, A + B from TABLE" and the database failed at adding A and B together correctly.

At another point I could run the program on a Windows-based laptop and a similarly spec'ed Granny Smith one and show that the Windows version was 10 times faster.

I spent a lot of time debugging someone else's software that year. Any one of those four items being in beta was a risk to the project. Having the four major components all in beta was a recipe for frustration, sleepless nights, and lots of conference calls.

The Lesson

Limit the moving parts. I didn't have a choice, but maybe you can push back against too many untested tools in your project.

At least I got a nice dinner out of it at the 4GL makers annual conference at Disney World the next summer. And a cool T-shirt, of course.