T-SQL Tuesday #107 - Death March

Page content

This month’s T-SQL Tuesday is hosted by Jeff Mlakar and he asks us to write about a project that went horribly wrong. My story isn’t really worthy of the name “death march” but it was a pretty rough project.

The project started sometime in mid-2003. I was working as a web developer (Classic ASP) for an insurance company and they wanted to modernize the underwriting process with a web-based “workflow” application.

A platform was selected, the team was picked, and we set about customizing the devil out of it (it was more of a “framework” than turnkey application) and integrating with our existing systems. Of course, being an insurance company everything ultimately had to land in the mainframe.

The tech team was pretty large - upwards of a dozen and a half programmers across various disciplines. Four of us were kind of off on our own building the ASP front end and COM-based middleware - myself and Ned on the ASP side, Tony & Phil (all names changed) on the other side of the cube walls working on COM objects. The four of us worked really well together, with Phil & I each taking the lead in our respective disciplines. Our PM was great; we got along well and she knew the right questions to ask me to nudge me in the right direction.

As Phil & I dug into the platform API, we realized we were going to be implementing more features with our own code than expected as they weren’t actually built into the API yet.

We brought a few consultants from a large company you’ve certainly heard of to help us work out a UI design. After 5 weeks, we presented our design proposals. Every - I mean every - proposal put forth by the consultants and me was shot down. The response was “those are very nice, but we want what we described to you 2 months ago, so build that.” I was flabbergasted. I asked the consultants if they’d ever seen anything like that and they admitted that they hadn’t. We just spent five weeks and a bunch of money to produce nothing.

February was approaching and while Phil, Tony, Ned and myself were firing on all cylinders, others weren’t doing as well. We were supposed to launch in March, but it was becoming apparent that we’d miss that date. The release date slipped to mid-April.

The four of us spent the last half of February and all of March tightening up our code. We had some performance concerns but hadn’t figured out where they were coming from yet. Acceptance testing hadn’t started yet; the project was expected to be “nearly complete” before that’d start.

Acceptance testing started in late March. Ned & I knew our code inside and out, and I could suss out pretty easily if a reported issue was there or in the COM objects so I could throw those bugs over the wall to Tony & Phil as necessary.

I returned from lunch on the last Friday of March and Ned said to me “I’m leaving at the end of the day.” Um…OK, thanks Ned, I am too. Then he clarified. See, Ned was a contractor and he he was ending his contract at the end of the day - and starting a new gig in Miami on Monday! We were unable to bring in a replacement, so I was flying solo for the rest of the project.

I spent April fixing bugs as fast as I could, re-implementing some features because they sounded good on paper but once seen in action, they weren’t acceptable. Mid-April came around and the project release date was pushed back by another month.

I went to the PM and expressed my concerns about the release date changes. I reminded her that I was getting married Memorial Day weekend and I’d be unavailable for 2 weeks due to that and the honeymoon. She assured me that this was the last reschedule, and even if that turned out to not be true, I wouldn’t be asked to change my plans.

Early in May was the final blow for me. We had an all-project meeting with several people from the highest levels in the company in attendance. The missed deadlines, the incredible strain placed on not only the development team but their families, the importance of the project to the company, all brought up. People laid some very personal feelings out. The response from one of the top-level people was cold, harsh, uncaring, impersonal, tone-deaf, and completely disheartening.

The project went live 10 days before I left for my wedding and honeymoon.

Epilogue

No project implementation is perfect out of the gate. Right before we went live, Phil figured out what was causing our performance issues and over the next month or two, more or less re-implemented that portion of the vendor’s API himself. We pushed out an updated version as quickly as we could; users were complaining about performance from day one. We got a lackluster “yeah, I guess it’s better now” but no real acknowledgement of the improvement Phil had made so quickly. It was too late; the perception was that this system is slow and that wasn’t going to change, now matter how fast we made it.

The project was started in the summer or early Fall of 2003 and went live in the last third of May 2004. By the end of the first quarter of 2005, the company had decided to go in a different direction and the system was on track to be mothballed; no new business was being scanned into it, and once the last piece of business that was in it was processed, everything was shut down.

The system spent more time under development than it did in use.