One Project Disaster

I was involved with a project in Australia for one of the major banks. The whole project was a fiasco. One marketing guy wrote a one-page "wouldn't it be nice if..." proposal. Bank management liked the idea and allocated $10 million to the project. A project manager was assigned. He didn't have a lot of experience, but got permission to hire a company which I will call MT (not their real initials).

MT brought in 5 programmers at $2000 per day each and their own project manager at $3000 a day.

The bank project manager and MT project manager discussed the one page proposal, and designed (on a white board) a 3 tier system; the backend would use the existing IBM mainframe, the front end would be PCs located in the retail branches and used by tellers, and in the middle would be a Tandem minicomputer collecting transactions from the PCs and converting them into transactions the mainframe could handle. This took a few days, maybe a week, I forget. In the meantime, the programmers had nothing to do, so they took long lunches, read magazines, and played cards. At $2000 a day.

When the two project managers finalised their concept, the bank ordered the Tandem minicomputer.

Neither project manager thought it necessary to write a system design document. The MT project leader had a one hour talk with his 5 programmers and allocated tasks: you will work on the GUI on the front end PCs running Windows, you will work on the communications with the Tandem, you will work on SQL server, etc.

No documentation was produced.

I was hired to manage a test team of 8 testers. I asked for the system design documents, and was handed the single piece of paper which was the marketing proposal. I knew right then, half an hour into the job, that this project was going to fail.

I cajoled and nagged the MT and bank project leaders for a couple of weeks, while my testers sat around with nothing to do, trying to at least get some high-level design documents and requirement specifications that we could test against.

After a couple of weeks, the MT project manager handed me a single page flowchart showing in reasonable detail what the system would look like. That was the sum total of the documentation. I looked at it, and 30 seconds later, said, "Where's the audit trail?"

He said, "We don't need no steenking audit trail, we're using SQL server."

I said, "This is a financial system. You need audit trails at each step so you can trace transactions from start to finish, and prove that no transaction gets lost, changed, or diverted."

He said, "We don't need no steenking audit trail, we're using SQL server."

Sigh.

After a couple more weeks, the MT team delivered the first version of the GUI, which runs on Windows PCs. We finally had something to test, and my testers were all fired up. We found a number of problems, which were fixed, and the cycle repeated.

MT delivered several versions of the GUI over the next few weeks, and my testers were happy.

The Tandem arrived at the bank's datacentre and was installed and running 24 hours a day, doing nothing.

The MT people working on the communications code, that allows the PCs and the Tandem to talk to each other, had major problems. The communications frequently failed and transactions were lost.

The Tandem support centre was in the UK so there were many phone calls. Eventually, the Tandem support guy flew to Australia to work first hand on this problem, which took a further month.

By this time, we were approaching the deadline for application delivery. The bank had rented a large office, installed 50 phone lines, and set up 50 work stations with PCs for the new call centre. They had also hired a professional trainer plus 20 people to work in the call centre. This all came out of the $10 million budget.

Since there was no software, there was nothing to train with, so all 21 people sat around for a couple of months waiting. Several of the call centre people quit out of boredom. The bank decided not to replace them until the software was delivered.

To make a long story shorter, the deadline arrived and was passed with no software delivered. the money ran out so the bank allocated another $5 million. Eventually, the MT project leader said, "We've got it working, I would like you to test the flow of transactions."

I said, "You mean validate the audit trail, which doesn't exist."

"Yes," he said, "we have no steenking audit trail and now I wish I had one."

I said, "You have SQL Server's logs, but they are inadequate."

"Correcto mundo," he replied.

So he and I sat down in a conference room with stacks of paper a couple of feet high trying to trace transactions. In two hours, we had validated 5 or 6 transactions out of the thousands.

He said, "Piss on this, that's enough. Let's go get a beer."

So that was the end of the audit.

Shortly after that the project was terminated, and everyone but the bank project manager had their contract cancelled.

For more information on Software Testing visit http://www.qa-software-testing.com