A Guide to BI Project Failure

Originally Posted January  4 2011,  3:53 PM


This story is not actually true but is inspired to a real world project. I’m changing many particulars to make the characters not recognizable. Anyway it’s still a story worth telling. 

Once upon a time there was a midsize company in a green valley in the northwestern part of Italy. It was a successful, consumer goods manufacturer. While they still were a personally owned company, with a very personal management style, yet they had an international reach. From the halls of Montezuma to the shores of Tripoli, they fought their commercial battles.

The company leader, once, woke up and realized he did not have a firm grasp on company’s operations and commercials anymore. So, he asked a large and famous consulting firm to come in and find a solution. He thought that he was going to need something on the computer side, but he could not define that properly because he started out when he was 15 and, at that time, pencil and paper were more than enough.

When the famous consulting firm analyzed the issue, they immediately realized what the problem was. Actually, the only honest reply would have been “learn to delegate and let your people do what you stubbornly do by yourself”, but this reply would have guaranteed a kick in the ***, so it was never voiced. The famous consulting firm, instead, as the quarter was closing, mounted up a spectacular demo and sold a famous reporting tool and an even more famous  budgeting tool.
“Can I do this? Can I do that? Can I do it twice?”, asked the leader.
“Yes,Yes,Yes” was the answer.

After a toast to celebrate the signatures on the contract, the project took off and the grunts came in. The grunts discovered that a reporting and a budgeting tool was sold, so they decided to start implementing the reporting part. It was not that difficult, and they built a nice, small, data-warehouse and placed the reporting tool on top of it. When the result was shown to the leader, something totally unexpected happened.
“Nice, but I already have the mainframe printouts. How can I change the list price for a customer and a product and see the result?”
“Actually you can’t, this is reporting.”
“I do not need reports, I need to change the price and see what happens”
“Well, you can do it on the budget app, but it’s phase 2”
“I do not need a F****** BUDGET! I need to CHANGE THE PRICE!”
“Sir, on the contract that you have signed …”

Then a heavy rainfall submerged the company, till the grunts, somehow, reverted to the Swiss Army Knife of professional data management, that is, Excel. They built something interactive, just to let him work the way he wanted and plug the hole in the hull.

The project, then, turned into a perilous death march that lead nowhere, and I spare you the rest of this sad story.

Now let’s see the good part, the lessons learned.

First: listen carefully to the customer. Do not think that a canned solution can fit all the needs just because the company is small or the requirements are not clearly stated. This is not true for BI only but is true for every consulting job.

Second: your solution is not necessarily something IT related. In this case the fundamental process flaw was the one man show. A more decentralized process, probably, would still have required some IT support, but that’s not the focus.

Third, and far more important for our purpose, the decision maker was not interested to reporting. The old printouts were supposed to give all the relevant information. What was needed was a tool to do simple calculations to easily assess the impact of a price variations, sales drops or surges, new freight contracts etc. The models implemented were very simple, probably too simple for  those who design business models for a living, but they could yield a fast and prompt reply. Sometimes, it is better having a moderately accurate model but simple and fast than a laser focused model but complex and unpractical.
This is a requirement I keep seeing popping up here and there. And yet I have to see someone addressing this requirement (with few, partial, exceptions).
Probably this appears to be a problem too specific to be solved by an app, or maybe Excel is perceived to be the perfect fit; at the end of the day, there’s a new class of issues to be addressed.

This topic is becoming paramount in these last weeks, so stay tuned!

  • Jan 23 2011,  5:28 PM

    Ellie K liked this post.