Jan 20, 2014
How to Fail an Analytical Project
Let me first clarify on what defines as a success in an analytical project. I like to stick to the basics and I believe a successful project is one that is finished on time, meets the objectives, makes money, results in a satisfied customer and brings in more business. It is probably the classic of definitions and it comes to remind analytical projects are no different than any other. Analytical projects differ from other projects in few things but mainly in the fact that their goal is generating knowledge. This sometimes could be intangible and difficult to measure, particularly in case of descriptive and predictive analytics.
The ingredients in the recipe for a failure in my book include following topics:
No proper understanding of the business problem at hand
We may think we have seen it all and there are no business holds any secrets for us. And we are wrong. Even if the most common of businesses has its intricate details and specific context. Jumping to conclusions about the nature of the problem and how to solve it may play a a bad joke on us. The more time spent with the potential or existing customer the better the understanding what really the problem is. The analytical professional should never forget the customer is contacting them because of a business problem that needs to be solved. This problem has another side - understanding the problem better defines better and more appropriate solution. I would put here also the lack of proper understanding the lingo of the company that may result in a finding that no one in the company cannot understand.
Lack of communication with the customer during project execution
Analytics requires even greater number of assumptions many of which become required in the course of the project. Also, every problem could be solved in many different ways. For good or bad many of the analytical concepts and methods are widely unknown among the general public. A general mistake analytical professionals could make is to keep radio silence with customers in the course of the project and serve their findings at the deadline with tons of explanations leaving the recipients overwhelmed with information, terminology, charts and recommendations that they cannot relate to. This way exposes the project to the great risks of wrong assumptions made somewhere on the way or delivering outcomes and models customer could not understand or properly use. A remedy for this situation is keeping the customer in the loop - asking for opinions on assumptions, communicating intermediate results, discussing incomplete or mock models, sharing views on what is more appropriate, etc. Communication has to be in a proper language as well - we need to remember few people know that p-value is and what the chi-square statistics is all about.
Bad choice of methods or too much focus on them
A quote from a favorite movie says "I am not here to solve friggin' math problems". Every analytical professional has a bag of methods they are very confident with and naturally tend to apply in all the problems they face. Some are very keen on trying new and newer algorithms and software. Others really love puzzles and take a great pride of solving difficult math problems. All this is great and beneficial of course. But to a degree! These tendencies could move the focus of the project, obscure the better solution, increase project's costs and duration and bring in doubts in the outcomes. The analyst could be very proud with the way she completed the projects and could brag about how advanced and cutting-edge it is but the truth is the customer is not so impressed with that and have other criteria on their mind. I am all in for constant improvement of methods and research on the new ways but I sign up for using the most appropriate and solid tools on hand.
Wrong communication of outcomes
As I mentioned earlier - customer do not always understand the "analytical language". Using too much terminology, lengthy bragging on the methods, lack of focus on the business implications, lack of estimates in resources and check against other sources, etc are among the mistakes that could be made in the course of delivery the outcomes. Forget that recipients are people who have a problem to solve and decisions to make is probably the worst mistake. Sometimes analysts think that using too much of terminology make them look better professionals. And they are wrong. Talking the language of the customer is a great asset that pays back big time.
Delivering trivial and uninteresting results
No customer pays for something they already know. However, lack of understanding the business and the problem at hand could lead to findings are very well-known to the customer but seem very interesting to the analyst. Some results could be quite amusing and seem as a break-through for the analyst but have little or no value at all for the customer.
This short list of issues just touches the surface of the topic and I would welcome any opinion.