BeyeBLOGS | BeyeBLOGS Home | Get Your Own Blog

« November 2009 | Main | February 2010 »

January 5, 2010

Decision Trees in Project Portfolio Management

PMBOK recommends using decision trees in project portfolio management decision-making analysis, focues on areas such as risk management and project go/no-go decisions.

Here is an example of a simple decision tree by Tom Mochal @ TechRepublic where risk implications are mapped out with a decision tree based on expected monetary value:

Another common use of decision trees is making decisions on projects in portfolios that will take into account a number of different attributes including ROI, costs, resources, risk score, strategic alignment, etc.

For simple modeling, drawing a decision tree in a tool such as PowerPoint, Word or Visio works just fine. I find it very useful to decompose a problem and to gain visible, graphic insights to be used for analyzing a problem before making a decision.

However, for more complex modeling, you may require a tool to consolidate different data sources and to build models based on complex rules & algorithms.

Now, I don't wish to offer my personal recommendation on which vendor to utilize to feed attributes and criteria to derive, model and analyze decision trees from your project portfolios. Instead, what I'd like to do is give you links to look into 3 data mining vendors that I've used for this purpose in the past:

Microsoft, in my experience and opinion, is one the simplest and easiest to use in terms of data mining algorithms such as decision trees. The trade-off of simplicity vs. complexity is that you may run into limitations with SQL Server and Excel based data mining. But if your intent is to use just a simple decision tree algorithm and do not require customization or extensive algorithms, then Microsoft is a good fit.

Oracle has both a native data mining engine in the database as well as prebuilt analytics in their enterpise performance management tool, Hyperion. Use the data mining engine if you wanna open the hood and get greasy or use Hyperion if you wanna take advantage of prebuilt attributes and algorithms. The problem with Hyperion is that is does not yet have a solid prebuilt set of project portfolio solutions, so you would probably have to use the financial analysis applications to fit your PPM data into that model.

SAS is generally considered the big daddy of data mining and predictive analytics. My experience is that total cost of ownership increases with a SAS based solution because of the licensing & support costs as well as the skill & talent levels needed to build models in SAS for activities such as decision trees. But the benefit is that you have the best-in-class data mining tool set and limitless possibilities.

Posted by Mark Kromer at 10:15 AM | Comments (5)