Data Mining in Excel Part 12: What-If

Today, we'll be talking the next component in the Table Analysis Tool set, What-If.

What-If

In our previous post, we talked about the Goal Seek algorithm which would give you the optimal value needed to obtain a particular result.  The What-If algorithm accomplish the converse.  You tell it what you want to change, and it will determine the likely result of that change.  For instance, "Would this person's buying habits be affected if we gave them a 10% discount?"  Once again, our data set consists of attributes about people, which we cannot change.  Therefore, this analysis is purely theoretical and is designed to demonstrate the tool.  Let's get going.

What-If Window

The first section allows us to choose which value we are willing to change.  There are two options here, "To value" and "Percentage".  "To value" allows you to specify the exact value you would to change it to, while "Percertange" allows you to increase the value by a certain percentage.  Obviously, "Percentage" only applies to numeric variables, which is why we couldn't select it here.  The next section allows us to choose the variable we would like to see the effect on.  For instance, our question here becomes "Would these customers be more likely to buy bikes if they moved to the Pacific Region?"  Let's see.

Relocation to Pacific

The algorithm appends two columns onto our data set.  The first column tells us whether the customer would buy a bike if they relocated.  The second column tells us how confident the algorithm is with the result.  Higher confidence is a good thing.  Now, let's ask "How many children would I have if I made 20% more money?"

20% Higher Income

As you can see, this only takes a few clicks.  Most of the work goes into figuring out an insightful question to ask.  Now, let's combine this with the "Detect Categories" algorithm and look at the data in Tableau to see some relationships.

Dashboard

Within a couple of minutes, we were able to take a set of data and create some interesting visuals out of it that really tell us a story about our data and help us make better decisions.  Stay tuned for the next post when we'll be talking about the Prediction Calculator.  Thanks for reading.  We hope you found this informative.

Brad Llewellyn
Director, Consumer Sciences
Consumer Orbit
llewellyn.wb@gmail.com

http://www.linkedin.com/in/bradllewellyn