Hello. My name is Chris Reyes, and I'm the content manager for Toad BI Suite and Toad World for Dell. Today, we're going to discuss two simple and effective ways to use scatter plot when making decisions.
To do that, we'll be using sales data on several products over multiple quarters, shown here. I'm looking to find which two or three products are best for me to apply some additional resources to increase profit. I look at my typical aggregate [INAUDIBLE] graph and see that it's not going to give me much of an idea.
So I sort by product, and that just hurts my eyes.
I try area graphs.
And I try bar graphs. But unless I'm looking for new argyle sock designs, I'm out of luck.
When you create a scatter plot, you're looking for relationships or patterns among the different variables. To do that, I've combined my sales data with my product catalog, so that I can compare fields across both databases.
In addition, I created new calculated fields that were not available in those original databases, such as profit and sales cost. It's important that the variables or fields you choose to interrogate in the scatter plot are quantifiable.
Also, come in looking for goals, rather than a random pick of two variables trying to find some sort of match. In my original question, which was to find profit, I identified three variables-- products, units sold, and profit. I specifically wanted to increase profit.
As I create my scatter plot, I find that products are not a quantifiable variable. They're a string. You can see here that the scatter plot prevented me from using it on the plot.
It doesn't matter where I put each variable on the scatter plot. But for consistency, I prefer to put my monetary variables, such as profit, on the y-axis, and my non-monetary variables, such as units sold, on the x-axis.
Now that we've created the plot, let's identify those two methods of evaluation that will allow us to come to a decision. The first is called clustering, and its groupings of data. Normally, it's everything in a certain area. Or it can be something showing the same pattern.
When they share something like that, they will also normally share a common attribute, or attributes, besides just profit or units sold, for example.
Another visualization that scatter plots can provide you is non-linearity of results. As you can see here, there's a definite curvature in the data. And this curvature will allow us to determine the sweet spot where you're able to gain the most profit for the fewest units sold. In short, clustering allows you to group data into relevant subsets, and to decide on those subsets for further exploration.
The second method of evaluation is identifying outliers. Outliers are groupings possessing characteristics beneficially or negatively outside the norm. Whether you want faster sale cycles, higher profit margins per sale, or fewer customer complaints, focusing on clustered outliers identify what you want to reproduce or prevent.
As a rule of thumb, run your business or division based on norms. But improve your business by concentrating on outliers.
Now, because these outliers are clustered and not one-offs, you can use filtering to identify the something in common. You can fix-- or, something common, you can adopt that will make your organization more profitable.
In addition, use filtering to zero in on just that subset of data. In our example, I see my subset exists where profits are greater than $120, and the units sold are between 75 and 150 units.
This much smaller subset will then help to reveal what attributes these points share in common. In our example, it's the product name. And it allows us to decide that the 100-milliliter Detafast Stain Remover and the 1 liter Super Soft are the products of choice.
We hope this video helped you use scatter plots more effectively in decision making. We value your comments and suggestions on these and other subjects here on Toad World. If you'd like to try these techniques on your own data or our sample data, download a trial copy of Toad Decision Point. Thank you for your time.