Yielding Results With a Yield Model

If you’re reading this article, you probably have some interest in analyzing customer data. After all, it’s the title of this particular ClickZ column. I started to assume that if you’re analyzing your customer data, it’s so you can use it to make better decisions. But is that really true?

Perhaps you’re so inundated with data, reports, and research that none of it is sinking in anymore. Perhaps you’re looking at results that took so long to compile you’ve lost the window of opportunity for making any meaningful changes.

I find it beneficial to periodically step back and review whether or not I’m using my data (and my time) wisely. Over the years, I’ve found myself always coming back to what I call the “yield model.”

The yield model is simply a documentation of the start-to-finish experience of getting an anonymous person in the door, through the store, and onto your list of customers. I used to take it for granted all business were closely monitoring their overall yield model. Sadly (and surprisingly), they’re not.

Remember the story of the six blind men and the elephant? Each felt a different part of the elephant, so each came away with a different impression of what they were experiencing. That’s how a lot of businesses are with their customer data. Each person is often looking at a small portion of the data and drawing conclusions without seeing the whole yield model. I think that’s dangerous.

Here’s an example of a yield model, in its most basic form, for a hypothetical online store:

  • 10 million people hear of the online store through a variety of means over the course of a month.

  • 5 percent (500,000 people) decide to visit the site, for various reasons.

  • 50 percent (250,000 people) venture beyond the home page and explore the site.

  • 10 percent (25,000 people) go on to make purchases.

  • The average number of items purchased is two, for a total of 50,000 items sold.

  • The average price per item purchased is $20, resulting in $1 million in revenue.

This is a very basic concept and a very basic example. Still, you’d be surprised how many businesses never look at the pieces all together. Whether you create an impressive PowerPoint presentation or draw it on a whiteboard, when you look at the big picture and begin to evaluate the possibilities for improving each result, you’ll often find gaping holes and new ideas. For example, why did only 10 percent of people go on to make a purchase? Couldn’t the rest find what they wanted? Did they think it was overpriced?

The yield model can easily be broken out to compare various customer groups, such as people who responded to an email or clicked a banner ad. Perhaps that comparison leads the company to ponder why customers who click banner ads spend less money than those who click an email link. The possibilities are many, but the practitioners are few.

I always go beyond a basic yield model to diagram activities happening behind the scenes. For example, the above-mentioned store might add a step to show what percent of people filled a shopping cart, then abandoned it. Perhaps a metric shows what percent of people referred to the site’s help section before completing their purchases. Obviously, this will be different for every company.

Over the years, I’ve seen some predictable reactions as I’ve shown each new employer or client a diagram of the yield model. Light bulbs going on. Individuals starting to understand how their actions (or divisions) fit into the big picture. Brainstorming and ideas popping out everywhere. Hidden problems coming to light. These reactions are not a result of the yield model itself. They’re the result of people finally understanding what happens between each step in the model, then applying that knowledge to their roles.

A word of advice before you pick up your pencils and start diagramming: It doesn’t have to be perfect to be effective. Don’t think because you don’t have access to 100 percent of the data or because your data isn’t 100 percent accurate, you can’t complete your own yield model. It’s perfectly fine to show a step with no statistics. This can also help convince others better tracking is necessary. In many cases, your inaccurate data may be consistently inaccurate and therefore still provide a meaningful measurement.

Don’t get me started on those unruly workers who are convinced they don’t need to understand the big picture to do their jobs better. I’ll save that for another column.

Related reading

site search hp