Building a Yield Model, Part 2

Part one of this series discussed what a yield model is and why you should have one. That mastered, you’re ready to determine response rate, your first yield. How many new people visited your site as a result of your marketing efforts?

Think of this model as pushing large numbers of people through a funnel in the hope they’ll come out the other end.

When crafting a yield model for a specific campaign, your goal may be to compare campaigns to see which are most effective. If you’re creating a yield model to summarize all your marketing activities for a certain period, your goal may be to compare periods to see if your marketing programs are more or less effective than in the past.

Whichever the case, knowing your response rate is a basic step for all good marketers. I’ve read a lot of articles slamming “old school” marketers for spending money with no regard to how campaigns perform. True, those people exist. But many offline marketers have been measuring campaign yield for years, using coupons, promotion codes, surveys, comparison testing, and so on. Learn from them! Some are today’s most successful Internet marketers, because they applied what they learned offline to a new medium.

Businesses have been asking “How did you hear about us?” far longer than I’ve been around. The Web takes that question to a new level. Some companies go to extreme lengths to determine where their traffic comes from.

In the early days of online advertising, it wasn’t unusual to purchase a banner ad from a site that could not report the number of impressions delivered or click-through rates. Now that better methods are available, sometimes we forget there are ways we can enhance the tracking we receive from outside sources.

A company I dealt with long ago, before the days of the infamous pop-up ad, served a “How did you hear about us?” survey to every person who clicked beyond the home page. The survey response (clicked a banner ad, email link, etc.) was embedded in the link and stored for every visitor — whether they clicked beyond the home page or not. If a user went on to register, the code was transferred to her customer profile, making it possible to attribute sales and other behavior back to the marketing effort that landed that customer. An analyst’s dream, having that much good data on hand.

I don’t recommend going to such extreme measures today. Web users won’t answer surveys like that any more. The story is an object lesson: Sometimes we have to take matters into our own hands rather than throwing in the towel because the data we want isn’t easily available to us yet.

A more recent example is I have no direct knowledge about its marketing practices or what it does with its customer data, so I’m guessing here. I get a postcard from every two or three months, as well as some emails I don’t read. Every postcard directs me to a different “promotional” URL. If I use that URL, I get free shipping. I assume the purpose of the special URL is to control the number of people who get free shipping. But hopefully, its marketing people have arranged to track which customers use that URL (subject to the usual privacy concerns) to evaluate the success of each campaign. If it’s able to compare who it sent the postcard or email to versus who responded, over time it’ll be able to segment customers based on the type of offers they do or don’t respond to. The possibilities are endless.

If you don’t have a method for determining where your site visitors come from, don’t despair. Hit the Web and read up on what other companies do and what products are available. (Hint: ClickZ’s archives are a great place to start!) Talk to your tech folks. Show them postcards or email you’ve received from other companies. Get their input on how you might track. You may have to use a variety of methods depending on your type of marketing initiative, but that’s OK.

For those of you already measuring response rates on a regular basis: Why do you need a yield model? Isn’t response rate enough? Why take the calculations any further? Because seeing your yield in its entirety helps generate ideas and understanding about how everything works together. Think your job is to generate traffic? Think again. Nowadays, we’re all responsible for the bottom line.

John may be responsible for driving traffic to the site, while Jane focuses on increasing order size. John cancels Banner A because the click-through rate is so much lower than Banner B’s. Jane’s average order size decreases. She has no idea why. A quick comparison of the yield models for the two banners reveals that although Banner A had a lower click-through rate than Banner B, Banner B customers spent much more than Banner A customers. John’s activities had a direct impact on Jane’s, yet neither had a clue.

A full yield model takes comparison testing of marketing initiatives to a new level. No longer are you testing just the response rate to two different banners or click-through rates of two different email subject headers. Now, you’re comparing the revenue potential of each, evaluating which steps in the conversion process lead to the difference in results and learning how to improve them.

Next, we move on to step three: Determining how many visitors interacted in some way relevant to your business. This is what happens between customers arriving at your site and generating revenue. For some, the desired interaction will be registration. For others, putting an item in a shopping cart. Measuring this step is crucial. It can help you figure out where you lose potential customers. Until next time, put on your thinking caps and work on step two.

Related reading

site search hp