Take Attribution for a Test Drive Using a Secondary Profile


Attribution has been a popular buzzword in the industry for years now. It’s a topic at nearly every digital marketing conference as well as in most trade publications and blogs. Yet, for all the talk and general acknowledgement of its importance, shockingly few marketers really take advantage of it.

Sure, some marketers feed their data into an attribution system and get back pretty charts and reports. Others make modest media mix tweaks because they can see that some channels are helping others more than their analytics package gives it credit for. But many retain a healthy fear of running a complex, cross-channel, multi-exposure model that feeds into their optimization, modeling and analytics.

Well, as FDR said, “the only thing we have to fear is fear itself.” In 1933, President Roosevelt was talking about citizens and businesses fear of investing in a very shaky U.S. economy. Those words helped pave the way for a boldness which reinvigorated the country, leading to prosperity.

Now, many marketers are afraid to change the way they do things, to move away from tried and true last click models, even though they know they’re faulty.

For many marketers faced with an all or nothing choice, they choose the relative safety and consistency of their existing model. After all, their budgets and goals were likely set up based on it. But why not dip their toe in the water to get a feel for what could be without immediately breaking away from what’s known?

If you could take attribution for a test drive, would you?

If you answered yes (and why wouldn’t you), the good news is it doesn’t have to be an all or nothing proposition. Before you move away from your current model, test out some changes using a secondary profile to allow you to experiment with the benefits and glean some important insights. Let’s look at some examples of areas to test.

Partial Credit

There are several levers you can pull. Most obvious is to start giving partial credit across multiple exposures. But for many, knowing how much to give remains a mystery. By setting up a simple even-credit model across a number of exposures, you can start to see how channels are working together without changing the way you optimize off the bat.

Exposure Sequence/Latency

Another tactic is to prioritize channels differently regardless of exposure sequence or latency. By moving acquisition-driving channels higher than, say, email, in a secondary profile, you can get much more insight into how search or display brought in new users, moving your focus higher up in the funnel as opposed to just saying “email closed the deal.”


You can also change look-back windows for each channel within a secondary profile. This allows you to see if you should start giving more time for certain channels to convert while potentially excluding obviously unrelated exposures such as a display view that occurred three weeks before the sale.


Another lever is weighting. If signups from one channel tend to lead to higher value customers, you may want to give more value to those sign ups to see if you’re able bring in more valuable customers. Similarly, if customers from email specials are more likely to return an item, you can lower the weight for email to see if that improves your gross margin.

The First Step is Testing

There is room for improvement on your media optimization if you are not leveraging attribution or are still stuck on a last-click model. Guaranteed.

You owe it to yourself to test out advanced attribution models and see how they can help you better understand your media and your customers. By testing a secondary attribution profile, marketers can have the best of both worlds – putting attribution through its paces and getting comfortable with it before trading in the old model.

Related reading

Your new secret weapon for better and robust analytics
A summary of Google Data Studio Updates from April 2019
What defines high-quality links in 2019 and how to get them
On-site analytics tactics to adopt now Heatmaps, intent analysis, and more