October 8, 2021
The Ideal QA Process in Digital Advertising
Google announced in May 2015 that mobile searches surpassed desktop searches for the first time ever. As the path to purchase stretches across more and more devices, it becomes increasingly difficult to monitor and evaluate each step in the purchase process. However, there are several methods you can employ to make cross-device evaluation easier.
Let’s look at a real-life example of a situation where we need cross-device tracking to see the full value of our marketing efforts.
I’m a night owl, but my young children don’t exactly follow this schedule, so I’m always looking for solutions to help me wake up earlier and be more alert. Of course, this thought only occurs to me as I’m getting into bed around midnight, when the only device handy is my phone. So I do a few searches and find a product that I like – an alarm clock that shines white light to wake me up.
But I don’t purchase on my phone. It’s hard to fill out purchase forms on the phone, and my wallet is across the room. I can’t muster the courage to brave the cold outside my bed, so I decide to complete the purchase tomorrow. When I get to work, I open my laptop and do another search to do a quick price comparison, deciding to purchase from Sonny’s Wake-Up Emporium. But then something comes up, and I don’t complete the purchase until that evening, when I get home, open my personal laptop, and search specifically for Sonny’s.
I’ve now searched on three devices, and visited Sonny’s site each time. These visits would look like 3 separate site visitors with traditional tracking models, so the first two interactions look like wasted effort (and spend). In reality, each visit was critical towards my ultimate purchase. The ability to recognize the value of my first two interactions is critical to an advertiser who wants to be in the right place to offer a solution to my problem (i.e. being awakened by my child jumping on me.)
There are two types of walled gardens we come up against as we try to measure the effectiveness of marketing efforts. One is based on devices (specifically browsers) and the other is based on a website or product.
Browser-Based
Cookies exist only on a single browser, so as long as someone stays on the same device, their website visits are recorded as coming from the same user. However, once they move to a different device (and more than 90% of searchers do) we lose that visibility, and see visits on a second device as visits from a completely new user.
Product-Based
The second type of walled garden exists in a specific product, like Facebook. As long as you are signed in, and you must be signed in to use the site or mobile app, then Facebook always knows that you are you. But once you sign out, Facebook doesn’t necessarily see you as the same person.
There are a number of strategies available to connect the dots across this divide, and plenty of companies offering solutions to this challenge. They generally fall into two camps: deterministic and probabilistic.
Deterministic Approach
This is most similar to Facebook’s tracking method, where users are tracked based on their login information. Google has introduced a similar functionality with the Universal Analytics upgrade in Google Analytics. Advertisers can provide a unique GA User ID for signed-in users, and then record a user’s movements between devices. This approach faces challenges when users are not signed in, meaning that the deterministic approach is most valuable when used on platforms with a wide reach, such as Google or Facebook.
Probabilistic Approach
The probabilistic approach looks at various signals to identify users regardless of device. Vendors providing this approach monitor browsing behavior, connections to WiFi nodes, location and more, and claim match rates above 95%.
Indirect Approach
The indirect approach is by far the least sophisticated, as it requires no technical changes to employ, but it can allow you to measure cross-device performance without investing in new software. The idea here is to test and measure for uplift or growth.
For example, if I increase my investment in mobile by 20%, what overall impact does that have on my performance?
You will not have the ability to track user behavior across devices, but you can approximate the effectiveness of investment in new areas by looking at overall growth. Just make sure to look at a large enough range of time to draw a statistically significant conclusion, and to control for other factors that might affect your metrics.
The goal of any of these approaches is to demonstrate the value of marketing efforts that cannot be tracked by pixels and cookies. Our ultimate goal as marketers is to speak directly to an individual, not a device, and offer him or her a valuable solution to an existing problem. The focus on individuals, not devices, should be a reminder that 100% accurate tracking is not possible, and that there are limits on how far we can and should go to track an individual’s each and every action.