Bridging the Identity Gap

Should you use a probabilistic or deterministic data matching strategy? Or both? Here's how to decide.
Customer Experience

Mobile devices are at a tipping point. Consumers are more connected than ever and use multiple devices for a wide range of activities from checking email and reading the news to shopping and paying bills. In 2014, 91 percent of consumers owned a laptop or PC, 80 percent owned a smartphone, and 47 percent owned a tablet, according to a survey of about 170,000 adults conducted by market research firm GlobalWebIndex.

As the mobile landscape becomes more fragmented, it's critical for marketers and advertisers to provide consumers with a consistent experience across all these devices. Marketers have a lot of first-party data assets including customer data tied to emails, physical addresses, and cookies but they are often disconnected.

Enter cross-device targeting. A growing number of vendors tout cross-device analytics solutions that are designed to extend sequential messaging across multiple screens based on on a user's particular journey and other features. But it's still early days and brands face a number of obstacles before they can connect their data assets.

The Cookie Crumbles
For several years, cookies or text files containing user information have supported most of the advertising industry's online targeting efforts. Publishers and advertisers largely depend on Web-based cookies to serve targeted ads to Internet users. And while cookies do exist on mobile web browsers, they can be reset every time users close the browser and they can't be shared between apps, making them unreliable. As a result, two other tracking methods, probabilistic and deterministic data matching, are gaining traction as alternatives to traditional cookies.

The deterministic method relies on personally identifiable information (PII) from a login system to make device matches. For example, when someone uses the same email address to log into a website and an app, advertisers and publishers can use this unique identifier to target those users on multiple screens with a high degree of precision.

This tactic is most effective when companies have a huge user base, like Google and Facebook, which have a number of mobile and desktop properties that require logins. At the same time, even companies with smaller user bases are increasingly asking users to sign in through a social login or by creating an account.

Probabilistic cross-device matching occurs by algorithmically analyzing thousands of different anonymous data points such as IP addresses, device type, location data, time of day, and a host of data points to create statistical matches between devices. For instance, if a smartphone, laptop, and tablet connect to the same Wi-Fi hotspot in the same coffee shop every weekday, it's likely that all three devices belong to the same person.

And because probabilistic device matching doesn't rely on PIIs like an email address, the data is considered more private. This tactic can also be scaled across more platforms and applications beyond those connected by deterministic methods. The tradeoff is that probabilistic tactics are not an exact science and accuracy rates vary.

Although deterministic matching appears to be the more reliable solution, users don't always stay logged in or use the same email address. And although Facebook has an impressively high percentage of mobile-only users, its overall desktop usage has been decreasing, which might diminish its ability to create cross-device matches if future users rely more heavily on its mobile properties.

So which tactic-deterministic or probabilistic-is more effective? The answer depends on the marketer or advertiser's goal, notes Jennifer Lum, co-founder of Adelphic Mobile, a cross-channel programmatic ad platform. "If you want to reach a specific group of people like existing customers who you already know are interested in certain products, then a deterministic solution is your best bet," Lum says. "But if you want to reach a broader group with a less specific offer, a probabilistic tactic that can give you a lot of reach makes sense."

The match or accuracy rate isn't important, agrees Nicole Kerr, director of product management at Monetate, a personalized marketing technology provider. "What's important is understanding how many uniques are visiting your site, where at least one other cookie or screen can be associated to that individual and if the brand is comfortable with the confidence score," Kerr maintains.

One cross-device vendor may have a 91 percent average accuracy rate and another offers 95.6 percent and "no one knows which one is actually better," Kerr adds. "They have different processes and unique scoring methodologies. What actually matters, in our opinion, is whether the data accuracy assists the use-case. For example, high-level personalization could benefit from greater reach but lower confidence, whereas recommendations may require higher confidence."

Keeping an Eye on Privacy
Deterministic matching may provide highly accurate cross-device tracking results, but it includes privacy concerns and questions about the security of the data. It's up to clients to decide to what extent they want to track customers, adds Arj?ahn, chief technology officer and co-founder, at Hippo, a content performance platform.

"If you only want to use data that's available in one session, then you don't have to store the data in a database and you don't have to worry so much about privacy. But if you want to give a richer experience, then you have to track multiple visits and you can be held responsible for the data," Cahn says. "We advise our clients to be as transparent as possible to their customers and to focus on making sure they're providing an excellent user experience-not just collecting as much data as possible."

Hiding first-party data with anonymous hashed identifiers and providing opt-out options are common practices for cross-device matching purposes but these practices are inconsistent, critics argue.

For instance, Verizon and Turn, a digital ad company, were hit with a scandal in January when Stanford University researcher Jonathan Mayer noted that Turn appeared to be reviving opted-out cookies via Verizon's unique identifier header.Facing pressurefrom the privacy advocates, Capitol Hill, and negative press around what the media referred to as Verizon's "zombie cookie," Verizon ultimately said it would allow users to completely opt out of all tracking if they choose to.

"Companies should be up front about what they do, and realize that not everyone wants a hyper targeted, personalized experience-there's also value in serendipitous discoveries," maintains Justin Brookman, director of the Center for Democracy and Technology. Brookman also dismissed the idea that consumers are not bothered by data collection practices as long as they're receiving valuable content or offers. "People are uncomfortable with companies having information about them that they didn't consciously share and that's why we're seeing people install ad blockers and uninstall apps," he says.

According to a survey of about 2,250 adults, 57 percent of app users have either uninstalled an app over concerns about having to share their personal information, or declined to install an app in the first place for similar reasons, reports the Pew Research Center.

The Future of Data
As brands seek to streamline and optimize their messaging across devices, data ownership will become a critical factor. Advertisers and marketers are also likely to challenge the "walled garden" practice where players with large logged-in user bases like Facebook and Google can control prices because they have the most scale.

"Facebook's strategy is incredibly smart, it offers great targeting capabilities and scale, but while these capabilities allow publishers to make money in the short-term, ultimately Facebook has all the control," notes Chris Cunningham, head of mobile at IronSource, an app discovery platform. "What if publishers themselves could extend their targeting capabilities?" A few companies like AdBrain, Appnexus, and Apsalar already offer their own unique IDs and analytics for targeting ads and we can expect to see more competitors, Cunningham notes.

The concept of cross-device identification is also rapidly expanding. While it currently means linking PCs with tablets and smartphones, the rise of connected TVs, wearables, and the Internet of Things will provide marketers with more opportunities-and challenges-for connecting with consumers.

But regardless of the number of devices that are being leveraged, being transparent about your data collection practices and allowing customers to control the level of personalized communications should be part of every company's customer strategy. And just because you can provide a high degree of personalized messaging, doesn't mean you should, notes Monetate's Kerr.

For instance, if a new visitor shows interest in sundresses over swimsuits, curate the site content accordingly, but avoid cross-device messaging that's based on what's in the customer's cart. "Those are powerful techniques, but best reserved for deterministic identity where the users have opted in, they've made a choice and understand this is utility and not 'dark-marketing,'" Kerr says.