Disadvantages of using deterministic techniques for cross device identity matching

People based marketing techniques can be broadly classified into two categories-probabilistic and deterministic. In the former, complex algorithms are used to make a guess about which visitor profiles (regardless of the device or interaction channel used) might belong to the same physical person. In deterministic techniques, on the other hand, cross-device/channel profiles are allocated to the same underlying physical user based on concrete mapping of identities using some common link field such as customer id, email, phone or some other personally identifiable information.

Both techniques have their pros and cons when it comes to one-to-one marketing. Running personalized, cross-channel campaigns that target individual consumers across multiple devices requires that data points from across multiple touchpoints be consolidated under a single identity profile. Achieving this using deterministic techniques has some key challenges as discussed below-

Data accuracy-While identity matching accuracy is the single biggest driver for adoption of deterministic techniques, in reality, the accuracy is rarely 100%. Consider for example a user who uses an anonymous login from a pc at work to browse through items in an e-commerce store. The same user then visits the same store from a different device at home and this time as a logged in user. Using a deterministic technique, there is no way to merge the browsing data of this user from across the two devices. Or consider the example where the user opens a different browser on the same work pc for further product research. Again, the data from sessions from across browser types cannot be allocated to the same user using a deterministic identity matching model.

Privacy-Using deterministic techniques requires that users provide personal information such as email, mobile number or some other unique ID that uniquely identifies their account. This may never be the case for certain categories of sites such as media publishers where users are not normally required to create accounts and forcing them to do so would inevitably degrade the user experience.

Lack of scale-Deterministic techniques only work on devices where the user has logged in at least once. So for example, if a user has never logged in to a site on a tablet, it may not be possible to show personalized content or offers to her if the targeting is based on deterministic techniques. Secondly, these techniques rely on explicit data points generated by visitors through submission of personal information. This implies that targeting based on deterministic techniques cannot consider any behavioral trail generated across third party sites not owned by brands. Without this larger pool of information, the options for targeting can be limited and may not justify the investments in cross-device identity matching.

Lack of deeper audience understanding-Identity matching using PII data is simply linking multiple profiles that share some common link field (E.g. email, phone number etc.). While this project the behavior of consumers across devices, it fails to provide any deeper understanding of why users adopt certain usage patterns. It is rare for consumers to provide detailed personal information around demographics, profession, social inclinations etc. as part of regular profile registrations and using pure play deterministic techniques, it is not possible to make inferences about these. The quality and efficiency of messaging inevitably suffer without a deeper knowledge of drivers for consumer behavior.

Conclusion

The long-term utility of designing customer contact strategies based on simply match rates in identity matching is questionable. Instead, Marketers must strive to understand the underlying behavioral drivers for values of the various data points that are deemed relevant for marketing. That being said, the practical need for balancing accuracy and precision that can be generated through deterministic identity matching with the scale and reach of probabilistic methods would imply that the use of pure-play deterministic methods is unlikely to decline anytime soon.

A teaser article designed to provide preliminary insights into the 2 main techniques for tracking cross-device visitors as part of unified profiles. Meant for ‘top-of-the-funnel’ consumption, more as syndicated content appearing on third-party sites.

A Retail Analytics consultancy targeting Enterprise Clients with a single view of customer proposition for enhanced digital intelligence.