It may seem to be a given that an identity management solution would center around an individual. Still, without person-centricity at the forefront, solutions can fall into a transaction management approach. While transactional information is important, value is unlocked when the uniqueness of identity records can be established. This is most easily done through biometrics but can also be accomplished with biographic, document, and contextual information. The key is to establish uniqueness as early in the identity management process as possible.

We understand that identity resolution, data management, and downstream uses will rely on the context of the identity data just as much as the data itself. For example, knowing the name “John Smith” was obtained from a Canadian passport on April 1st at Ft Meade by a user, JaneDoe can mean much more than a name field simply containing “John Smith.” Without the contextual information at the time of collection, many operations are left to infer context or look for it elsewhere, typically in a very inefficient manual approach.

Matching (biometric or biographic) is critical to any identity management solution, but it does not need to be the core driver of the identity repository. With proper identifier utilization, data feeds can be sent to matching services that are already available and make the most sense for the need. Matching results can then feed the identity repository to establish links or uniqueness of records. This allows for multiple matching services (e.g., fingerprints by one service, facial photo by another, fuzzy biographic matching by yet another) to be utilized based on data availability. Whether you utilize an existing matching capability or not, our approach would utilize matching as a service.

Operations change; policies change; data needs change. For a solution to be adaptable, sustainable, and valuable to the broadest set of needs, configurability is critical. For example, if an operation interfaces with two “classes” of individuals – one requiring a high level and one with a low level of identity scrutiny and information verification needs – parameters must match score thresholds, expiration timelines, and several required data points should be configurable for each class. This configurability approach not only improves the application of resources where it’s needed most but improves the useful life of the solution.

While identities themselves are in most cases independent of your operations (the consumer of the identity information), the nuances of your use cases and collection points can make all the difference when making identity-based decisions. For example, determining whether Jen = Jennifer can be done more confidently if operational scenarios (e.g. a renewal of an existing credential, a change in benefit status) can justify it. This is where pure “AI/ML” data science can fail – but when combined with business-driven rules engines, the institutional knowledge that human experts have accumulated is combined with the scalability and speed of systems to form robust and accurate decisions.