Fool’s gold – Leaving behind “match rate” as a success metric for identity

The topic of identity resolution is gaining increased attention among marketers. Brands are coming around to the notion that personalization in marketing and advertising is only as good as their ability to know who they are talking to (Read more...) time.

Marketers are growing more comfortable with utilizing first-party data in combination with ID resolution or “ID graph” suppliers to fuel their people-based ad campaigns, site personalization, and analytics.

Exploring current limitations of ID graphs

The identity market is set to increase 188 percent in the US alone from $900 million in 2018 to $2.6 billion by 2022. The increased focus on identity is promising, however, today’s predominant ID graphs are “public” – largely external to the enterprise and often compared to a utility grid.

Today’s ID graphs are presented to marketers with a very tidy proposition: input your first-party data and we will match it to a person who can be reached through addressable platforms and channels i.e. Facebook, or programmatic display media.

The quality and effectiveness of the ID graph is presented to brands through an easy-to-understand metric known as “match rate.”

It’s pretty simple; a brand sends in a file of known individuals and the ID graph sends back a percentage rate at which it can link those consumer IDs (email addresses for the most part) to a cookie or a device. The bigger the match rate the better, right? Not so fast.

If the marketer could crack open the ID graph black box and look at what’s underneath that match rate, the discussion quickly gets more complicated and troubling.

What is perceived as a homogeneous set of email-to-cookie or device ID pairs in ID graphs, neatly arrayed one-to-one, is hardly that. Yes, there would be some of that, but marketers would also find matches built using math that counts every combination of person and device in a household or building.

They would find linkages between emails and cookies that are as old as 90-120 days counted the same as those 15 days old or less. Even more disconcerting, the marketer would identify a significant number of these linkages that make no sense at all, and in some cases represent botnets.

How else would you explain a single cookie associated with 1000+ people?

With this visibility, the marketer would quickly realize that the idea of a simplistic match rate metric is fool’s gold.

A metric based on quality of identity vs quantity

The black box associated with ID graphs combined with increasing privacy compliance pressures and insufficient depth of available first-party data as Internet browsers deprecate use of third-party cookies pose some real questions for the long-term sustainability of the ID graph in its current form.

The future of identity will be about quality; marketers cracking open the black box of ID graph providers and taking ownership and control of identity.

Marketers must demand a better metric – a confidence or quality score – to confirm and validate the identity solutions they are paying for, and more importantly, to enable the kinds or use-cases and results marketers expect from their identity investments.

A confidence level or quality score on identity is the only way marketers can truly know and trust what makes up the identity of their customers and prospects.

Imagine a match-rate driven, homogenous ID graph being used to deliver personalized offers on a customer’s birthday. The predominant ID graphs of today give no transparency into wider person-based identifiers that are matched, or method to score and use a group of ID’s and data based on higher confidence/accuracy.

In a world where match rate is replaced with a metric such as person match confidence level, marketers can use level 1 ID’s to deliver highly personalized messaging with limited risk of getting the customer experience wrong.

Display media prospect targeting can loosen the accuracy aperture to a lower level like a 4-5 to maximize targeting scale, while analytics staff performing attribution modeling can use levels 1-3 for understanding scaled reach across a journey with the flexibility to tune attribution models based on use of confidence level 1 IDs; removal of duplicate reach and maximum accuracy of person exposure to closed-loop actions like sales.

Analytics teams can also build audience segments and targeting models off the most accurate ID’s and data of customers and a prospect universe to drive best performance.

Marketers and publishers should understand if there’s a common data model and infrastructure among identity providers that allows all sources of linkages to be resolved back to a known, identified person, and in a privacy-safe way abiding by all regulations.

In this regard, the whole population, gathered from first-, second-, and third-party sources, becomes the marketer’s new “CRM,” or ID graph.

And similar to traditional CRM databases, data in this “private ID graph” is constantly scrubbed of noisy signals and scored based on quality metrics like person match confidence.

Marketers can then take complete control of how they monetize their private ID graphs based on use cases and value, and rest easier knowing they’ve discovered real gold, a durable and sustainable basis of identity leaving metrics like match rate in the dust.

As Chief Strategy Officer of Merkury, Merkle’s identity resolution and data platform, Gerry Bavaro leads sales and marketing, solutions, and business operations efforts aimed at integrating first-party identity and data products and solutions into Merkle and Dentsu agencies’ various services. He joined Merkle in 2015 and has held leadership positions such as SVP, Enterprise Solutions, SVP, Digital Strategy, and Global Chief Strategy Officer, M1.