Story Time: AI Probabilistic & Deterministic Targeting

Stories of industry, for your enjoyment and intellectual curiosity

When we first made the decision to commercialize our first-party data properties, two labels seemed to follow us everywhere: Probabilistic and Deterministic.

I remember hearing them for the first time on a call packed with very smart, very young people. As far as I could tell, our VP of Digital Audience Solutions and I were the only two on the call in the “50-plus club” — and the only two with 25+ years of data experience including:

Standardizing, normalizing, matching, and appending large disparate data sets to create successful postal mail campaigns. If you get that wrong, the mailer pays the cost of each mail-piece and the postage, while watching revenue tank. Good way to lose a mailer. On the flip-side, get it right and the orders (and mailer continuations) roll-in.

Building and managing email data, good corporate citizenship, and best practices necessary to deploy successful email campaigns. When done right - email is still the perennial marketing ROI winner.

Watching social-media arrive overnight, only to over-promise and under-deliver.

Developing and deploying 1st-party digital audience targeting a decade before it was mainstream.

On the call I asked for a description of what the two labels meant. My first reaction: Probabilistic sounded like a recipe for problematic; and Deterministic - detrimental.

Over the years that first reaction proved fateful.

Audience “expansions” with little foundation in 1st-party data sources ensued. Scale for the sake of scale abounded. Attention to source data accuracy, freshness, and processes became the exception, not the rule. Advertisers footed the bill.

Today, we're told AI is going to absolve us from problematic processes and detrimental sins of the past.

Perhaps that’s true.

But not without feeding it a healthy diet of 1st-party data.