Solving Marketing’s Black Box Problem: Predictability with Explainability
Updated: Mar 5
Modern marketing has a major black box problem when it comes to the kudzu-like spread of AI across the industry. While marketing AI is uniquely able to make quality predictions from enormous amounts of opaquely correlated customer behavioral data, in most cases these results are delivered via unsupervised learning algorithms and therefore lack human accessible explanation. This lack of explainability is problematic for multiple reasons, but perhaps most importantly, because it prevents humans from being able to build upon AI-driven predictions with common sense insights that could further improve upon results
Recall your high school math classroom (if you care to), particularly the exams. A boxed correct answer in the absence of supporting ‘work’ (explanation) always prompted a ‘show your work’ comment in red pen. The correct answer is, of course, an excellent place to start, however full mastery of the building blocks is critical in order to work through next level challenges yet to come. Bridging this math test analogy back to marketing AI, when predicted results have no grounding in a human understandable ‘why’, we can’t learn from the process...or put such learnings to work on next level optimization.
Put succinctly by the Marketing AI Institute, “Although machine learning algorithms are highly precise, if they can’t tell you why someone is likely to respond, purchase, be a high-value customer, cost money or churn, then it’s impossible to have insights that are contextually relevant.” Marketing is an industry grounded in the why, and without it, marketers are working with one arm tied behind their back.
Pinpoint has a solution for this explainability conundrum (at least for the AI use cases we support!)
All of our analyses are grounded in the intuitive psychological understanding that humans have distinct, measurable personalities, which underlie most decision making and behavior. Most of this we understand innately; for example, we know extroverts are more likely to decide to spend time in group settings and cynics respond well to messages appealing to their deeply held skepticism.
While Pinpoint’s predictions are powered by machine learning application to massive data sets, the output is uniquely associated with the detailed personality dimensions of the audience. This is effectively how we show our work, and it enables marketers to build upon our results with their own native understanding of personality and compelling communication styles to amplify the effects of the findings.