Way back in 1982, Thomas Dolby famously quipped “She blinded me with science and hit me with technology.” While musical tastes have since changed, the sentiment is stronger than ever. Science, especially data science, can be an intimidating subject for marketers and other non-mathematicians.
While risk managers have embraced data science for years (e.g., assessing credit worthiness, detecting transaction fraud, etc.), marketers still reside much further left on the technology adoption curve. My theory is that this hesitation is due to the opaque nature of data science. Many marketers are unable to trust a practice that cannot be explained (at least not in terms that they can understand).
Consider the extraordinary marketers you’ve known. What often makes them great is their ability to anticipate. They anticipate what their competitors will do, what their customers will respond to, and how their markets will evolve. Better yet, they can explain the reasons behind their theories in great detail. Their segmentation strategies and workflow rules all appear to make complete sense. They even measure their success with improvements in click-through rates, conversion ratios, and other marketing KPIs. After all, marketing is inherently a metric-driven profession.
Now consider how Big Data science could be used to make our ability to anticipate even greater. For starters, let’s segment data science (at least for common marketing purposes) into two primary categories: predictive and descriptive. Both methods have a role in anticipating future behaviors.
Descriptive analytics are used to uncover (or describe) the root causes of known outcomes. These methods utilize machine-learning techniques to discover complex correlations and patterns scattered across many sources of disparate data. For example, imagine the complexity (and benefit) of finding the contributing factors behind unscheduled maintenance on aircraft engines. While people can address the problems once known, the actual root cause discovery relies on data science to scour thousands of machine log files and find common factors leading up to device failure.
These same techniques will play a critical role in the evolution of attribution marketing. As marketers, we need to resist the urge to over-simplify the problem with logical explanations. Instead, trust the science to find the most impactful attribution elements. Focus on interrogating the results, not the methods, and you’ll learn to trust the unknown.
For marketers, predictive analytics are most useful when forecasting individuals’ future behaviors or responses. These techniques are incredibly useful in attrition forecasting (fading) and nano-segmentation campaigns (propensity) focused on cross-sell/up-sell opportunities. The data science techniques being employed have a long track record of success in non-marketing use cases. Understanding the lineage of these techniques may increase your comfort level when applying them to your next campaign.
It turns out that rocket science can be used to determine real-time pricing models. The navigation systems on guided rockets continually evaluate changes in wind, velocity, density altitude, and hundreds of other variables in order to hit their target. The same techniques apply to real-time pricing adjustments in auctions or other dynamic pricing environments. In these cases, the real-time ”navigation systems” are evaluating inventory levels, product features, recent sales prices, and consumer alternatives to determine a target price. Same math, different use case.
In the banking world, the algorithms used to determine unusual credit card transactions are very similar to the algorithms used to determine whether someone is at high risk of canceling an online dating service. In both cases, the science is focused on detecting anomalous account activity at the individual level. What’s unusual for one person may not be unusual for another, and that’s where the science excels. Again, same math, different use case.
As data science becomes more ingrained in our day-to-day lives, the extraordinary marketers will learn to trust it in the same way that risk managers and rocket scientists have for years. Hopefully you’ll find that the techniques are not that blinding after all.