Evan Hanlon is the Director of Audience Strategy at Xaxis.
There’s a certain type of person I’ve learned not to trust: someone who insists that they can make things simple. I recognize how problematic that is for me, a person whose job it is to in fact make things simple. But I suppose it’s because I dive into what lurks behind the design studio slides and know just how deep the rabbit hole goes.
We don't shill.
Check out TNW's Hard Fork.
It’s ironic that almost every eye-roll inducing invocation of the phrase “big data” is accompanied by a sales pitch for a tool that makes it easy. Data is wielded as a magic wand, a cure-all for the new information age, wedged firmly between dotcom bubble one and dotcom bubble two.
It’s then boxed up like so many other solutions, placed on a well-manicured shelf, and then promptly flies off it and into the hands of brand marketers, their agencies, and dozens of third parties in between.
The main flaw in this system, however, is the reductio ad absurdum that something as chaotic, unwieldy, and often times unstructured as large, non-intuitive data can be turned into something easy. It can’t. But that’s not necessarily a bad thing.
There’s no simple solution for winning at chess. However, by embracing the complexity that comes with today’s intricate systems that knit together the programmatic audience buying landscape, we actually arm ourselves with more tools.
Where complex data matter
Take predictive modeling. Instead of targeting just one or two obvious signals (think buying a pre-built segment of ice cream intenders when you’re looking for people who want ice cream) a competent audience buyer looks at hundreds or thousands of non-intuitive, anonymous inputs to create models that are scored against a certain event (in this case, ice cream purchases).
Gym membership status, car ownership, the last time someone searched for flights, level of social media sharing, propensity to buy designer shoes, enthusiasm for team sports; these are all data points that at face value have no relation to ice cream. But with high enough volume, and when considered in the right combination, these things become statistically important.
The face value of purchase and intent signals is only the beginning; it’s the complex qualifications on the rest of the world that provide the power.
Elements of complexity
So complexity, or more specifically, the ability to successfully harness complexity in the service of brand goals, is good. For marketers in the programmatic environment the question then becomes whether there are simple and effective ways to evaluate these complex systems.
As a starting point for evaluating that shiny black box, we’ve assembled a short checklist of elements to consider.
Scale & Speed
Programmatic is based around the law of large numbers. The more information and the more information sources immediately on hand allow for better precision when it comes to identifying the best audiences for any particular marketing message.
Programmatic is a highly probabilistic endeavor. Bumping up the number of datasets you’re able to crunch in context with one another yields exponential improvements and allows ads to be delivered more quickly and with more confidence.
Science & Engineering
Simply throwing a bunch of data into a bucket and stirring, doesn’t automatically equal quality outcomes. Rather, the results will only be as good as the system’s ability to make sense of all the data feeding into it.
The best systems have able teams of data scientists and engineers steering the ship and are constantly self-improving. The more decisions that are made, the more results we have, which, in turn, leads to clearer insight and ever improving decision making going forward.
Open Architecture & Sequential Capabilities
Nothing exists in a vacuum, least of all digital media. As a result, you need to be able to ingest, integrate, and activate against data flows from every available source, and more.
For instance, you know how your mobile ads perform in context to display, online video, etc. That allows brands to create ad sequences to relevant audiences at scale that deliver better performance than single channel ads evaluated independently of the rest of the digital spend.
If you can’t create sequential buys or measure various ad channels in context, you’re getting a blinkered view of total performance and, consequently, are similarly limited in overall campaign effectiveness. And nothing is more important today than the biggest picture possible.
The last factor of a truly killer programmatic platform, however, is the human part of the equation. All the processing power in the world means little without a capable human hand steering the ship.
Even though technology extends our capacity to understand complex data sets, we should never forget that people are not data points. The course of history has shown that irrationality is one of our specialties, which means there’s no one better than a human to think like a human.
So the next time you run across someone bearing tidings of simplicity, should you run the other way? Absolutely not – after all it might be me. Instead, go in with a clear understanding that simple simply means the shiny logo capping what should in fact be a complex system.
Big data is inherently intricate. While “making it easy” is, of course, the goal of all the players in the space – when selling to non-players – the better your understanding of what exactly is being made easy, the more confidence you can have that your “simple” solution is amply delivering on its big data promise.