Predictive analytics sounds like every marketer’s dream come true. The premise is simple: look at past behavior and performance as an indicator of future behavior and performance, and react accordingly. It’s neat, tidy and predictable – just the way we like it. But there’s a major problem inherent in the design of predictive analytics that few people have acknowledged…
Predictive analytics is only just starting to find major use and adoption within the marketing arena, but it has long been a staple of other industries, including healthcare, insurance and financial institutions. In fact, predictive analytics was heavily used in the finance industry across the U.S. to create mortgage-backed securities Using hard science and predictive models that weighed the risks accordingly, the models predicted that there was no way so many people could default on their mortgages at the same time.
But what the models failed to take into account were the volatilities of the housing market – most notably, houses being worth less than the mortgage used to pay for them. Everyone trusted the models and few people bothered to put their results through proper checks and balances. A few folks noticed that most of the people given the all clear for borrowing were high risk, but by the time the domino effect began to happen, no one could slow it down. This is partially what caused the Great Housing Bubble of 2008 to burst, and with it, significant economic impacts across the country.
Do we blame “bad data” here? Not necessarily. Here, the predictive models failed to account for things like shifts and changes in how we lend and who we lend to. Those kinds of unknowns caused an already faulty foundation to crumble.
The Black Swan Theory
The inability of predictive analytics to account for these “unknowns” is known as the Black Swan Theory. It gets its name from the assumption (back in 16th century London) that all swans were white, because all depictions of swans and all knowledge of swans to that point showed that they were. But in the latter half of the century, when an explorer in Australia saw a black swan, the theory was disproven. All swans were not white.
But how could you possibly have accounted for that? You couldn’t, and therein lies the issue with predictive analytics. You can’t account for all of the unknowns, and even trying to plops you head-first into a bunch of complicated, complex models that even the best statisticians can’t wrap their heads around.
Painting a Smarter Picture of Analytics
Now, it’s worth mentioning that predictive analytics does have its place in marketing. We’re not painting all analytics platforms with the same broad strokes. Trusting major decisions or highly influential campaigns to predictive analytics is like tying your feet together before a marathon. Those are not the instances you’d want to rely on predictive data because there are simply too many unknowns to account for.
Here’s where it pays to rely on the cold, hard data you’re collecting. Predictive analytics is better suited to making small guesses and estimates. What could happen if we changed this? What might happen if we try this instead? Those are areas where predictive analytics truly shines. Of course, if you want to have true executive buy-in, you’ll likely be asked to prove that it’s worth it. But how can you prove anything with a platform that only predicts? You can’t – and therein lies the rub.
Why Is It So Difficult to Prove?
Predictive analytics does have its place in the marketing arsenal. But it’s not the magic pill that neatly wraps up all anticipatory consumer behaviors. Remember that the accuracy of your prediction is only as good as the data you feed it. There are just too many variables to account for, and so much uncertainty inherent in human psychology itself, that the platform just isn’t reliable for making concrete decisions. You simply can’t set it and forget it. Even complex calculations can’t be relied on entirely as truth – and as marketers, we’re busy folks, which means we’re tempted to grab at scraps of data and make pivotal decisions based on whatever we can get.
And with predictive analytics, that’s a recipe for disaster.
Getting the Right Help at the Right Time
There’s a lot of emphases (and responsibility) put on data scientists, not to mention a demanding set of skills in things like statistics, modeling and regression – skills that the average person simply doesn’t have. You can’t simply throw all your available data to the predictive analytics wall and hope it churns out something that makes sense. If you do, it will let you down every time.
Here, it’s best to hire an analytics expert to work with your team to help them make sense of the data while teaching them the necessary skills to harness predictive and probability analytics to the fullest. Analytics is a double-edged sword. The same skill that cuts through the big data clutter to give you actionable insights is also the same power to make drastic decisions based on imperfect information.
What is Predictive Analytics Good For?
From the tone of this article, it might sound like predictive analytics is a recipe for disaster – when nothing could be farther from the truth. It does have some excellent uses in the right scenarios, namely:
- Fine-tuning your buyer personas – There’s no better way to really hone in on what makes your customers tick than analyzing their past behavior and mining the rich seams of predictive analytics to brainstorm other options they may be attracted to.
- Creating more personalized marketing messages – How do your customers respond to personalization? And I’m talking about more than just a (first name) variable. Predictive analytics can show you the results of deeper, more unique messaging.
- Conceptualizing new products and services – Predictive analytics is great for brainstorming new product and service ideas that would appeal to your audience based on their past purchasing and browsing behavior.
- Determining the best route for your marketing spend – Predictive models can show you which marketing avenues and channels are worth your investment based on previous results that have produced a high ROI.
- Discovering which leads are worth the attention of your sales department – You can use predictive analytics to see which leads are most likely to respond well to nurturing campaigns and lead to a purchase, versus leads that are likely just tire-kickers. This helps free up time and maximize effort on your sales team, and more closed sales equals better results company-wide.
Making Data Manageable
The first step in making sense of predictive analytics is to start making significant strides in data organization. Gather all the relevant departments together and start leveraging all the data to make it manageable and actionable. You can even break it down into segmented “data sandboxes” to experiment with; you get all the benefits of the data you need, without compromising existing information. Here, you can play around with the models and get a feel for what predictive analytics is good at – and what it’s not.
That being said, it’s absolutely vital that if you want to leverage this new technology to the fullest, that you do so with a strategic plan in mind. Focus on one initiative at a time, whether that’s remarketing to existing customers or increasing order volume. This kind of laser-focused practice can help you get more out of predictive analytics than just hoping for a gold nugget to be sifted out of the data dirt.
Share Your Experience With Predictive Analytics
Are you using predictive analytics in your own marketing campaigns? How has it worked for you so far? Are you finding the results reliable or do you prefer working with more tried-and-true methods? Share your thoughts with us in the comments below and let us know your experiences.
*Featured Image Source