Universal Information Services

Public relations practitioners love a fancy new phrase and the phrase “predictive analytics” has been making the rounds in recent years. The idea of being able to predict likely outcomes, based on variables around your PR strategy, is very seductive for those wanting to better control desired outcomes. I’ve written about the False Glitter Effect in past posts, but I wouldn’t necessarily say that predictive analytics falls into that category. Rather, I think the media intelligence and PR software industry has given the wrong name to the right idea.

Predictive analytics denotes that you should be able to predict outcomes based on historical data. However, while that might work with marketing or weather, it’s very hard to do with human behavior when so many variables can’t be controlled. One might say the chaos of humanity makes the promise of predictive anlytics impossible. I would say predictive modeling is much more possible. So what’s the difference between the two and why do I believe predictive modeling is a better promise?

It’s possible the the problem with this term isn’t whether you use “analytics” or “modeling,” but rather the predictive nature of what you’re trying to do.

Can you truly predict the outcomes? Our research currently indicates only a very narrow margin of efficacy to predictive analytics.

Predictive analytics appears to be somewhat reliable when two conditions are present. First, the outcome is a binary choice between X and Y. Second, those choices are competing for the same result. For example, we’ve been somewhat successful using predictive analytics to digest share of voice, key phrasing, messages, and sentiment when choosing a Super Bowl winner or a political race between two people. In all fairness, the odds makers in Las Vegas were also on our side with those scenarios which shows they are undoubtedly also predicting a winner based on additional factors not necessarily found in the media.

As far as being able to actually predict outcomes for your company or client, predictive analytics may fall short for some time. However, using predictive modeling to develop the most likely scenarios could be very helpful for public relations planning and future actions. For example, like your crisis management plan, having a variety of models that reflect potential outcomes could better prepare your manufacturing and support teams for increases or decreases in sales. Depending on the success or failure of a model, you could measure data points along the path to the point in time you targeted, further refining the accuracy of the potential outcome. When viewed like this one can easily see the value of preparing such models. Modeling is also more practical because you are planning for a general situation, not on hitting the bullseye of what was predicted… something that could be impossible.

Modeling for potential scenarios might be much more valuable than gambling on more targeted predictions, and the value of this science is growing rapidly. According to Zion Market Research, the global market from predictive analytics is expected to grow from $3.49 billion in 206 to almost $11 billion in 2022. If we include modeling as one of the forms of predictive analytics.

You are going to hear more from service providers like Universal Information Services as we push the science further into predictions. What I hope to deliver to our clients is a much more efficacious modeling tool, rather than a tool you’ll try to use to make predictions. In my experience, a crystal ball is very difficult to create but educated plans can be much more useful.

Leave your comments and tell us what you think of the next wave of PR science. Do you like the promise of predictive modeling or predictive analytics?

Todd Murphy  CEO Universal Information Services