Premium POV

Chaos Theory in Marketing

Updated. Originally published April 27, 2015.


What do you think of when you hear chaos theory?  The butterfly effect?  A crowd tipping cars after their team won the championship?

Or perhaps you remember chaos theorist Dr. Malcolm from Michael Crichton’s Jurassic Park:

They believed that prediction was just a function of keeping track of things. If you knew enough, you could predict anything. That’s been cherished scientific belief since Newton.


Chaos theory throws it right out the window.

What is Chaos Theory?

Chaos theory is a mathematical way to deal with complex systems.  The name is misleading, because chaos theory is really looking for order in the chaos. The theory is typically associated with the fields of physics, astronomy, biology, engineering, or sociology. It was born out of the study of weather systems.  A man named Edward Lorenz made a weather model, and one day ran it using a certain set of parameters to create a particular weather pattern.  Later he wanted to run it again, but cheated a bit, and started the simulation halfway through by inputting rounded-off values from the earlier run. The outcome was completely different.

Complex systems, such as the motion of planets, got scientists thinking about how to solve non-linear systems.  Not every system can be described by a summation of their parts.

It usually requires a supercomputer or, for a marketing company, a data scientist! The field of marketing has to find a pattern in its own chaos, so the chaos theory is quite applicable.

As the world gets more complex, the field of marketing deals with more target audiences, more avenues to reach them, growing data sets, complicated marketing funnels, and on and on.  The complexity grows by the day.  Marketing is no longer simply hanging a flyer at the local grocery store.  Marketing is a complex system of messaging directed at certain consumers, at certain times, in certain ways, on certain devices, with just the right message for their stage in the purchase cycle.

Marketing Campaigns are Chaotic Systems

Consider the complex marketing plan for a product launch. Each media buy has a function and is put in place to obtain a certain number of impressions, clicks, buys, or a set of other key performance indicators. Summing those KPIs individually, as if you had run each one by itself, for each piece of media across each channel (TV, radio, print, paid search, social media, online banners, etc.) will not show the possible success for the entire campaign. That would make this a linear problem.  Other variables such as the number of impressions before conversions, change of messaging through the funnel, and how the pieces of media work together in general, lead to results that are different from a summation of what each piece of media could do on its own. Marketing is a non-linear problem.  Welcome to chaos theory!

Chaos Theory in Predictive Marketing

The idea of predictive marketing has been around for a few years. Data has grown strong enough that marketers can look historically to build relationships between what a consumer has done and what they will most likely do next. Simple examples include people researching baby names before buying diapers, or people looking for a real estate agent before looking for a loan agent. Of course, there are more complex relationships in which AI can help modern marketers reach out to the audience most likely to convert based on the countless signals being watched.  This type of innovation is behind campaigns such as smart display on Google and conversion-based, look-alike campaigns on Facebook.

However, this is completely based on historical behavior. This type of marketing assumes that the future is going to be just like the present. There is no accounting for what will come that may change the path a consumer is on.

Even predicting the outcome of a cross channel campaign assumes a steady baseline, which is impossible.  What if AI could advance to the point of predicting the future of even chaotic systems?  Then, rather than basing the success of a plan on historical data, the predicted success could include future fluctuations.

When using machine learning to predict the outcome of a chaotic system, the equations describing the system are not actually known, which can be scary. All that’s needed is the data. There are certain problems in which the best answer is found by man acknowledging that a machine may come up with better modeling to get to an answer than our trusted historical ways.

A team from the University of Maryland led by chaos theorist Edward Ott have shown that machine learning is a powerful tool for predicting chaos. Results reported in Physical Review Letters and Chaos cover predicting the future of chaotic systems out to distant horizons. The effects of this on mankind could be far-reaching. Of course, there are basic applications such as weather predictions. But what about looking for early signs of a stroke, traffic patterns, the spread of a pandemic, continental drifting, or climate change?

Complex marketing plans are chaotic systems.  With multiple inputs, variables, and potential outcomes, predicting what will happen once everything launches is hard to do. All marketers want to promise their client that the plan in place is the best one that could have been created and to express the exact ROI. What if that was possible? What if, instead of relying on old equations to calculate results, marketers worked with machine learning to use data to predict the most likely outcome? Not only could the most optimal channel, fund allocation, messaging, and targeting combination be run, but results could be nearly promised.

At Warren Douglas, we thrive on complex problems and complex solutions. We believe in functioning as a consultant, not a contractor, seeking to solve the most complicated marketing problems using modern solutions. We use layers of data, scientific thinking, proprietary tools, and the newest approaches to give clients solutions as elaborate as their problems.

What problem do you need solved?


*Photo: Fractint/Wikimedia

, , , , ,

Learn more about the WD