What will you buy tomorrow?

There is a constant race being run in marketing analytics. In lane one, there is data. As we all know, as each year passes more and more data is created that can be used to describe consumers, touchpoints, journeys and decisions. In addition, more data is being merged, fused and connected to amplify this effect. Additionally, the ability to process this data continues to become faster and the costs continue to fall.

In lane two is complexity. Marketing executions are far more complex, involving ever more touchpoints across more platforms. Consumer choice continues to grow. Prices change in real time. Life is getting more complicated.

If complexity rises faster than our ability to understand it, our ability to create effective marketing and advertising is actually going to go backwards. Could it be true that we had a better understanding of how marketing worked in the 80s than we do now? It’s a bit more complicated than that.

I believe that we have become better at understanding how certain parts or silos of marketing work, and therefore have been able to exploit these successfully in isolation. For example building, adjusting and flexing the content and user experience on a website is now a finely tuned operation. The same goes for adjusting pricing to optimise profit and erode consumer surplus, or driving demand into websites, call centres or stores through search (both paid and organic).

In fact, all three of these examples are so well understood that they can almost be completely run using only intelligent software. However when you consider the whole marketing effort, or equally look at things through the eyes of the consumer, how everything works together (or not), remains difficult to understand, let alone ‘optimise’. To address this challenge, the marketing analytics industry has started to look at mathematical simulations and, in particular, agent based modelling. Mathematically and statistically speaking, this isn’t a new area. But the increases in data and processing power have started to open this up as a viable avenue to explore within marketing.

Agent based modelling has a number of features that make this a really interesting, and quite different, approach to the challenge. Agent based models work ‘bottom up’, rather than ‘top down’ – in plain English this means that instead of working with weekly chunks of data such as sales or exposures, we are working with individual people (although often scaled up, so one agent in a simulation might represent a thousand consumers). This moves us away from averages and allows us to take advantage of the growth of data. Instead of treating all consumers the same and working to average effects, all consumers can be different, and the data we work with, can be specific to them.

Because agent based models also integrate probabilities, you rarely get the same result from a simulation twice. While this might initially sound terrifying or just wrong, it’s actually a far more accurate reflection of reality, and also a more useful one.

At MEC, we have been working with agent based simulations over the last two years. In the process, we’ve developed a tool, MEC Velocity, that runs simulated effects of our clients marketing, predicting consumer behaviour and brand interaction into the future. One of the big advantages we have found is that we are able to work with different groups of consumers (audiences) within a single model. This enables us to see how different types of consumers react and to build different activation plans accordingly.

Another strength is that by their very nature simulations are time based. We are not only modelling the outcome, but we are working out when changes happen, how long they last for, and when they end. In some categories, it can take months and years for the effects of marketing plans to work through and deliver improvements in sales. This opens a completely new window in thinking about return on investment (when is the return?) and testing different plans to deliver shorter or longer term effects.

Furthermore, agent based simulations lend themselves incredibly well to visualizing the models, and this helps to get clients and stakeholders engaged, and to some extent to unravel the mathematics and rules that drive them.

The biggest challenge, however, is still creating confidence with clients in using a relatively new approach and, linked to this, the validation of the simulations. But the nuance of thinking about simulations and likelihood, rather than a fixed forecast, is an important one. Of course no one can predict the future, but visualizing how it might play out, and how that could be different as we change our assumptions or marketing plans, is incredibly powerful.

Right now predictive marketing is probably the most promising approach in the race to keep growing our clients’ business in a competitive marketplace.

James Northway is head of analytics and insight at MEC Global Solutions and EMEA