Many marketers and agencies find predictive modelling to be a frustrating and complex discipline. The reasons for this are simple: predictive modelling demands patience, an analytical mind set, and a realistic understanding of the real-world limitations and benefits of today's tools and statistical models.
But marketers that set out with the right expectations and a long-term strategic commitment to using predictive modelling quickly find that it grows into an indispensable part of the marketing toolkit. Used sensibly and with the understanding that it is not a silver bullet, predictive modelling can help marketers to get better returns from their marketing investment.
Predictive modelling defined
So, what is predictive modelling? It is essentially a statistical technique for predicting the probability of a future outcome based on historical and current data. It involves collecting data, formulating a model based on that data, and using that model to make forecasts. As the model is used more often and refined as more data becomes available, its validity grows.
This is the sort of stuff that actuaries in insurance companies are comfortable with, but it can be quite intimidating to the layman. But used wisely, predictive modelling can help marketers to understand, predict and influence the future behaviour of their customers, according to variables like gender, age, and purchase history.
I've been careful to avoid suggesting that predictive modelling can "predict" the future -it's more accurate to say that it helps us understand the probabilities and possibilities of certain behaviours based on the data we have to hand. Human behaviour is notoriously hard to predict with complete accuracy because it's impossible to account for and model every variable that influences what people do.
But with that caveat, some benefits that predictive modelling can deliver include:
A better understanding of the customer journey.
An effective way to measure spend against return on investment and its key performance indicators.
Insight into human behavioural patterns.
More efficient use of marketing budgets.
Predictive modelling must be combined with a marketer's intuition and experience to really deliver on its potential. It should be used to understand and shape an analysis, rather than being the only factor one uses to measure ROI and make business decisions.
Here are some key success factors for using predictive modelling in marketing applications:
- You need to have an understanding of the business's objectives, including ROI goals.
- You must review results regularly.
- You need a tactical plan to execute your media buys based on the ongoing statistical analyses.
- You should use different models for different phases of the customer purchase funnel (acquisition, retention, etc.).
- You must take into account that diverse customer bases consume various media types differently.
- Put focus on operationalising your findings rather than on evaluating different models.
As I hinted earlier, the past cannot always reliably predict the future. We are dealing with complex systems of human behaviour, where the conditions (or constants) are changing all the time. What's more, there many variables that influence how people act, and it is difficult to capture and measure them all.
Yet organisations that use care in planning and executing on predictive modelling can achieve great returns from it. It's not the right tool for every job and it depends on using the right statistical models, but it can be an invaluable supporting tool for marketers and agencies trying to make complex spending decisions.