In this the age of big data we can get our hands on an almost incomprehensible volume of information, including transactional, behavioural, demographic, compiled, user generated and experimental. But the key questions we have to ask ourselves are where do we start? Which pieces of data shall we use? And what value can we get out of this data iceberg?
The most vital tool for marketers to utilise, which for some tends to be the most surprising, is testing. Testing is a tool which has been around for a long time and links very closely to the final data type mentioned above.
Testing is a tool that is going to become even more imperative over the coming years as data volumes grow at an even more rapid pace.
The Testing Process
Data mining techniques can be used to look at the data and understand what has happened based on criteria such as classification, clustering and regression.
Predictive models can then be built to try and understand what could happen, or if it is a good model, what should happen, providing valuable insights.
Testing is the only way to understand how to make improvements now. It allows us to understand how to use the insight gained from data mining and predictive modelling to monetise the almost unlimited opportunity that resides within that data. Any good data scientist will look at their initial predictive model and know that a percentage will miss the target. That is the essence of a predictive model and the process you need to go through, you invest in a model over time to make it more robust. You also need to test your models in the real world to refine them, this is where it can work well for the marketer.
I would always recommend building a testing plan not only of your functionality and creatives but on your data. A simple place that all businesses can start with is segmentation. Set up a testing plan based on segments that you have been able to build from your available data sets. Over time augment these segments with other data that becomes available, to test the data’s value. As you test and learn, you will create more data that can be built back into those segments. Consider experimental data such as what gets a response or what is created and your programme can become more and more sophisticated, relevant and valuable.