Predictive – The Latest Buzzword

by Vasudha Khandeparkar, Senior Data Analyst, RedEye RedEye
Predictive Blog

If you have spent any length of time with anyone involved in marketing you have undoubtedly heard of predictive analytics. Everyone wants to know how they can predict when their customer is about to make a purchase. Or perhaps when someone from their database is about to disengage. This knowledge could revolutionise our marketing – no wonder it is the next big buzzword!

So what actually is predictive analytics?

Predictive analytics is the broad term used to define all analytical methods which allow you to predict future probabilities and trends. When implemented correctly, the output allows you to understand what the impact of your actions would be. For example, it could help you to predict how many customers will go on to make a purchase off the back of one of your marketing campaigns.

Having performance knowledge before the campaign has been sent out is powerful, you could decide to exclude individuals who are most likely to unsubscribe from your send. You may also choose to only target those customers who are most likely to convert.

Before jumping on the predictive bandwagon

In order to build or implement a predictive model, you need to be backed up by two things – good business understanding and good data.

Business understanding ensures that you are trying to predict the correct thing – think of it as finding the answer to the correct question. Building and implementing a bespoke model takes time and costs money. If the entire premise of the model is wrong, you could lose a lot of valuable resource.

Predictive analytics looks at the trends you have in your data. If your data quality is questionable, you could identify trends that do not really exist, leading to model predictions that are incorrect. You also need to make sure you actually have data to build a predictive model around. Sounds obvious, but if you don’t have any data around what products your customers purchase, then you won’t be able to build a predictive model that identifies the next product a customer is most likely to purchase.

Let’s look at some key predictive techniques

Do you have a revenue or sales forecast sheet? Congratulations! You are already using predictive analytics within your business. In the crowded marketplace of technical analytical terms, it is important to remember that all of us in some shape or form are probably already using predictive analytics without realising it.

  • Product Recommendation – Product association analysis and market basket analysis that help understand what products an individual is most likely to purchase next are a simple way of using predictive analytics. These can be implemented quite easily, both on websites and in campaigns.
  • CHAID – This method gives you a probability tree and allows you to decide which option you want to go with at each stage. However, while it is theoretically visual, it can be difficult to implement this seamlessly in a SaaS environment.
  • Regression Models – You have probably heard the term ‘logistic regression’ mentioned umpteen times in relation to predictive modelling. This is one of the most common techniques used as it’s incredibly powerful, not just in providing a probability but also identifying variables that have the maximum impact on customer behaviour.

Don’t forget the past while looking into the future

So why not jump straight into predicting the future? Campaign management becomes a lot easier if you know what will happen in the future. Well, predictive analytics is not a crystal ball! It does not tell you what will happen in the future. It only gives you an indication of what could happen. In fact, any predictive model uses historic data to help you understand what could potentially happen in the future.

It is important to remember that predictive analytics is a step in the analytics journey and not the end goal. Descriptive and diagnostic analytics have an important role to play, as does business and market knowledge in making the right marketing decisions.

Access more insights like this on the RedEye Blog.

Topics

Analytics Data