In its simplest form, predictive analytics is the use of data along with mathematical and statistical modelling to predict future behaviour based on observation of past behaviours. By analysing data, brands are able to predict what consumers are likely to do next.

An increasing number of brands are starting to recognise the benefits of using predictive analytics to unlock the value in existing data in order to gain an understanding of consumer behaviour. This understanding can then be used to target consumers more effectively. This approach is good for both brands and consumers. Brands can see how individuals respond and apply this to future communications to deliver more relevant content that fits their needs, leading to greater engagement and loyalty. This helps brands to ensure that marketing spend is allocated in the right areas to deliver the best return on investment.

With predictive analytics, ‘behaviour’ is the key word. Consumers tend to show a consistency of response to certain stimuli, whether in the form of marketing communications or the likelihood of renewing a subscription at the end of the 12 month period, for example. Of course, this varies depending on the segment or type of consumer and this is the challenge of predictive modelling. Brands have to understand how different consumers respond to different stimuli and then use their personal attributes and past behaviour to predict their likely response in the future.

Here is a guide to getting predictive analytics right:

1. Objective: As with any analytical exercise start with clear objectives and establish what you are trying to predict. This needs to be decided before setting out the analytical process and will allow you to select the right data, build the right model and then measure the success, or lack of.

2. Sampling: Once you know what you are trying to predict, it is possible to source past data that is representative of what you are trying to predict. This data can then be used to inform the analytical process. For example, if you want to look at 18-25 year olds, you would not want to include 40-year olds in the sample. The mistake that can often be made at this point is choosing too small a sample size, which means the results are likely to be too imprecise to be able to draw reasonable assumptions from, so make sure you have a robust sample size.

3. Choosing the modelling technique: It is important to choose the right modelling technique and this is where the skill and expertise come in. Choosing the wrong technique could leave you with a model that cannot be implemented or even worse, a model that delivers false or misleading results.

4. Test and learn: Once the model is built, test it out on some previously unseen data not used in the model build. If the model is built right, the results will be predicted accurately. If not, the model has probably been incorrectly specified, or there may be a fault with the sample.

5. Applying the model: There are many ways to apply the model, whether through decision rules or appending data on a regular basis. Make sure your technical teams and architectures are able to put it into practice.

6. Monitor throughout: As the market changes and consumer behaviour develops, your approach to modelling will need to adapt to ensure it continues to deliver accurate results.

7. Evaluation: The best predictive analytics comes from testing and learning to continually improve the modelling. To allow proper testing a control group should be used, this allows you to compare the results of your live sample against different scenarios. This way you can establish which techniques work and respond accordingly, giving you more accurate results in the future.

The ability to predict future behaviour through analytics allows marketers to optimise spend on activities that deliver maximum returns. This has contributed to an increase in the prevalence of predictive analytics, moving from cost saving combined with targeting in the days of direct mail to ensuring every opportunity is taken to increase income and achieve cut through in a crowded market, even though media today is reasonably inexpensive.

For many the use of predictive analytics is still a relatively new concept and getting the process right, from setting the right objectives through to measuring the results, is not as straightforward as it may first appear. However, as with anything that appears difficult on the surface, predictive analytics can be successfully implemented by having people in the team with the right level of skill and expertise. Get it right, and marketers have a real opportunity to predict how consumers are likely to react to certain stimuli and develop campaigns and initiatives around this that deliver both financial benefits for the business and better experiences for their customers.


By Nick Evans, Consulting Marketing Practice Director at Jaywing

PrivSec Conferences will bring together leading speakers and experts from privacy and security to deliver compelling content via solo presentations, panel discussions, debates, roundtables and workshops.
For more information on upcoming events, visit the website.

comments powered by Disqus