Predictive marketing and machine learning: The next step for customer engagement

Predictions rarely come true. Take the CIA’s report on what 2015 would look like as an example. As expected, most of its projections were incorrect.

The soothsayers were right in some cases (the internet’s growth), but many didn’t even appear in the report – 2008’s financial crisis is one significant omission.

So if political estimates, social prophecies or even the weather are difficult to get right, is there anything that can be predicted with confidence?

Understanding human behaviour

It might seem surprising to say ‘us’, but we as humans are more predictable than you might think. Aside from irrational exceptions, we are creatures of habit and over a long period of time tend to show distinct behavioural patterns. What we eat, wear, watch and shop for are consistent, especially if our lives are analysed over a number of years.

The advent of analytical, algorithmic and diagnostic technologies shows how true this statement is. Analytics is definitely the buzzword of business at the moment. Retailing is poised to capture human behaviour to its advantage and marketers are wise to this change. Many marketing teams are now implementing tools that predict human behaviour commercially and boost campaign performance.

Essentially, if you are a retailer with a customer database, gathering information from human behaviour offers plenty of opportunities for growth. Even the most seemingly individual shopping habits can be logged, stored and analysed with the right solution.

The result: real-time, personalised content on a level never seen before.

Automated predictions for improved performance

Having advanced capabilities such as the above is becoming increasingly essential for the modern marketer. Most retailers are working harder than ever to attract, engage and retain customers. Brands know they have to keep customers engaged across every channel and device. Data is a crucial element in this battle to engage and retain.

The efficient use of data is crucial. An example of this is when customer data is used to create automated messages that are triggered in response to specific activities. This might include shopping basket reminders to address cart abandonment or buying recommendations based on a customer’s known tastes.

Behind these triggers is complex statistical analysis. Machine learning software works in the background to process diverse behavioural data including page views, check-outs, add-to-basket events and website search queries.

Customer interactions with thousands of products are then processed to provide real-time, personal recommendations with each page refresh. Solutions continuously test and update databases to establish a detailed record of known customer behaviour so that it can be predicted in the future.

Underlying trends

As you would expect, this involves a substantial amount of data. Currently, most marketing platforms still require some level of human interaction. Customer segments need to be built, discounts created and exact timings agreed.

The human element will not be abandoned any time soon, however, retailers can start to look beyond the limits of their own knowledge with a more progressive outlook on customer intelligence.

Meeting customer expectations, competing online and in physical stores, and incorporating new channels are huge challenges. The benefits of deeper customer insights are many and will change your overall strategy and expand your thinking on customer behaviour.

Customer analysis enables a more detailed understanding of every customer, supports product promotion and offers that will trigger positive responses in customers and drives higher engagement and a closer buyer community.

One detail to bear in mind: each retail business and customer is different. Recommendation models for your business should be designed for each stage of your customer’s journey (from research and discovery, basket purchases, to post-purchase) and take into account channel-related behaviour.

Furthermore, if you are concerned with false negatives, only genuine behaviour is captured by the most advanced machine learning solutions. Complex learning algorithms filter out irregular online activity (i.e. automated bots – so only genuine human behaviour is captured).

For retailers with large, regularly updated catalogues, solutions can also manage long-tail items involving little or no behavioural traffic.

Predicting customer actions based on individual behaviour and habits is becoming the reality of retail marketing today. One prediction we can make will be that predictive marketing will become far more important than was first expected.

By Steven Ledgerwood, managing director, UK at Emarsys