Machine learning improves customer engagement and lifetime value

Marketing managers have long been running email campaigns, which take into consideration many factors that would impact campaign performance, such as RFM, geographic, gender, and purchase history. In running these email campaigns, marketing managers have been using segmentation and targeting techniques to deploy personalised campaigns.

So what capabilities does machine learning offer to retailers beyond those already provided by segmented campaigns?

First, and most important, ML can understand the specific tastes and interests of each consumer – even if you have millions of them in your database. For example, ML can analyse all the real-time and historical interests of each consumer during the past five years. By understanding the product attributes that each consumer finds valuable, computers can identify an ‘interest map’ for each consumer and identify product recommendations for each consumer.

Second, by using Natural Language Processing, Machine learning can build a ‘word cloud’ of each consumer’s preferences. That process allows marketers to rank products and other content for each consumer, even if you have vast numbers of SKUs in the product database.

 Third, algorithms trained through ML can identify customers likely to churn. From propensity models and Learning-To-Rank, trained algorithms can find relevant micro clues that indicate the consumer may be about to churn.

Fourth, because machine learning is based on continuous evaluation of massive databases, it can reduce the human biases that undermine sound marketing decisions. These cognitive distortions include excessive confidence, avoiding regret and chasing trends.

However, ML has its limitations. It may have biases derived from the data used to train algorithms or statistical quirks in its methodologies. To address this issue, marketers should review their software provider’s Guiding Principles for Machine Learning.

 Machine Learning derives its conclusions from existing data points fed into trained algorithms. It cannot predict the future to the extent that future patterns are not rooted in the past, such as significant discontinuities during the 2008 recession. Marketing professionals must make judgments, based partly on intuition, their editorial calendar, and their marketing strategy.

 In short, ML supplies powerful tools for improving email personalisation because of the recent leaps in computing power, available data and better algorithms. Nevertheless, talented marketing professionals will still be needed to drive strategy, manage the marketing calendar, and translate 1:1 personalisation into increased customer loyalty and LTV.

Now email marketers can sleep while the robots work

Among the benefits of ML is that marketers can now deploy genuinely personalised campaigns with higher speed and far less effort.

Much hated by marketers, email campaign production is the task of manually creating email campaign content, a job that can involve 80% of the time involved in getting an email campaign out the door.

Late lights and long days spent locked in a room are considered a standard practice in the email marketing profession. Now, though, artificial intelligence will help email marketers avoid the monotony.

The past decade has seen a series of leaps and bounds in artificial intelligence. Email marketers used to spend hours doing this work manually. Now, personalisation software platforms powered by artificial intelligence offer time-saving tools.

Google is using similar techniques to boost search engine ranking efficiency

Google is implementing a similar method, which marks the most significant change in its ranking algorithm for at least five years, and one of the biggest ever.

Until now, Google’s algorithm has tried to single out the keywords in any search query, ignoring the most common or smaller words that seem less significant. This enables it to identify the primary subject matter of a query, but often results in it misunderstanding a user’s precise request.

The new technique, known as BERT, relies on a vast, general-purpose language model that has been built up from the analysis of enormous amounts of online text. Rather than reading the string of words in a query sequentially, it reads them all at the same time — including common words that would have been ignored before.

Using similar techniques, also based on Natural Language Processing, email marketers can automatically curate and rank products within their email provider, with a sense of serendipity for consumers. This process is fully automatic, doesn’t require metatags, and eliminates 60 to 90 per cent of campaign production times.

Conclusion

To avoid falling behind in this brave new world, retailers must be prepared to adopt and test the benefits of machine learning personalisation; to learn what works and adapt accordingly. These innovative qualities are likely to favour smart marketers who understand the value of data and will deliver a competitive advantage against their slower-moving competitors.