Why are marketers kidding themselves that AI is about more than sales?

There will be many uses for machine learning in the future, but right now marketers should focus on improving sales by continually tweaking communications.


Potential applications of machine learning are very broad indeed. But, although AI conjures up images of robot butlers and promises big changes to customer experience, marketing teams who are already making use of AI-powered tech are doing so to sell.

Machine learning refers to statistical approaches to train models which incrementally improve the output of a system. This sort of predictive modelling is used to increase the likelihood that a customer will take a particular action – this could be opening an email, clicking an ad or viewing a recommended product.

Marketers are therefore mostly using machine learning in their push for personalisation – fuelled by a desire to improve sales. To put it simply, data scientists will use a variety of data – purchase history, demographics, browsing behaviour, etc – in order to predict a next best action such as what message to send, in what medium, at what time, and so on.

Colin Lewis, my fellow Marketing Week columnist and CMO at OpenJaw Technologies, warns in Econsultancy’s ‘Embracing Predictive Marketing’ report that marketers need to consider whether this technology is best employed purely with sales in mind: “If you’re using predictive modelling for the right reasons, you’re doing it to build the best customer experience. The problem is when personalisation is designed around the idea that you only want to sell more stuff. Then you’d be just as well giving the customer an offer.”

Of course, Lewis is right that the best uses of machine learning are all about improving experience – Spotify’s recommendations, for example – but it cannot be ignored that most uses in retail, for example, are all about sales.

Fifteen years from now, the question may become to what extent AI will help shape creative? But for now, marketers need to be much more hard-nosed.

This is nothing new either. Helena Andreas, CMO of Nordea, is quoted in the same report: “I worked at Tesco 15 years ago, and Tesco was advanced at the time because it had a vast database, and we used that to try and do smart analysis to figure out what customers would do if we put this on promotion or that on promotion, if we sent coupons, if we raised the price or did [two-for-one offers]. We looked at what would yield us the most profit and the biggest sales in the end.”

A classic use of machine learning in retail is to better predict customer lifetime value (CLV) and then allocate marketing spend accordingly (through retargeting, for example). Asos is one business doing this, and its data scientists have published papers on their work. One paper, ‘Customer Lifetime Value Prediction Using Embeddings’ by Ben Chamberlain et al, details how the Asos CLV model incorporates four broad classes of data – customer demographics, purchase history, returns history and web and app session logs (the last of these being the biggest data set).

In another marketing application of machine learning, Saul Lopez, customer lifecycle lead at Virgin Holidays, told the Econsultancy blog how the company has used Phrasee to optimise email subject lines, seeing an immediate increase in open rate of two percentage points (with no change in campaigns or segmentation). Given that Virgin Holidays sends 22 million emails every year, this is not insignificant. Furthermore, AI-optimised campaigns have seen a 33% increase in web traffic.

READ MORE: How to use email marketing effectively

Ben Barrass, head of data and analytics at Marketing Week and Econsultancy’s parent company Centaur Media, compares AI to blockchain, saying: “Marketers should understand the concepts and if they find out that something they need to do falls into the realm of machine learning, then be prepared to spend some cash and call in the experts.”

However, where blockchain has been described as a solution looking for a problem, the ability of machine learning to find meaning in large data sets and to optimise performance is an opportunity for all marketers.

Of course, there’s a big ‘if’. There are multiple, in fact. Legacy issues have to be sorted. Is data accessible in the right format? Is the data of the right quality? Do you have buy-in for a process which could take a year to bear fruit? Will you be able to flex resources?

Fifteen years from now, the question may become to what extent AI will help shape creative. Can machines be copywriters, art directors and video producers?

But for now, marketers need to be much more hard-nosed. As machine learning is plugged into the marketing stack through integrations with specialist software, more and more marketers will be able to incrementally improve sales by continually tweaking their communications.

That means if you’re still relying on a hunch about that product recommendation or email subject line, it’s time to start getting your house in order.

This is an edited extract from Econsultancy’s Digital Outlook 2019 report.

Ben Davis is editor of Econsultancy