There is a huge amount of buzz and excitement about at the moment, all centering on the potential for artificial intelligence (AI) to transform marketing. For example, the EIU reports that executives were on board. Three quarters of those surveyed, were reported to be expecting AI to be used in their company within the next three years. Forrester takes it further, and advises CIOs on what they should be doing to ensure that their organisation can leverage AI in marketing over the next few years. The enthusiasm is both compelling and infectious. But it is also, in my opinion, worrying. Are we in danger of trying to run before we can walk?
Much of the hype is still very much about potential. There are very few good examples of AI or advanced analytics being used to provide insights in a coherent way. There are pockets of developing good practice, such as a recent announcement of a partnership involving McKinsey’s Periscope cloud and analytics platform and Nielsen’s point-of-sales and other retail data. The alliance will give retail and FMCG clients of Periscope access to Nielsen’s data. The idea is that the combination of Nielsen’s data with Periscope’s analytic capabilities will mean better, faster, data-driven decisions in sales and marketing. Customers will be able to connect applications and data sets more easily, improving collaboration. This is, however, not only at a very early stage, but it is also described as an industry ‘first’.
The missing ingredients
My real issue with all the AI hype is that customers repeatedly complain that they are still unable to use existing data effectively in marketing. The first of two particular issues that comes up a lot is transactional data. This is data about each customer’s transactions with the company. It includes their sales history, but also goes a lot further. The second issue is merging behavioral and social data with transactional data.
Properly held and managed, transactional data allows you to see how customers have responded to previous marketing campaigns. Have they ‘clicked through’, gone to the store, or done nothing? You can then target marketing much more effectively to suit customer preferences. Co-operative databases, including data from other organisations in the same sector, allow you to see whether your customers are still buying, but from elsewhere, and give you a more comprehensive picture of the market.
Siloed processes and data-holding
It is pretty obvious that it is going to be easier to sell to someone if you know what they have previously bought and under what circumstances. It is even more obvious that you can target marketing if you know how customers have responded to previous campaigns, so why are companies failing to capitalise on this data?
Sometimes the answer is simple: Transaction data is not managed by the marketing department, and access is limited. Siloed processes and data-holding — where data held in one department is not available to others within the company is a major problem. It leads to problems like customers being invited to buy something that they already own. This can be an even bigger problem if your customer actually bought the product from you, because you will actively put them off buying again. This is a concept that has been described as artificial stupidity, and often emerges from inappropriate automation of processes, using the wrong, or insufficient data to drive them.
If you were starting from scratch, would you end up here?
The problem often lies in marketing processes. They draw on the wrong (or no) data, or use different data in disconnected processes or customer journeys, and therefore do not result in the desired behaviour in customers. Careless automation makes this worse, because simply automating the same processes will give the same results. But it is actually only a symptom of the original problem with the processes.
A useful question to ask with automation, and with process design more generally: If you were starting from scratch, is this where you would end up? In other words, if you were designing marketing processes again, but this time starting with the data, would you end up with your current processes? If the answer to that is no, then simply automating existing processes is not going to make you data-driven.
Instead, you need to start defining your process from an end user perspective and map the data view on these journeys. Good marketing processes need to bring together data from across the company that can provide insights into customer needs and drive buying behavior. Only when you have integrated data can you begin to generate insights, and even think about using AI to improve those.
