By Elena Samuylova, Head of Marketing and Business Development, Yandex Data Factory.
The retail landscape has changed rapidly over the past decade, as has the way consumers are making their purchasing decisions and interacting with brands. Shoppers are overwhelmed with the amount of choices, messages and advertising they are exposed to, and easily switch between different outlets. The challenge for retailers is to differentiate themselves by delivering a compelling customer experience that builds loyalty among buyers.
To succeed retailers need to personalise their customer interactions – mimicking the quality of customer experience that was delivered by a small corner shop 50 years, where the owner knew all his clients by name, as well as their tastes and needs. Nowadays this personal touch can come through intelligent website content management systems or direct marketing, still staying the route to higher volume and value conversions. Yet there are important nuances as to how to deliver it.
We all know the key to true personalisation is to understand your customer as an individual; to provide the illusion of conscious person-to-person contact. Such an approach depends on holding deep, detailed information on each customer's preferences, engagements and history that would allow deducting their wishes and making appealing and timely offers. But here's the rub: the illusion of this personal touch actually depends on as little involvement from people as possible.
Retailers now accumulate vast amounts of customer data, including purchase history, behavioural and demographic information, logs of apps and web-site usage, and so on. They are even beginning to incorporate mobile operators' data including, location information and channel preferences.
This data is a precious asset that many other industries lack. At the same time, many retailers continue to underuse its full potential, limiting it to just being a source of insight to guide human decision-making. The data is gathered, visualized, analysed by the marketing team or similar; assumptions of trends and causation are applied to create charts and rules around customer communication. In effect, response mechanisms are integrated into the wider marketing activity, to add an "individual", tailored touch to it.
Technology, as such, is used to execute the plan, delivering the messages via the chosen channels, but humans determine the tactics of the most appropriate response.
And that is exactly where a huge portion of economically measurable value is lost. Humans' brainpower is, of course, essential to guide a strategy, to develop creative and to manage teams - but it is definitely not the right tool when it comes to dealing with big data on an operational level.
Humans lack the processing power to interpret the volume and complexity of the data, and to create the requisite thousands of hypotheses that are compared against each other to determine the most suitable course of action for each and every subscriber. Humans can and make great decisions on a case per case basis – remember that corner shop owner – but when presented with a task to deliver personal offers on a scale of the whole customer base, they have no other option than to generalise. Typically to the point that the data's power is dramatically undermined: the customer base is limited to a number of segments, and the suggested actions are affected, among others, by biases humans have in the decision making process, whether based on their personal experiences or perceived best practice.
In effect, big data's usefulness is often being downgraded and sporadic.
Data per se has no value, and generates costs of collection and storage. The value, on its side, is measured by the effect the data usage brings. And when it comes to big data, machine learning arrives as a complementary technology that allows overcoming the drawbacks of human analytics – by allowing the machines not only execute the decisions, but actually make those.
For retailers to really succeed with personalisation, the operational decision-making – such as which offers to send to each exact client – needs to be handed over to the machines. Using sophisticated algorithms machines make intelligent predictions based on data and recommend the best action, to be executed in an automatic fashion. For retailers, that means the ability to treat each customer as an individual, rather than a part of an assigned group. The algorithms continuously test, refine and reapply many hundreds of thousands of hypotheses, and learn from every action and reaction, thus giving every exact customer the offer that has the highest likelihood to be accepted – and makes them feel like they are taken care of.
Not only is the complexity of the algorithms and the speed at which they operate beyond the capability of humans, but so are the hypotheses that it tests. The chains of causation and the correlations that machine learning discovers are often ones that humans simply would not consider. And yet, because they are based solely on the incontestable data of what works and what evidently does not, they result in improved personalisation and greater conversion rates.
Machine learning's greatest advantage over previous generations of data analytics is that it does not require the retailer to have a deep understanding of personalisation techniques, nor to have data interpretation capabilities. The key is simply the provision of relevant historic data – which unlike many sectors, retailers are typically blessed with a wealth of - for the "machine" to interpret, test and act upon. This "black box" approach can seem scary – but the beauty of the technology is that it makes everything measurable: experimenting and A/B-testing allows comparing the effect of the machine learning solutions with traditional rule-based marketing and making a conscious choice on what simply works best.
It's important to note that this is not just blind hope for the future use of technology. Retailers are already enjoying the benefits of entrusting personalisation strategies to intelligent and predictive algorithms. For instance, 35 percent of what consumers purchase on Amazon and 75 percent of what customers watch on Netflix comes from product recommendations determined by machine learning algorithms. These companies born online made technology part of their strategy early on – and now it's time for "traditional" retail to catch up.
With an ever-growing and fragmenting market, retailers need to embrace marketing activities that help deliver a better customer experience in order to stay on top. Data has rightly been championed as the key to this for many years, but personalisation tactics have been too reliant on a person making decisions.
Ironically, with the arrival and proven use of machine learning, reaching customers on a personal level depends on the person stepping back and empowering the machine to do the job.