With cheaper storage and faster computer processing, comes a vast accumulation of data driven by increased social connectedness and internet of things, among others. This has, in turn, led to drastic changes in consumer behaviours such as increased expectations regarding convenience and great shopping experiences.
This means that businesses need to dig into the volumes of data to get a 360 degree view of the customer behaviours, influences and motivations. Given the vast variations in structure and velocity of the data, technologies have emerged to help scientists build the tools needed to navigate these waters. There is an increase in open-source systems which result in faster adoption of new methods and cloud-computing which provides a lot more flexibility in how the data is consumed.
As a data science platform, dunnhumby has been at the forefront of using retail data to understand the customer. We have then applied this understanding towards implementing ‘Customer First’ solutions and provide unmatched shopping experiences through activations such as 1:1 personalised marketing activities, both online and offline.
Before moving on to the concepts of predictive and prescriptive analysis, let us also talk a little of our old friend, still relevant – descriptive analytics. It is looking at the historical data and condensing it into useful nuggets of information. This is only about “what has happened”.
The vast majority of big data analytics used by organizations falls into the category of descriptive analytics.
Predictive analytics is using historical data to predict what might happen in the future. For example, this set of customers might lapse from your brand. This has improved over time with the advent of strong machine learning algorithms and, more recently, deep learning.
Prescriptive analytics is when you use data-backed insights to tell businesses how to solve a problem. These prescriptions lead to the realisation of business goals.
But this must not be looked at as a separate discipline, but just an aspect of what has always been done. For dunnhumby, it is never predictive over prescriptive, rather they are interrelated – we believe in supporting the businesses around how to solve a problem at hand, not just identifying it.
How does predictive analytics work?
Predictive analytics is a way to train the machine, where a model learns the trend of historical data, using a variety of variables. If you need to predict the demand of a certain commodity, then the model will need information on its historical sales and what triggered them. For example, what happened on the day when it was sold the most/least, how was the weather, any event trigger; we work with hundreds of such variables to train a predictive model. Eventually, the model learns that given a set of conditions this is how the commodity performed in the past and then it is used to predict the average demand of that commodity at any day in the future.
Of course, the model goes through validations before it can be used for predictions. The types of models used are dependent on the business scenario. There is no one-size-fits-all formula.
Demand for predictive analytics in retail
India is an enticing market. We have seen organized retail, e-commerce and social media, all developing together. With mobile devices and their consumption on the rise, the online market place has evolved rapidly and has grown leaps and bounds.
With this tremendous growth, consumers are more mobile-savvy and demand personal attention. The noise created by so many players offering so many mediums and promotions makes it even more difficult for retailers and brands to make themselves stand out. Consumers have started to voice what they think is best for them and the retail businesses will need to keep their interests in mind before taking any major strides or decisions.
Analytics/Machine learning is penetrating every industry and functionally becoming an increasingly important ingredient for data driven business decisions. Most of the leading organizations have already tested their hands with machine learning tools and making it mainstream. In fact, analytics capabilities have made a leap forward in recent years and the coming years are going to witness machine learning as an integral part of analytical strategies even by relatively small-scale businesses as well, backed up by investments on big-data analytics.
Harnessing the customer data for decision making, machine learning tools and automated products are not going to be restricted just with big companies or any sphere in coming years. The huge potential and untapped opportunities with small sectors, specific economies and certain scale will offer immense growth to the analytics world. Strategies need to be aligned with the direction of market growth and one must be ready for future opportunities.
Future of predictive analytics in retail
Predictive analytics is the present and future of retail. With increased competition and with customers moving online, the game is changing. We ask ourselves questions whether brand loyalty is still a thing? It is getting even more challenging with Amazon offering lowest prices and with discounters like Aldi, Lidl.
So it is imperative that we know beforehand what the customer wants, when they want it and where they would be at a point in time. It is about seizing the opportunity and personalisation is key. The answer to these questions lies in predictive analytics.
Authored by Shikha Tara, Analysis Director at dunnhumby- The world’s first customer data science platform