Retail Reflections Part One: Is My Machine Learning About Retail?

Guest blog by Alan Morris. Alan is the co-founder and brand ambassador of Retail Assist.

Even if you don’t take an interest in machine learning, it doesn’t mean that it won’t take one in you. We are all someone’s customers, and our digital footprint is being increasingly tracked and used by retailers who want to profile us. They want to know our interests, observations, comments, likes and dislikes so they can produce for us a tailor-made customer experience. The reason being? To encourage us to spend more money: this is the future of retail.  

Machine learning will be the key that unlocks retail’s big data and enables this to happen. It will provide insights so comprehensive and accurate that retailers will know exactly what their customers want to buy, and how and when they want to buy it. This will give retailers a new data driven confidence. Going forward, these insights will be automatically actioned, and the results will improve the supply chain process to a level that the buying, merchandising and logistics teams never thought possible. If you are cynical, remember that the history of innovation is the story of ideas that seemed dumb at the time.

Considering how retailers currently perform data analytics will help position machine learning. Retailers use spreadsheets, planning and forecasting applications, and Business Intelligence (BI) tools, to generate insights to improve trading performance. These solutions forecast what should happen and playback what has happened. They are strong when it comes to broadcasting good, bad and indifferent performance, but they fall short of explaining why some things work and some things don’t. This insight comes from how the people that use the technology interpret the information they are presented with, and the validity of their opinion is based upon their experience and understanding of retail.

The problem for humans when analysing data is the more you learn, the more you realise what you don’t know. You begin acknowledging that to understand something totally, you must sometimes break away and look at things differently; then, once you have looked at things differently, you need to look at them differently again. The best insights are those that are achieved when you completely understand the relationships between all the variables in play that can affect the scenario you are considering. Traditionally, retailers analyse data by looking for recognised patterns such as ‘what was bought with what?’, ‘which products sold best?’, and ‘how do sales compare to this time last year?’. However, they don’t fully help you to understand the different relationships that exist within the data, so they can’t fully appreciate the importance these may have when it comes to making informed decisions. This requires working through thousands of computations, testing different theories based upon how to improve performance. This arduous process is how you learn and build the experience so that the insights you give are meaningful.  

Some have suggested that there is a lack of data in retail businesses. I know from personal experience that there is no shortage in terms of volume, but based upon current demands, I believe that, in most businesses, the data sets used for modelling are either incomplete, inconsistent or both. Retail systems aren’t traditionally known for their capability to garner, store and process the type of unstructured data generated by Twitter, Facebook, Instagram and other social platforms. Retailers are also prone to creating many silos of data containing different truths. As I have said already, real insight needs every variable, and every relationship in the data to be explored, but it is crucial the data used is complete, accurate and consistent in order for retail businesses to thrive.

Read Part 2 of Alan’s blog on machine learning on here.

If you are looking for expert retail or hospitality IT support, get in touch with us here or email marketing@retail-assist.co.uk with your request.