SUBSCRIBE TO OUR NEWSLETTER

    SUBSCRIBE TO OUR NEWSLETTER

    Business and Predictive Analytics
    Predictive Analytics in Retail and Mass Market: how to predict your future receipts

    Never before have the words Artificial Intelligence (AI), Predictive Analytics and Machine Learning been so overused and inflated as in recent times.

    One thing, however, is certain: few people really know what they are about, and even fewer have tried to evaluate the practical impacts of their use on the forecasting processes of the Retail and Mass Market sector.

    Predictive Analytics: Artificial Intelligence and Machine Learning, let’s understand it better

    Let us first try to understand what lies beneath these terms. Artificial Intelligence is a very broad area that represents the ability of a machine to act with human capabilities such as: reasoning, learning, creativity and decision making. The earliest studies in these areas date back to the 1970s but it did not have much of a following due to two problems: lack of data and computational power.

    Among the various branches of AI we find Machine Learning, which involves the machine learning to solve problems by studying data about similar situations that happened in the past. It is no longer the programmer who writes the algorithm but it is the machine that deduces autonomously the rule.  These techniques are also increasingly coming to the fore thanks to Big Data: the more information I can give the machine, the more comprehensively it can understand and interpret the past. The problem, rather, is figuring out what data is useful to give to the machine so that it does not generate ‘confusion’.

    How and why to use Machine Learning in Retail

    Focusing on the Retail domain, a typical application of Machine Learning (ML) is in the prediction of future receipts. It is necessary to provide the predictive engine with historical sales and map events deemed significant in the past (and therefore also in the future) such as, for example: store opening hours, Calendar Effect (holidays such as Easter), inflation (internal/external), particular market trends, promotions or other events (see COVID). This is all fed to the ML algorithm (there are several in the literature) so that it automatically learns and deduces the cause-and-effect relationships that each event had on past receipts.

    Once the learning phase is over, the machine is ready to apply the experience gained to the future; all it needs to do is to hypothesize how these events will happen in the future and the ML algorithm will define the collections. Such a project is not to be called “plug and play” you need to have a deep understanding of the business and, at the same time, of AI techniques.

    Certainly leaning on a consulting firm that has already experience in the field helps but does not trivialize the matter.  One proceeds with continuous fine tuning in order to identify the path that produces the best results.  One has to identify the right dimensional detail to consider (day, week or month? Retail Department or Aggregate Department?) and the historical depth to use in the learning phase (receipts from 20 years ago are probably not meaningful for predicting the future). In addition, all the events we are going to map must be measurable both in the past and in the future: for example, weather might be an impactful event (on a rainy weekend, attendance in a mall increases) but, at the same time, it is difficult to make medium- to long-term predictions (will it rain in 8 weekends?).

    Human and Artificial Intelligence: the pair of the future

    In conclusion, the introduction of these new processing techniques can certainly lead to benefits, such as: 

    • capability of a more ‘unbiased’ and deep interpretation of past data;
    • a faster speed of reaction and reprogramming (which is very important in the complex world in which we live).

    All of this, however, must always be done with a view to providing help and not replacing human intelligence.

    It is therefore important that the introduction of these techniques in the company goes hand in hand with the cultural training of the people who must use them, understand them and perceive AI as an aid and an added value and not as a black box that magically pulls out numbers that are difficult to interpret and explain.

    CONTACT US

    If you desire to get in touch with us and get more information, please fill out the form below

    9249