Amid the retail apocalypse that has led many stores to face breakdowns, several brands have gradually woken up to the reality of the power of data. The highly competitive industry, which is transforming into a data-driven landscape, is adopting data science to get a competitive edge upon their business. Proactive investments in data science are being made by retail companies to stay ahead of others, and to expedite better customer experiences.
According an estimate by McKinsey, the retail sector can use data science to raise the margins by up to 60%. Moreover, in the presence of the internet, digital technology has become ubiquitous, making companies look for innovative ways to collect online content (structured as well as unstructured data) about customers’ behaviour, such as by using data science tools. It was observed that data science adoption for all companies reached 53% in 2017, compared to 17% in 2015.
Challenges where data science can help
The concept of mass marketing both in terms of traditions and experience started getting replaced by data science around two decades ago. Consequently, verticals like healthcare, finance, and evidently, retail also aimed to channelise and manage data. This shows that retailers are willing to spend more on niche technology to get better results. Moreover, the retail analytics market around the world is estimated to increase by $5.1 billion to $8.64 billion by 2022. From optimising shelves to forecasting sales, data science is now being applied to each facet of the retail business such as:
· Forecasting and planning
· Competitive price
· Retail transportation and logistics
· Promotions, markdowns, cross-promotions
· Allocation and replenishments
· Inventory management
· Dynamic product recommendation for up-sell and cross-sell
· Customer experience management
· Inventory visibility across the enterprise and available to promise
· Buying and merchandising
· Reputation management on social and web
· Determining store formats as well as merchandising in them and many more, such as planogramming
By coming to terms with the fact that data science analytics is now being applied at each step of the retail process, there are a wide of range of tools accessible to retailers. Ideally, marketers and brands engage in several types of analysis based on the combination of structured and unstructured data. They can make effective use of the data science tools to evaluate the most effective route to attain increased sales and customer engagement.
Time series and spatio-temporal data analysis
A specialised division of statistics that is extensively used in the field of operations research and econometrics is called time-series analysis. There can be several data sets, which are cross-sectional and signify a single slice of time. On the other hand, data collected over several periods, such as weekly sales data, weather, weight, and home energy usage are the examples of data which can be gathered at regular intervals.
An additive model is considered one of the most powerful and simple techniques for evaluating and forecasting periodic data. It is mainly used to present the time-series as a combination of varied magnitude patterns like daily, seasonally, weekly, and yearly, together with the general trend. Mr. A’s energy usage might increase in the summer and drop in the winter, but the overall trend reduces, as he alleviates the energy efficiency of his house. An additive model displays both patterns/trends and gives predictions on the basis of these observations.
Recommender systems are methods that offer suggestions to users with any kind of content that a user might prefer. It plays a major role in the decision-making processes. To put it simply, personalised recommendations are provided as ranked lists of suggestions. In order to get these rankings, recommender systems tend to predict an appropriate list of products or services on the basis of the preferences and constraints of users.
Creating a recommendation system is commonly confronted by users on platforms like Google, Amazon, Netflix, and Spotify. The main objective is to personalise content as well as recognise relevant data for audiences. Content may be available in varied forms such as movies, games, articles, and many more. Most popular types of recommender systems are content-based, collaborative, and popularity-based, to name a few.
Game theory is concerned with finding an optimal solution for a specific situation. Suppose Mr. X is driving down a road in heavy traffic. He switches to another road where traffic is moving faster. Although, after some time he observes that traffic in the previous road is now moving at an even faster pace. In this case, Mr. X will have to take a strategic decision—should he stay where he is or should he change back to the previous road?
Therefore, all game theories have to do with is understanding how well an individual performs in comparison to others and vice-versa. Apart from games such as football, poker, and chess that are suitable for Game theory, there are several other strategic decisions like customer engagement, picking a job amongst many job options, investing, etc. game theory is applied in many strategic decision-making areas, including economics, geosciences, sports and politics. It helps retail enterprises to foretell likely results for individuals, societies, and, of course, businesses.
Optimisation is used in transforming the knowledge into making effective and efficient decisions. It is an important element in the process of system design that leads to cheaper or higher cost, lesser processing time and much more. In addition to other fields, optimisation techniques are being widely applied in the retail sector as well. An effective application of optimisation technique needs at least three conditions to be fulfilled, such as the ability to solve problems through mathematical models, a complete understanding of the optimisation technique, and acquaintance with computer programs.
Data mining is fetching the information from databases and using it to make significant business decisions. It is also called unsupervised machine learning and is used extensively in CRM (customer relationship management) software for cluster analysis and market segmentation. Recently, there has been an increased interest in knowledge discovery from data, which is an essential tool to identify information for decision-making from the presented data set. It is used by retailers on a large scale through a set of different data mining methods for knowledge discovery for differentiating the correlation and patterns that were not known earlier. It is technically a bridge between machine learning and statistics. The major objective of the data mining process is to extract information from different data sets to turn it into an appropriate and comprehensive structure for end-use.
Artificial intelligence - Superforecasting
Artificial intelligence encompasses machine learning, data mining, and artificial intelligence (AI) itself, where it deals majorly with reinforcement learning methods. AI is a complex subject, as it seeks to replicate the human brain stimuli-action-reward/punishment process. Mainly, it is attained by building an artificial neural network that represents human intelligence such as logical reasoning, self-correction, and learning. Retailers are inclining towards this technology to eliminate errors and achieve precision. With the appropriate training of the machine, a retailer can probably estimate his shop-fill inventory more accurately that would not let him mark down the prices of his products. Logistics and transportation actually use a lot of operation research methods along with AI for optimisation.
The biggest challenge in the retail industry is to maintain profitable stocks that allow customer offtake without investing significantly in inventories and retail spaces (real estate on rental or on lease to store such inventory). The core is inventory management that links with the entire process of doing retail business, from merchandising, procurement, allocation, supply chain optimisation, mark downs and space management to offline and online reputation management of the brand and customer service. Retailers, irrespective of their size can benefit from data science quite significantly. Moreover, innovative thinking and approaches in data science are more likely to show its impact in the retail industry in the coming days.
The author is director and faculty of data science at SP Jain School of Global Management. Views are personal