Knowing what tomorrow brings

Amidst the retail thrust to put the most-modern face, retailers have begun their tryst with the next-gen retail. As changes sweep through retail encompassing customer demand, customer interactions and customer experience, next generation retailer is making a mark by tapping into the potentials of digital technologies to embrace new ways and charm the customer. What makes the next gen retail tick is the trait to know what tomorrow can bring.

Looking forward, an apparel retailer sets a target of 700 shirts of a particular brand, of a particular color, to be made available for the subsequent week – to be distributed across 20 outlets. It is not the retailer’s intuition that has allowed him to arrive at the number. It was the retailer’s ability to forecast future demands. In another case, a toy retailer wants to run a promotion for a specific toy vehicle with a period fixed in mind. That plan was mooted once the retailer sensed that a specific toy distributed over outlets was about to lose traction in the coming weeks, resulting in the campaign idea to send those vehicles off the racks. Pre-emptive and proactive are integral traits of the next-gen retail, courtesy demand forecasting in this case.

With the next-gen retail keen about knowing things in advance, keen to acquire predictive and prescriptive powers, data has emerged as the zing thing to the next generation retailer. Retail in the future is more about managing the proliferation of data, using data to know how the future would unfold, what’s the right course of action and what is the ‘transformation’ in question to thrive in the future.

Data conundrum

The data scenario at one of the leading apparel retailers in UK portrays the ordeal of mining intelligence from data. Where data from internal and external systems proved to be of immense value, accelerating the data-to-nuggets-to-intelligent actions cycle hinged on the retailer’s ability to bring disparate data together. That was what that set the next-gen retailer on road to create a data lake leveraging Hadoop enabling multiple data access options like batch, real-time and in-memory processing.

The Ask

What datasets play a pivotal role depends on what the retailer wants to predict, foresee, forecast. As in the case of the apparel retailer, there was a growing need to know if there were customers in line to defect, a dire need to read demand well in advance and use price to their advantage to increase sales and beat competition.

For the apparel retailer, customer data, order data, delivery data, catalogue data and web logs, purchase histories, product preferences, competitors’ pricing and inventory, product demand, sales data, inputs from weather patterns, social data, events data contained hidden treasures and showed promise of providing answers to the problems at hand.

Machine learning algorithm for the problem

With every bit of relevant data available in the Hadoop-based data lake, moving from data to prediction and intelligence depended on leveraging right machine learning algorithms to address specific problems. The need to use price as the lever to increase sales magnified the relevance of Gaussian process regression, Bayesian linear regression, regression tree. While forecasting demand deserved the use of relevant ML algorithms, Time-series, regression analysis, clustering models were leveraged resulting in predictive forecasting model with an accuracy of 82%.

As to the demand forecasting problem, next generation retailer leveraged big data and ML algorithms to improve forecasts in the range of about 8-14% across SKUs. The predictive power gained helped the retailer to pin point slack periods as well as unfulfilled demands, create attractive promotions when the ‘dull demand’ periods were read early through predictive algorithms.

The transformation into a next-gen retailer means acquiring the power and ability to know what tomorrow is about to bring – it pertains to the diligent use of big data and advanced analytics to predict, foresee, forecast what’s about to come, prescribe the right set of actions and promote intelligence-based actions to accomplish successful business results.

As the next-gen retail gets better and better at predicting, anticipating events, and in addressing problems with actionable intelligence, Saksoft helped a leading apparel retailer take advantage of big data, machine learning algorithms and predictive modelling to drive customer churn analytics, customer 360-degree, sales forecasting, demand forecasting and price optimization strategies, with heartwarming results for the retailer.