Retail Sales Forecasting Case Study
Our client is one of the largest FMCG retailers in the world, with more than 2,000 stores and 30,000 employees, operating spacious stores in a wide range of formats and locations. Product lines include general groceries, cosmetics, stationery, glass and kitchenware, electronic appliances, book, textiles, and other items along with other products.
With over 35 product categories across 2,000 stores, accurately predicting category-based sales volumes in different stores had the potential to deliver considerable operational efficiencies. Being able to track individual stores performance and processes would also provide an insight into business activities.
While the analytics team of our client had started to develop modelling for daily sales forecasts one month ahead for each category in each store, handling over 20 million of sales was a challenge – and the team were finding they were ending up with an almost unmanageable number of different models.
Our customer was interested in the solutions we can offer for inventory management and forecasting – our expertise was directly aligned with the problems they needed to solve. They approached us with a clear project objective in mind – to decrease the error rate in their modelling by at least 10%, and to increase the accuracy of category-based forecasts for each individual store. They were also keen to explore the potential to reduce the number of models they were generating by 50%.
Our Solution and Results
Our team at iERP created an innovative multi-layered next-generation algorithm to allow our customer to achieve more accurate sales forecasts and to reduce the number of models. We then created a solution that would process historical sales data and enrich this using external indicators which would allow our solution to understand the factors driving demand more accurately in order to predict demand more effectively. To achieve this, we considered a range of external parameters including timestamps, days of week, store categories, promotions, GDP growth rates, unemployment rates, inflation and credit card spending.
The results were impressive. We developed a series of blind tests to compare our forecasting values with the previous modelling used by the client and even with the number of models reduced by 50%, we managed to improve the error rate for sales forecast predictions by 10%.
The iERP sales forecasting software, deployed with the Supermarket retail chain, achieved a 10% reduction in forecast errors delivering potential savings of around €1.8 each month – that is equivalent to a potential €21.7m every year!
Our new solution is currently cloud-based and includes the potential to also be transferred to the secure on-premises environment at our customer facilities. We are now working with our customer to develop an algorithm and precision improvements for our predictive models by experimenting with different parameters.
iERP’s mission is to provide an end-to-end business prediction platform with modules that address multiple business scenarios and ZERO required knowledge of artificial intelligence or machine learning technologies.
We are helping companies to predict sales and inventory demands, what their customers are going to purchase next, or identify customers that are going to be late with the payment.