Customer-Centric Demand Forecasting

In the past few months, the unprecedented pandemic has impacted well established demand forecasting workflows Demand for essentials broke the seams, but DIY, apparel, footwear, consumer durables’ retailers witnessed a sharp decline in sales. The prolonged store closures and the staggered return to normalcy is forcing buyers and merchandisers to relook at ways and means to predict these demand aberrations and tweak their replenishment algorithms to generate near perfect purchase orders.

Non – essential retailers are having to deal with slow-moving / dead spring stocks in the stores, and excess summer stocks in the warehouse with slowdown in allocations to the stores. Grocery retailers, on the other hand, are grappling with optimizing fill rates on the shelves, removed multi-buy pricing, and imposed purchase restrictions. They have reported sharp rise in slow-moving stocks of private labels / store brands as shoppers focused more on availability and accessibility as a result of panic buying. Inventory that would have taken longer days to turn, was moving faster resulting in stock-outs in matter of hours if not in minutes.

Whilst there is no silver bullet to resolve these aberrations overnight, the solution lies in how robust or scalable the demand forecasting applications are, with focus on the following:

• Cost Savings – Reduce transport, inventory holding costs by leveraging big data effectively, and reducing the frequency of forecasting to weekly analysis. The closer the forecasting is to inventory turns, the lower the costs of moving inventory across the order to cash lifecycle
• Less Touch, Contactless – with the help of AI / ML demand forecasting should be self-driven with minimum manual interventions. The pandemic has taught pertinent lessons to everyone, and the replenishment engines would have to be trained to be smarter and be able to generate forecasts for the peaks and troughs.
• Supply Chain Efficiencies – As the fear of the pandemic recedes, the pent – up demand in categories like DIY, Cosmetics would potentially lead to a spike in sales thus creating a a halo effect for other categories. Retailers could witness a shift in destination categories, albeit briefly, and all this would compel them to ensure that efficiencies like IoT sensors, automated warehouse processes, paperless transactions are focus areas. This would lead to faster inbound shipments and result in store determined dynamic allocations.

As re-mediation efforts start picking pace, and stimulus packages are disbursed, the emergence of the New Normal will re-baseline all retail strategies, and demand forecasting will be at the forefront. Predicting shelf level visibility, real time replenishment algorithms, and centralized view of inventory would be key in driving the near and long term business benefits.

At RCG, we are focusing on delivering an AI driven Demand Forecasting Solution– powered by Stylumia, built on proprietary one-of-its-kind machine learning algorithms with data at internet scale. Whether it is forecasting trends, spotting winning products, predicting demand of new products, balancing width and depth of assortment or localizing assortment for a store, our Demand Forecasting solution covers it for you.

Some of the organizations that have witnessed 60% improvement in sales velocity, 30% optimization in inventory and 20% increase in full price sales include Vero Moda, Jack & Jones, Only, Pepe Jeans, New Balance, Pepe Jeans London, Fossil and many more.

To accomplish this, our Demand Forecasting solution focuses on 4 specific areas that can help clients overcome the common hurdles of demand forecasting and achieve accuracy:

• Apollo – Test new product potential before investing, grade relative potential of new ranges, and buy just the right amount of inventory
• Market Intelligence Tool – Laser sharp fashion forecast, instant validation of trends and products, gain entry into new, unexplored categories
• Storey – Localize assortment for every channel / store, optimal first allocation, dynamic intelligent replacement / replenishment
• Fashion Intelligence Tool – Simplify In-season analysis, decision making, & action, post-season diagnosis for range correction

The Demand Forecasting solution would accomplish a robust pilot in a 12-16 weeks for a cluster of the client’s Departments, Categories and Sub-Categories and deliver a state of the art machine learning model for the client’s use case, process the data, identify and handle anomalies, engineer features as required, establish a baseline lift to the current forecast accuracy using train-test-validate method and showcase the output accuracy across various metrics.