Optimal Demand Sensing, Shaping & Response: What's in Store When Retailers Fully Leverage AI/ML
Digital transformations that leverage technologies like artificial intelligence and machine learning have increasingly had great success in driving profitable growth in many retail sectors.
Retailers have gone through the most tumultuous times in recent memory, as the COVID-19 pandemic delivered a growth windfall for grocers but decimated the demand in the fashion and apparel sector. Meanwhile, changing consumer preferences and the channels in which they shop will continue to challenge all types of retail supply chains.
Maintaining margins and service with these unprecedented changes requires a digital transformation of supply chain processes so that retailers can quickly analyze, optimize, and evaluate complex decisions before taking action. Key areas where retailers are digitally transforming their operations include demand sensing, demand shaping and ensuring product availability with a robust demand response.
Digital transformations that leverage technologies like artificial intelligence (AI) and machine learning (ML) have increasingly had great success in driving profitable growth in many retail sectors.
Here are some of the ways that AI/ML can transform a retailer’s demand sensing, demand shaping and demand response capabilities.
Leverage Technology to Improve Demand Sensing Capabilities
Retailers face increased demand volatility due to rapidly changing assortments, shorter product life cycles, increased promotional activity, social media influencers that go viral, and order volatility due to e-commerce. As a result, it is becoming more difficult to predict where demand will occur across brick-and-mortar and omnichannel and efficiently source and fulfill the right quantity of products to thousands and even millions of locations.
To be able to quickly respond to changing consumer buying patterns, forecasting techniques that leverage demand sensing capabilities are being deployed. Demand sensing focuses on eliminating supply chain lags by continuously learning and reducing the time between demand signals—such as order frequency, order size, and local events—and the response to those signals.
Newer ML mathematical techniques enable demand sensing with pattern recognition and ways to overcome latency issues associated with traditional time-series statistical methods. These techniques improve the accuracy of forecasts across all channels by incorporating and leveraging:
Newer algorithms like Gradient Boosting, Support Vector Machines, Tournament and Ensemble methods
Internal drivers like everyday shelf price, sale price, product placement, offers, digital coupons
A host of external causals like weather, GDP, new housing starts, mobility indices, local school and sports events, interest rates, inflation, and debt to income ratios
Building a Robust Demand Shaping Process
Retailers devote extensive resources trying to shape the demand through promotions and campaigns at both the store and online channels. They resort to various in-store promotions with temporary price discounts, displays, and feature inserts in local publications. There are also omnichannel demand shaping activities such as placement on the web site, special offers like free shipping, price reductions, email offers, digital coupons, and social media campaigns which drive incremental sales.
Robust modeling of these demand shaping activities can greatly benefit from ML techniques and there has been considerable success in using ARIMAX-based algorithms. Category managers can run “what-if” scenarios to look at the impact of changing the timing and duration of promotions, try different product placement strategies, offer different levels of price discounts or free shipping, and understand the impact of in-store sales or online orders. The expected demand can be broken out by fulfillment method (in-store sales, ship from store, pickup at store, ship from DC) to drive the inventory replenishment needed to deliver high customer service.
ML techniques are particularly good at understanding the impacts of these promotions on other items in the category or cross-category with cannibalization and halo effects. The category manager is looking for promotions that create net incremental revenue and margin, after considering cannibalization and pantry loading.
Improving Demand Response with a Digital Twin
Even with more accurate demand forecasting and robust modeling of demand shaping, retailers will still encounter out of stocks and inventory in the wrong locations, leading to expedites and unnecessary transfer costs. AI/ML techniques can optimize product availability, by anticipating customer fulfillment issues in advance and making prescriptive recommendations to take actions and mitigate poor customer service.
This requires end-to-end visibility with a digital twin — a digital representation of the physical supply chain where every asset is represented with their capacities and connections — available to be analyzed in real time to determine the next best action when exception conditions are detected.
With the application of AI/ML techniques on the underlying digital twin, retailers can evaluate trade-offs between demand, sourcing, transportation, flow path alternatives, inventory, and service in a holistic fashion. These analytics leverage a variety of AI algorithms starting with simple methods like business rules with scoring or scalable heuristics, linear models like logistic regression, and non-linear techniques like Gaussian mixture modeling.
Collaborating with Retailer Data Science Teams
Many retailers have data science teams that have developed cutting edge algorithms in critical areas such as store and omnichannel forecasting, labor capacity planning, assortment optimization, promotion and price modeling and out-of-stock analysis. However, a significant portion of these efforts never end up being fully deployed. Deploying AI/ML projects into usable applications remains a principal barrier to delivering business value.
A key competitive advantage for retailers is the ability to efficiently turn their algorithms and models into production grade deployed applications. This needs to be done quickly and efficiently without the need for a large software development team. Therefore, data scientists need a platform to productize their models that allows iterations safely and securely, as well as scalability to handle retail volumes.
This platform should contain a robust digital twin with current master data such as items, locations, capacities, suppliers, and policies, as well as transactional data on sell-thru, store orders, inventory, in-transit, and supplier orders. The platform should also support enhanced cross-functional coordination with role-based access, scenario management, workflows, and flexible reporting capabilities.
Overall, technology and automation will continue to play a big role in the transformation of retail with a relentless focus on supply chain design, localized assortments, anticipating consumer demand, shaping consumer purchases with pricing and promotions, and fulfilling demand across all channels in the most cost-effective manner.
When you’re one of the largest CPGs in the world, understanding consumers has become more important than ever. For The Kellogg Company, this means leaning into such technologies as AI and machine learning, and connecting the dots between their data across all channels.