Photo by Tech Daily on Unsplash
How Will AI-Driven Inventory Management Revolutionize Retail in 2024?
As we wade through a post-pandemic world, the retail industry is reinventing itself with artificial intelligence (AI) technology. The prime focus? Gaining a more accurate prediction of shopper demand and resurrecting supply chains that have been rattled by capricious consumer buying models during the recent pandemic and the challenges with “just-in-time” inventory management. One such challenge lies in erroneous inventory data, where inaccurate sales floor quantities of specific SKUs lead to complications like imprecise backroom locations.
The pandemic forced a dramatic shift in consumer behavior, leading to the pitfalls of previous forecasting models that relied almost entirely on historic sales data alone. During the pandemic, consumers’ buying preferences constantly and rapidly switched between items such as cleaning supplies, home gym equipment, and office apparel and then into services like travel. The flux in shopping patterns led to an overflow of unsold goods in many retailers’ inventories, while demand for other items soared unchecked, adding to shortages. Streamlining these concerns is paramount for retail businesses aiming to save both time and money.
During the 2020 and 2021 peaks, companies ordered large-scale stocks as a buffer against supply-chain disruptions, often leading to a surplus of unwanted goods and shifts in consumer spending. According to the Census Bureau, by August 2022, retailers held about $760 billion in inventories, a 30% jump from August 2019, double the sales growth rate. Accurate demand forecasting is vital for profitability. Understanding customer behavior enables retailers to detect trends, make informed purchases, and devise appropriate pricing and promotional strategies.
AI Arrives on the Retail Scene
AI has emerged as the knight in shining armor, offering the potential to enhance inventory management with a variety of innovative strategies. Take this fascinating use case: AI can scrutinize consumer expenditure patterns to predict shifts in the movement of specific merchandise on the sales floor. This intelligence encourages a more focused approach to manual auditing, allowing personnel to target specific areas that demand attention rather than embarking on the Herculean task of auditing the entire store.
Of course, not all solutions need to involve a sophisticated camera network. In situations where manual efforts are necessary, machine learning algorithms can still provide valuable assistance. For instance, rather than performing a tedious item-by-item audit of empty shelf spots, a retail worker could simply capture a photograph of each shelf section using their device. Leveraging object recognition and deep learning capabilities, these photographs could be compared to a planogram to identify missing items swiftly. However, the system is not without its caveats. Achieving success with this method relies heavily on meticulous zoning, and challenges might arise in estimating quantities if items happen to obstruct one another.
Retailer Implementation of AI Inventory Management
Retailers such as Walmart, Walgreens, and online fashion dealer ASOS have started implementing advanced AI technology for retail inventory management, according to The Wall Street Journal. By harnessing parameters like weather patterns and social media trends, AI-driven algorithms can process vast arrays of data and aid in strategic decision-making regarding inventory placement. With this technological support, retailers aim to fine-tune their longstanding practice of employing internal historical sales data to forecast consumer demand. This would, in turn, help them stock their shelves with the right items at the right moment, addressing the many years of challenging and expensive inventory imbalances.
The complexities for retailers extend beyond merely gauging demand. The evolution of consumer shopping habits has made it increasingly challenging to determine where to place inventory. In our modern shopping landscape, consumers expect swift fulfillment for purchases made online for home delivery or store pickup. Therefore, retailers must strategically plan where to house their merchandise for fast and efficient movement.
Additionally, previous forecasting tools were unable to adequately consider how factors such as viral social media videos and local weather patterns impact customers’ shopping decisions. However, advancements in AI and machine-learning technologies have made it possible to incorporate this data into forecasting models.
Walmart, for instance, has programmed its inventory management system to consider weather forecasts and online search trends. With this data, the retailer can then use AI to anticipate regional demand for specific products and accordingly distribute inventory.
Walgreens also leverages AI technology to forecast demand, using data from social media and seasonal illness reports. The insights gleaned are then used to position inventory close to where consumers are expected to shop for these items.
For example, Rajnish Kapur, chief sourcing and supply chain officer, said that “the company’s AI-driven forecasting model last year helped predict regional and local trends during cough, cold and flu season so Walgreens could get over-the-counter products onto shelves. The model had predicted higher rates of fever and lower rates of congestion and cough, leading the retailer to stock more pediatric fever reducers in the areas where there was expected to be the most demand.”
Meanwhile, U.K.-based ASOS has started to employ AI for demand forecasting for items such as T-shirts, denim, and dresses. By combining past sales, returns data, product popularity, and trends, the AI technology provides a granular and accurate forecast that wasn’t previously possible.
Macy’s improved its gross margin despite a decline in sales during the last quarter of 2023, owing to tighter inventories. Its inventory was reduced by 6% in the third quarter compared to the previous year, and 17% compared to 2019. Tony Spring, the upcoming president and CEO, indicated that the retailer’s focus is shifting toward offering a wider variety of goods rather than stocking redundant items.
“The customer today does not want an endless aisle. They want the best aisle.”
Tony Spring, upcoming president and CEO of Macy’s, via The WSJ
In conclusion, as retailers align with the operational realities of a post-pandemic world, artificial intelligence emerges as the secret weapon to predict shopper demand, reduce inventory imbalances, and optimize supply chains.
Discussion Questions
What other research approaches could enhance the robustness of AI-driven predictive models in retail, especially incorporating sales data, customer behavior, social media trends, and external factors like weather? How will AI impact the evolution of the retail landscape, particularly in terms of inventory placement for online and in-store orders and variety of stock? Is the future of retail focused on variety rather than redundant stocking? Do you agree with the notion that customers don’t want an “endless aisle”?
Poll
Loading …BrainTrust
Inventory and supply chain management is a promising area for AI.
The ability of AI to analyze large data sets should help to improve the accuracy of demand forecasting and to allow for adjustments to be made more quickly. Potentially, AI can consider a much wider range of factors than traditional forecasting, which would make forecasts more sensitive.
If combined with RFID or other technology to track stock levels, AI can also automate stockkeeping tasks and ensure that inventory levels are accurate – which is a critical element of forecasting and inventory control.
Excellent points! Without real-time inventory accuracy, more accurate AI-based demand forecasts is not a complete win. If store inventory is inaccurate, which is often the case, replenishment orders may be over or under-stated and online order fulfillment from store inventory may still be a challenge. Combining AI-demand forecasting with real-time inventory is the Holy Grail.
There is no doubt that, at many levels, AI can improve inventory management. As an inventory predictor, AI can also predict revenues. What happens when predicted revenue comes up short of the corporate plan?
AI will impact the evolution of the retail landscape by enabling more accurate and dynamic demand forecasting, inventory optimization, and personalized shopping experience.
Using data from various sources, such as sales history, social media trends, weather patterns, and customer behavior, AI can help retailers anticipate and respond to changing consumer preferences and market conditions.
AI can also help retailers allocate inventory more efficiently across different channels, such as online, in-store, or delivery, and reduce inventory imbalances and costs as well as suggesting inventory pre-placement across the supply chain.
AI can also help retailers offer a products that match customer needs and tastes, rather than stocking redundant items. AI can also enhance customer loyalty and satisfaction by providing tailored recommendations, offers, and services.
The application of AI to inventory management holds great promise. However, it’s not magic. AI is driven by the quality and type of underlying data, as well as the assumptions used to weight any given factor. For example, how should weather influence inventory vs. social media input? The answer, it depends. Getting product mix right and minimizing excess inventory remains an ongoing challenge and while AI tools and models will help refine and improve inventory management, nothing is completely fool proof. As for “endless aisle” – I doubt that consumers think about it. Consumers want what they want…they want their own aisle.
AI is not a new phenomenon in demand forecasting and inventory management. We’ve seen algorithms used to execute many different strategies for many different retailers. What’s important to note is that AI tech continues to advance, pushing the boundaries of how it can assist in reducing the human burden of managing across thousands of stores, millions of SKUs and tens of millions of customers. The complexity that can be managed is now addressable, and AI is a simplifying tool to meet strategic goals. Expect to see more closely matched inventories to demand and lower safety stocks, reducing costs for retailers. More important, however will be scenario and contingency planning, because even the best AI isn’t a perfect crystal ball. Retailers will leverage AI to execute but also to prepare. Redundancy of stock will continue to be a tool, but it will depend on the retailer and their appetite for risk combined with their specific market strategies. AI will continue to shift the inventory management landscape as more powerful algorithms are applied to larger data sets.
Weather forecasters (using AI) can’t predict the weather, so why we think supply chain software will do a better job is a challenging thought, plus supply chains can only move so fast and consumer responses to weather are pretty immediate – a week or less, we’re talking about. Social media and search gives more around intent data (rather than history), which, if there are patterns to discern, AI could potentially find them – I personally like that example better. But at the end of the day, it’s all about predictions, and anyone thinking that they can math their way to better forecasts of what is ultimately an extremely complex and unpredictable global system of consumers through to raw material suppliers, is frankly insane. Balance forecasting science with execution flexibility. Don’t over-index on the forecast just because AI is trendy.
It’s easy to see that AI is going to lead to new levels of accuracy and efficiency in the management and allocation of existing inventory. If a retailer owns 10,000 units of a given style, they will now be much smarter about how those units are positioned to serve current demand at the store and ecomm level. Better sell thru, less residual inventory, higher margins. Great. It’s forecasting future demand and allocation that is the trickier part. AI might even give us better visibility in forecasting future demand. BUT…that forecast is only actionable IF the supply chain can manufacture the product in a timely manner. That is, in the window where the data is still valid. Data has a shelf life. And if August/September selling is telling the retailer they are underbought in a certain item for Christmas selling, can the supply chain manufacture additional inventory in time to arrive on the selling floor and sell through? Highly unlikely in most product categories.
So like the article says, resurrecting the supply chain is equally as important as the new found genius of AI. AI is like a whole new rocket fuel, capable of taking the business to new heights. But, the engine…the supply chain…has a speed cap. We can pour in the most amazing rocket fuel/data, but the engine can only go so fast. The engine needs a whole new rethink at the raw material, manufacturing and transport levels. And…country of origin level.
Real-time inventory management and visibility will be a key competitive differentiator for retailers throughout 2024. With the evolving consumer behaviors, By leveraging the power of AI capabilities, retailers will have the ability to be far more prescriptive with their inventory and order management strategies.
The sheer amount of data that AI technologies can consume and provide in the form of actionable forecasts will be a game-changer, as retailers not only race against the clock to mitigate the last mile but also gain the upper hand in reducing the over-inventoried situations we experienced during the pandemic. RFID technologies are yet another emerging capability that will enable retailers to track, manage, and forecast inventory across the entire product lifecycle.
There is no underestimating the power and impressive capabilities that AI offers in the form of machine learning and automation. Naturally, retailers will have to take a crawl, walk, run approach to incorporating both AI technologies and RFID capabilities into their merchandising, order management, and inventory management strategies. However, investments in these areas will be a critical strategy for 2024 and beyond.
Managing and improving the purchase process with AI is going to improve with sales and customer satisfaction. But to imply that AI is going to vastly improve inventory management, such as procuring inventory from vendors, is highly optimistic. Over the years, we have been able to understand seasonality, local cultures, trends in product preferences and dislikes and have generated recommended purchase orders to stock distribution centers and stores and reorders to replenish them both. AI is not going to change the season’s or accurately predict the weather this winter (can it?). AI will continue to improve finding the right match between customer and product, but AI’s potential to actually manage overall inventory is, in my opinion, not its forté. Call me shortsighted.
In my opinion, the “secret sauce” of AI is the ability to find discrete correlations in massive, disparate data sets. For fulfillment and replenishment, this capability has enormous potential. The replenishment space has been chasing the more accurate forecast and the more sensitive demand signal tool for as long as I’ve been in the industry. (Which is a really long time) For some categories, like CPG, relatively stable demand signals and reliable availability allowed us to take our eye off the ball when it came to managing disruption and accounting for volatility. The pandemic exposed just how far off the ball we were in many cases. In categories like apparel, where volatility is inherent due to trends and changing consumer tastes, forecasting demand, especially down the item/location level, the level that really matters to the consumer, was almost impossible; we created an environment where we managed to sell-through instead of in-stock, hopefully managing this year’s inventory issues in next seasons buy.
I believe that AI has the potential to move the needle significantly in both scenarios by enabling inventory managers to plan and, more importantly, react to potential volatility in demand. The tools can consider more discrete indicators of changing demand and do so at a more granular level than we can today. This will significantly impact customer service, i.e., reliability, and drive increased incremental sales.
I don’t mean to be Negative Nancy, but I think AI’s ability to predict demand is somewhat limited. I’m all in on the supply chain, and maybe, maybe quantities and timings of receipts, but the rest is all subject to the vagueries of human taste.
It’s all about the data. As the founder of a data warehousing and business intelligence company in 1996, I can still make that claim. The more data sources we integrate, the more we will learn.
For example, integrating POS data with shipping data along with order data and RFID data can help us determine where along the supply chain issues are occurring.
Years ago, while doing work with a pharmaceutical company, we integrated script information with CDC information as well as internal data to predict when and where medications would be most likely needed.
When working with a large beverage company, we integrated the traditional, historical sales information with NASCAR event information to predict volume increases.
AI and GenAI are just part of the business intelligence evolution that has been going on for 30 years. Today, algorithms and machine learnings process information to glean more insights and therefore are getting better and better at predictions and prescriptions.
The more data we integrate, the more we can learn. But key to this is making sure business rules apply and the data is clean, valid & integrated properly. Because as the old saying goes… “Garbage in, garbage out.”
Consumers are happier when their shopping experience is more personalized. Customers who receive a more personalized customer experience are likely to spend more at any given retailer. With artificial intelligence, brands are able to accomplish this more easily than ever before.
It’s not new, but AI-driven personalization is becoming increasingly sophisticated. By 2024, AI will redefine personalization as machine learning algorithms analyze vast amounts of customer data (such as browsing behavior, purchase history, and even social media activity) to create highly customizable shopping experiences.
By tailoring product recommendations, creating personalized marketing messages, and implementing dynamic pricing strategies, brands can help consumers discover the products they’re looking for while driving conversions and revenues.
In 2024, AI will change the game when it comes to helping customers find the products they’re looking for.
A visual search engine powered by AI allows users to upload images or take pictures of products to find them. Using shapes, colors, and details in an image, visual search engines suggest similar items. A simplified shopping experience is what young shoppers want. The technology was developed to meet those needs. Db
In my opinion, to boost the effectiveness of AI-driven retail models, researchers can certainly explore real-time sales data, customer feedback, and social media insights more dynamically. Continuous updates would help retailers in capturing evolving consumer preferences; considering the local nuances like specific events and holidays to enhance predictive accuracy. AI’s impact on retail is evident, and it is reshaping the inventory strategies for both online and in-store orders. With precise data, retailers can streamline their inventory management to reducing costs and cater to changing consumer demands. It seems fair that customers may prefer a focused selection over an overwhelming “endless aisle,” emphasizing the importance of targeted inventory placement over a variety of excess stock, and aligning with customer preferences for a satisfying shopping experience.
AI most certainly has the potential to improve forecasting models and inventory placements beyond what has been historically available to retailers and their brand partners. As such, it is an exciting use case for this much-discussed new technology that has practical real-world benefits to all parties, including consumers who benefit from a greater likelihood that a desired product is in stock when they want it.
Improved in-stock performance, and better inventory management overall, are just one of the potential benefits of new technology that will provide a measurable impact on the business and the customer experience, in the near term. Exciting stuff!
AI will be highly valuable at first in Mass Merchandisers and Supermarkets. I believe it will take awhile before it shows value and return on investment in fashion and clothing categories,because size/style/color is very difficult. Always has been, and I am tempted to type always will be, but we will have to wait to find out.
The thing about demand forecasts is that they are always approximations. When used to drive decision-making the best you can hope for is to narrow the error bars. AI can be a very useful ally in this.
In grocery, computer-automated ordering has been viable for nearly two decades. Industry-leading systems incorporate a limited form of machine learning and forecasting to make nightly re-ordering decisions across thousands of items to keep stock levels within bounds, and reduce the decision-making burden on humans.
In apparel, the forecasting challenges are different in nature, incorporating more seasonality and longer lead-times. Here again, decision making assistance from AI is a reasonable expectation, with the goals of reducing markdowns and clearances that sap profits at the end of the selling season.
When it comes to optimizing inventory levels in the retail environment, I prefer to think of AI as “assisted” intelligence rather than “artificial.”
Computer vision will give new insights into shopper behaviors in-store to optimize the front-of-house while providing much better visibility to issues and areas of optimization in the back-of-house. AI thrives on data. The more data (computer vision, customer data, supplier data) and the more accurate data (RFID for inventory, CDP management, etc.) are the keys.