Volatility Is Killing Your Inventory Strategy - How to fight back (1)

Forecasting isn't failing because your team is bad at it—it's failing because the world changed faster than the models. Consumer packaged goods (CPG) companies and retailers are juggling SKU proliferation, fragmented channels, whiplash demand shifts, and razor-thin margins. The result? Many legacy inventory planning systems, built for a slower and simpler era, are buckling under the pressure. Consider that the average grocery store carries over 40,000 SKUs today, up from just 7,000 in 1970, yet the methods for managing this complexity hardly evolved in decades. And the cost of misalignment is staggering: global inventory "distortion" from out-of-stocks and overstocks is projected to hit $1.77 trillion in 2023, with lost sales from empty shelves comprising about $1.2 trillion of that. In short, the old playbook of forecasting averages and managing by category is falling apart in the face of today's complexity and volatility.

What changed? For one, demand signals are more chaotic—trends can spark and fizzle in weeks, consumers bounce across online and offline channels, and a social media post can create a surge in one region and a slump in another. Supply chain disruptions and lead-time variability have also become the norm rather than the exception. Meanwhile, SKU complexity and rapid product turnover mean planners are often chasing a moving target with blunt tools. Traditional inventory management was about efficiency in a steady state. Today, it's about responsiveness in a constant state of flux. As one supply chain expert put it, improving forecast accuracy alone won't save you if you can't respond - agility and responsiveness are now just as critical. The mission has shifted: from minimizing cost variance to maximizing resilience and revenue capture. In this new world, a stockout isn't just a planning miss—it's a direct hit to the top line, and an overstock isn't just a holding cost—it's margin bleeding out on the balance sheet.

Five Below's Inventory Reinvention: A Case in Point

Five Below, a value retailer with trend-driven teen merchandise, rebuilt its inventory operating system to survive in a world of volatile demand. Over 1,800 stores now run on AI-generated, granular forecasts and automated reorders—ensuring hot items are in stock and shelves aren't cluttered with the wrong products. To see these challenges and solutions in action, look at Five Below, the high-growth value retailer catering to teens and pre-teens. With more than 1,800 stores across the U.S. and an ever-changing assortment of trendy, $1-to-$5 items, Five Below was living the inventory nightmare daily. They faced the classic bind: too many SKUs and not enough shelf space, wildly localized demand patterns, trends that evaporate before the next replenishment cycle, and planning processes that couldn't keep up. The pain was tangible on both ends of the spectrum—too much of the wrong stuff, and not enough of the right stuff. "We didn't just have a stock problem, we had a decision problem," one might say. Every day, margin was bleeding in two directions: markdowns on excess inventory that didn't sell, and missed sales from popular items out-of-stock. Five Below's leadership knew that simply pushing buyers or planners to "forecast better" wasn't going to cut it—they needed a fundamentally different approach.

The turnaround began by reframing inventory as a living system rather than a set of tables and reports. In 2025, Five Below partnered with a retail AI firm to essentially re-engineer their inventory brain. But the technology was only part of the story—it was how they used it to change decisions that made the difference. They blew up several long-held practices and replaced them with new ones:

• From hierarchy averages to hyper-local precision: Instead of forecasting sales at the category or regional level, they now generate probabilistic forecasts for every product in every store, every day. In other words, the unit of planning became "SKU x store x day," accounting for local demand quirks and real-time events. A fidget toy might sell out in one suburb while languishing in another—the new system sees that and adjusts. Five Below's AI platform churns through hundreds of variables (store traffic, school calendars, social media trends, weather, etc.) to predict demand more granularly than any human planner ever could.

• From static rules to adaptive autopilot: The company moved away from static min/max restock rules and periodic reviews to dynamic, automated replenishment triggers. The AI now continuously calculates optimal reorder points and quantities for each item-store, essentially dialing up the supply frequency when demand spikes and dialing it down when demand wanes. This automation ensures no store is left waiting on a head office meeting to react—the system adjusts daily. Five Below's team reported that the solution "automatically determines the right amount of inventory at the right time" for each store, maximizing availability while turning inventory faster.

• From gross volume to profit optimization: Perhaps the boldest shift was focusing decisions on profitability instead of just volumes or fill rates. Five Below's new platform doesn't just ask "will it sell?" but "where will it generate the most profit?". Using the granular forecasts, it allocates products to the locations that yield the highest incremental revenue and margin. One executive noted that inventory is only useful "if it's in the right place, at the right time, for the right reason." The old mindset was, "get it on the shelf and we'll sell what we can." The new mindset is, "if it's not earning its keep, rework it or cut it." By embedding a profit lens in replenishment, Five Below ensures limited shelf space is used for the most productive items—a crucial tweak when fads can flame out in weeks.

• From monthly hindsight to daily course-correction: Five Below also collapsed its decision cycle times. They went from monthly or quarterly planning meetings to daily and intraday analytics feedback loops. Inventory decisions are now reviewed in near-real-time—every day is "closing the loop" on yesterday's forecast vs actual and re-tuning accordingly. In practice, this meant planners and merchandisers trust the system to handle the routine flows, while they focus on exceptions and strategic buys. The AI "learns" with each cycle, continuously improving its recommendations as it ingests more data on what actually sold and where.

Crucially, Five Below did not treat this as a one-and-done IT project. They piloted the approach in just three months (90 days to stand up the AI forecasting across initial categories) and then rapidly scaled it to all product categories across all stores. Even after implementation, they set up an ongoing partnership model—a "hub" of data scientists and planners tweaking algorithms, exploring new data signals, and ensuring the system keeps learning as business conditions evolve. This underscores an important point: building a modern inventory capability isn't like installing software; it's like coaching a team—continuous and adaptive.

What was the impact? While Five Below hasn't published detailed numbers yet, leadership has cited major gains. Inventory per store dropped over 5% year-on-year even as sales rose, indicating leaner, more productive inventory. The company's Chief Strategy & Analytics Officer called the AI platform "a game-changer for our operations," noting it helped reduce stockouts and overstocking, while ensuring each location has the right products at the right time. In short, they could carry less inventory and still grow sales—a holy grail combination for retail. Fewer dead items clogging shelves, faster reaction to local demand surges, higher full-price sell-through, and ultimately more confidence to expand stores without burying themselves in complexity.

The Five Below story is not about one retailer's tech upgrade—it's a playbook for any CPG or retail operator drowning in complexity. They faced the same problems many do: hyper-local demand, too many SKUs, slow central planning, and an inventory system creaking under growth. By reimagining how decisions get made—more granular, more automated, and more profit-focused—they turned volatility from an enemy into an advantage. The hero here isn't the AI platform; it's the operators at Five Below who rewired their processes. And their battle cry was essentially: "inventory is only useful if it's right, not just there."

Five Myths Holding Back Inventory Planning (and New Rules to Replace Them)

Five Below's tale highlights a broader shift in mindset that leading CPGs and retailers are embracing. It turns out a lot of the old assumptions about inventory planning no longer hold true. Here are five inventory planning myths getting busted—and what forward-thinking teams are doing instead:

1. Myth: "Forecast accuracy is the ultimate goal."

   Reality: Responsiveness is the real goal. In volatile times, chasing a perfect forecast is like chasing a mirage. You'll never predict every twist—but you can build a supply chain that bends and flexes with each punch. As supply chain advisors note, focusing only on accuracy can mislead—a slightly wrong forecast with an agile response beats an accurate forecast with a rigid plan. The new KPI is not a single-number accuracy percentage; it's a combination of service level, reaction time, and resilience. Top operators are shifting investment into demand sensing and rapid replenishment capabilities. They accept that forecasts will be wrong often—the key is detecting the errors quickly and responding before customers feel it. For instance, Five Below's move to daily re-planning exemplifies responsiveness over static accuracy. In practice, this might mean smaller, more frequent orders, flexible contracts, and local buffering—all to respond fast when reality diverges from the plan. In short, don't let "forecast error" become a boogeyman—a decent forecast with superb agility beats a great forecast with no agility.

2. Myth: "More data equals better planning."

   Reality: Better data (and better use of it) equals better planning. We've all heard "big data" touted as the cure-all. Yes, today we can track everything from social sentiment to weather to Google Trends—but dumping terabytes into a silo doesn't automatically improve decisions. Often, too much data, poorly integrated, just creates noise. What matters is using the right data at the right time. A small but timely data point (like a sudden spike in searches for a product in one region) can beat a mountain of stale sales reports. Modern inventory systems prioritize data relevance, quality, and speed over sheer volume. They're also breaking down silos—connecting marketing data with supply chain data, for example—so that signals are interpreted in context. A hard lesson from recent years is that late data is almost as bad as no data. That's why "real-time demand sensing" is a cornerstone of the new approach—whether it's POS sales by the hour, online Browse behavior, or even local events, the winners are those who integrate these signals quickly into their inventory decisions. In short: a few critical, clean signals beat a hundred noisy ones.

3. Myth: "Replenishment is a back-office, supply chain task."

   Reality: Replenishment is a front-line profit lever. Traditionally, once forecasts were done, the act of replenishment (i.e. how much to ship to each store or region, when) was seen as an operational routine aimed at keeping shelves stocked. It was about efficiency and service levels. But in an era of constrained shelf space and vast assortment, replenishment choices—what goes where, in what quantity—are directly linked to profit and growth. Think of each store (or fulfillment center) as a mini P&L: sending Product A instead of Product B isn't just a stocking decision, it's a decision that affects revenue and margin. Five Below's system, for example, explicitly positions inventory where it will generate the highest incremental revenue. That's a profit-driven view: every slot, every dollar of inventory is an investment—is it yielding returns or could it be put to better use? Many CPG companies are adopting "profit-weighted allocation" strategies, which might mean preferentially allocating limited stock to channels or regions that drive higher contribution margin (e.g. e-commerce vs. wholesale, or vice versa, depending on economics). It also means treating stockouts as not just a service issue but a lost revenue event—one estimate finds AI-based planning can cut lost sales from stockouts by up to 65%, directly boosting the top line. The bottom line: replenishment is no longer just about logistics; it's a strategic lever to maximize ROI on inventory.

4. Myth: "If we plan it right, we won't need safety stock or buffers."

   Reality: Uncertainty is inevitable—bake flexibility into your system. Old-school planning sought to eliminate variability (with big, upfront forecasting efforts) and then drive lean execution. But as the last few years showed, shocks happen—pandemics, supply shortages, demand crazes from TikTok, etc. The new thinking is to embrace uncertainty rather than wish it away. This means designing inventory policies for resilience. Leading firms are increasing optionality in their supply chain: more flexible contracts, dual sourcing, on-demand production for certain items, and yes, strategic buffers where it makes sense. The key is dynamic buffering—for example, holding more safety stock for high-volatility items or in locations where supply is unreliable, but not a blanket one-size-fits-all cushion. Planners are also using probabilistic forecasts (like Five Below did) which provide a range of outcomes and confidence intervals, so they can make inventory decisions with risk in mind (e.g. "there's a 20% chance promo demand will 2x—do we want to cover that risk or accept some stockout?"). In short, today it's better to be roughly right and ready, than precisely wrong and stuck. Agility beats purity.

5. Myth: "Our legacy systems and org structure can handle it (with enough effort)."

   Reality: You likely need a new "Inventory OS" and a culture shift to match. Perhaps the biggest myth is believing you can muddle through today's challenges by simply working harder or adding spreadsheets on top of ERP systems. The leaders are realizing that incremental tweaks won't cut it—a fundamental reengineering is needed. This includes technology (moving to cloud-based, AI-driven planning platforms capable of crunching millions of combos), but also process (cross-functional collaboration, faster decision cycles) and people (upskilling teams to trust and leverage advanced analytics). A McKinsey study found companies using AI for supply chain planning reduced errors by 20-50%, and in retail, better AI forecasts cut warehousing costs 10-40% and lost sales 65%—but capturing that value required breaking down silos and changing how teams make decisions. It's telling that BCG's supply chain lead advised retailers to focus AI on just a couple of "quintessential" things like nailing the store-level product mix and put their best people on it, because "if you don't do that better than others, you're cooked." The new inventory OS isn't one system or one team; it's an integrated way of working that treats inventory as a continuous, company-wide decision loop. Leading CPGs are forming agile "control tower" teams that span demand planning, supply planning, and finance, all looking at the same real-time dashboards and scenarios. The organizations that succeed are those willing to overhaul processes that were built for a different era.

The Modern Inventory: Key Components for a New Era

How exactly are leading CPG and retail companies redesigning their inventory planning operating system? While approaches vary, a few common building blocks stand out in what we can call the Modern Inventory OS:

• Real-Time Demand Sensing: Forget waiting weeks for sales reports—the new systems ingest data in near real-time. Point-of-sale transactions, online search trends, social media sentiment, even foot traffic and weather data—all feed into demand signals. This "live pulse" of demand lets planners (or the AI) spot emerging trends or disruptions as they happen. For example, if a particular snack flavor suddenly spikes in one city due to a viral video, a demand-sensing system catches it within days (or hours) and can alert the team to reposition stock. Demand sensing bridges the gap between forecasting and execution, making planning a continuously updated activity. Some companies have invested in control towers that monitor key metrics live, so they can react to anomalies (a sudden drop in sales might indicate a stocking issue, an unexpected surge might trigger an express replenishment). The goal is early detection of both opportunities and risks.

• Probabilistic Forecasting: As mentioned, leading firms are shifting from point forecasts ("we expect 10,000 units next month") to probabilistic forecasts that account for uncertainty ("there's a 70% chance demand will be between 9,000-12,000 units, but with a heavy tail if trend X takes off"). By embracing uncertainty, planners can make smarter decisions on how much risk to take in inventory. Five Below's adoption of probabilistic, granular forecasts is a prime example. This approach often goes hand-in-hand with AI/ML models, which can analyze a far wider range of inputs and scenarios than traditional methods. The result is not necessarily a more accurate single number, but a better sense of the range of outcomes—which is invaluable for planning contingencies. Probabilistic forecasts also enable techniques like expected profit optimization (weighing the cost of overstock vs. cost of stockout to find the optimal inventory level under uncertainty). This is a game-changer for new products or promotional items where uncertainty is high by nature.

• Profit-Weighted Allocation and Assortment: Modern inventory systems recognize that not all products, customers, or channels are equal. They use algorithms to allocate inventory and space to maximize profitability rather than just to meet a forecast. This might mean prioritizing high-margin SKUs, or those that drive attachment sales, or ensuring key customers (say, a major retail partner or a VIP segment online) get priority on scarce items. It's a more nuanced approach than the old "fair share" allocation. Five Below's use of AI to decide where each product should go to generate the highest incremental revenue is a perfect illustration. Another aspect is assortment optimization at the store level—using analytics to tailor each store's SKU mix to local tastes and profitability. Gone are the days of one-size-fits-all planograms. If Store A can sell item X at full price while Store B would end up marking it down, the system might send more to A and less to B. These profit-weighted decisions require breaking organizational silos (merchandising and supply chain have to work hand in hand) and often need advanced tools to simulate outcomes, but they can significantly lift margin. Every SKU is an investment; the modern planner asks: what's the return on this, and can we do better?

• Automated Replenishment & Execution Logic: In the new model, a lot of the day-to-day ordering and restocking decisions are handed over to algorithms with human oversight. Automated replenishment systems calculate optimal order quantities and timing continuously, reacting to the latest data on sales, inventory levels, and supplier lead times. This doesn't mean the role of humans disappears; rather, their focus shifts to managing exceptions (e.g., supplier issues, events) and strategic parameter setting (like service level targets). The benefit of automation is consistency and speed—it can reorder fast and early enough to prevent stockouts, and also cut orders when something isn't selling to avoid overstock. Some CPG companies integrate their systems directly with supplier networks for automatic triggering of production or shipments when certain thresholds hit. The inventory moves become more "pull" than "push"—triggered by real demand, not just forecasts. A side effect: teams save countless hours formerly spent on manual data crunching and can reinvest that time in analysis and improvement (one study noted AI forecasting can free up 30-50% of planners' time from manual tasks).

• Embedded Analytics and Continuous Learning: The modern inventory OS is not a set-and-forget tool; it's continually learning from outcomes. Companies are embedding analytics that measure forecast vs. actual, test various strategies (A/B testing in supply chain, imagine that!), and provide feedback to refine the models. It's a virtuous cycle: more data → better forecasts → better decisions → new data from those decisions → even better forecasts, and so on. For example, if the system over-forecasted a SKU for three weeks in a row, it flags it and auto-adjusts the model weighting. If a "what-if" analysis shows that a 5% increase in safety stock on a certain category would have prevented lost sales worth $500k last quarter, that insight is fed upstream into policy setting. The best teams also loop in human learning—they conduct post-mortems on big misses (e.g., why did we stock out of Product Z? Was it a data blind spot or an execution lag?) and then update processes accordingly. Over time, this continuous improvement approach turns inventory management into a competitive asset that's hard for others to replicate, because you're essentially building an organizational memory and intelligence around demand-supply management.

Think of this modern inventory OS as akin to upgrading from a static GPS to real-time Google Maps/Waze for your supply chain. The static GPS (old forecasts and set rules) might give you one route at the start of the day. But the modern system is constantly re-routing based on live traffic, road closures, and the driver's preferences (profit goals). It doesn't mean there are no jams, but you navigate through them much more effectively.

What to Do Monday Morning: A Playbook for Operators

All this talk of AI, probabilistic models, and reinvention can feel daunting. But you don't need a multi-million dollar project to start moving from chaos to clarity. Here's a straightforward Monday-morning playbook to get momentum:

• 1. Put Your Current System Under a Microscope: Take stock of where your current forecasting and inventory processes are weakest. Where do you see frequent stockouts or excess? What products or channels feel "unpredictable" and cause fire-drills? Make a short list. For each, ask "Is our forecast granularity and frequency sufficient here?" and "How quickly do we adjust when we're wrong?". For example, if you realize you're forecasting at a monthly level but sales vary wildly week to week, that's a red flag that you need more granular, frequent planning in that area. This is the start of your roadmap.

• 2. Audit Decision-Making Cadence: Identify any decisions in your inventory flow that are happening in big, infrequent batches. Do you hold giant monthly allocation meetings? Review safety stocks quarterly? Those are candidates to break into smaller, more frequent decisions. Ask your team, "What would it take for us to adjust XYZ daily or weekly instead of monthly?" Perhaps it's too time-consuming today—which hints that automation or better tools are needed. You don't have to flip a switch overnight, but even piloting a faster cycle (say, weekly plan updates for a pilot category) can reveal huge benefits in service and efficiency.

• 3. Find the Silos and Bust Them: Volatility exploits the gaps between functions—maybe sales knows something that supply chain doesn't, or e-commerce and store teams aren't syncing up on inventory. Pick one or two big disconnects in planning (for instance, promotion planning vs. inventory planning often misaligned, or new product launch forecasts vs. replenishment). Bring the stakeholders together and map out how information flows (or doesn't). Often, you'll find simple fixes: a shared dashboard, a joint weekly check-in, or co-locating a demand planner with the marketing team can improve the information flow. The goal is to ensure everyone is looking at the same real-time truth and working off the same assumptions. When a plan changes, who needs to know today? Make that lightning-fast and automatic.

• 4. Leverage Quick Wins with Existing Data: You likely have more data than you realize that can improve planning right now. For example, do you truly use your point-of-sale data in forecasting, or are you averaging at a high level? What about external data—have you looked at correlations like weather, Google Trends, or social media buzz for your key products? Get your analytics team (or a savvy analyst) to run a quick study: could any readily available data improve our forecast or alerts? Perhaps you'll find that a surge in online searches is a 2-week leading indicator for demand—that's something you can start monitoring immediately, even manually, to adjust orders. Quick win: set up a simple alert (even a Google Alert or social listening) for spikes in interest in your product/category, and feed that to planners. This doesn't require new systems, just curiosity and initiative.

• 5. Talk to the Front Lines: Your store managers, sales reps, or supply planners likely know where the bodies are buried. Ask them: "Where do you see us consistently overstocked or understocked? What decisions feel too slow or out of touch?" These anecdotes can pinpoint systemic issues. Maybe a planner will tell you, "We always run out of size Medium in Region X because the plan doesn't account for the demographic difference." Aha—that's a clue to incorporate new data (demographics by store) or to adjust the logic for that region. Front-line wisdom can also help you build the case for change. When you go to senior leadership or budgeting, it's powerful to say, "Our Southern sales director estimates we lost 5% of sales last quarter due to stockouts on just three items. An investment in demand sensing there could pay back in months." Tie pain points to dollars—it gets attention. As one CEO famously noted about AI in retail, if you don't get inventory right at the local level, "you're cooked"—so gather the evidence that shows where you're not getting it right, yet.

• 6. Start a Pilot (Think Big, Start Small, Move Fast): Identify one category, brand, or region where you can pilot a more advanced inventory approach. Maybe it's implementing a new forecasting tool that uses machine learning, or as simple as testing a more granular spreadsheet model by store. Set clear metrics (e.g. stockouts%, inventory turns, forecast error) and run the pilot for a few cycles. For instance, Five Below piloted their AI system in a few categories over 90 days—you can mimic that spirit even with low-tech experiments. The key is to create a sandbox where the team can try new methods without waiting for a company-wide rollout. Once you see results—say, the pilot stores cut stockouts by half—you'll have the momentum and proof to scale it up. And in scaling, consider a partner if needed: Five Below didn't build everything in-house; they teamed with specialists. Today there's a rich ecosystem of tech vendors, consultants, and even universities that can help jump-start AI and advanced analytics in inventory. Just remember to focus on the decision change, not the shiny tech.

By Monday afternoon, you might not have an AI brain running your inventory (and you probably don't need one immediately). But you can have taken concrete steps toward more clarity. Audit, align, and pilot—those are your initial moves. The beautiful thing about inventory improvements is they often self-fund: reduce a bit of overstock, prevent a few lost sales, and you've freed up cash to reinvest in better capabilities.

Final Thoughts: Clarity in the Fog

Volatility isn't going away—if anything, the pace of change in consumer demand and supply dynamics will keep accelerating. But as Five Below's story shows, you don't have to be a victim of the chaos. The winners in CPG and retail are those who turn unpredictability into competitive advantage. They do it by reengineering how decisions are made, leveraging new tech as an enabler but, above all, adopting a new mindset: flexible, granular, and focused on value. It's not about "adopting AI" as a slogan or buying a fancy new dashboard. It's about rolling up your sleeves and fixing the broken links in your inventory logic—sometimes with advanced tools, and sometimes with just common-sense changes. The hero of this transformation is you, the operator, planner, or supply chain leader who refuses to accept that "it's always been done this way." As you push for clarity, remember that every big success starts with small steps. Today's chaos can be tamed—one SKU, one store, one decision at a time—until your inventory shines as a source of agility and profit, not anxiety. In the words of Five Below's team (and echoed by many others on this journey): inventory isn't just about stocking products, it's about positioning possibilities—the right product, at the right place, at the right time, for the right reason. Get that right, and volatility turns from a headwind into your wind at the back.

Sources:

1. SKU Rationalization - What Is It & How Is It Being Used https://www.specright.com/blog/sku-rationalization-what-is-it-and-how-is-it-being-used

2. Why Inventory Distortion Costs Retailers Trillions - The Food Institute https://foodinstitute.com/focus/why-inventory-distortion-costs-retailers-trillions/

3. 5 Myths About Forecast Accuracy https://www.bciglobal.com/5-myths-about-forecast-accuracy

4. Five Below deploys invent.ai's platform for forecasting, inventory optimization and replenishment https://www.invent.ai/news/five-below-deploys-invent-ai-platform-for-forecasting-inventory-optimization-and-replenishment

5. Five Below Partners with Invent.ai to Roll Out AI-Driven Inventory Optimization - The Toy Book https://toybook.com/five-below-invent-ai-news/

6. Five Below taps Invent.ai platform for inventory optimization | Retail Customer Experience https://www.retailcustomerexperience.com/news/five-below-taps-inventai-platform-for-inventory-optimization/

7. Five Below looks to tech, analytics to optimize inventory | Supply Chain Dive https://www.supplychaindive.com/news/five-below-analytics-inventory-India-sourcing-earnings/701893/

8. An introduction to AI demand forecasting - Peak https://peak.ai/hub/blog/ai-demand-forecasting/

9. The AI Mistake Companies Are Making, According to a BCG Senior Partner - Business Insider https://www.businessinsider.com/ai-mistake-companies-make-bcg-tech-executive-2025-5

10. Citi Trends partners with invent.ai to drive profitability through AI-powered inventory optimization https://www.invent.ai/news/citi-trends-partners-with-invent-ai-to-drive-profitability-through-ai-powered-inventory-optimization

11. Asos taps Celonis tech as online retailer looks to unlock full potential of its supply chain - Retail Technology Innovation Hub https://retailtechinnovationhub.com/home/2025/6/23/asos-taps-celonis-tech-as-online-retailer-looks-to-unlock-full-potential-of-its-supply-chain