How to Right-Size Inventory Without Increasing Stockout Risk
Inventory conversations get uncomfortable when finance and merchandising are looking at the same problem from opposite sides.
Finance sees too much cash sitting in inventory. Carrying costs are up. Turns are below target. Aged stock is building. The obvious request is to bring inventory down.
Merchants and planners hear something different: cut receipts, tighten safety stock, and accept more stockouts.
That trade-off is real, but it is often framed too broadly. A retailer can be overstocked at the company level while being badly understocked where demand actually exists. There might be plenty of units in the network and still no medium in black at the stores that sell it. Six weeks of supply overall can coexist with 12 WOS in weak locations and two WOS in the stores doing the volume.
That is not simply an inventory quantity problem. It is an inventory balance problem.
The safest way to right-size inventory is not to cut everything. It is to remove inventory that is doing little to protect profitable demand, then make sure the remaining stock is positioned where customers are most likely to buy it.
Stop Treating Right-Sizing as an Across-the-Board Inventory Cut
A company-level target like "reduce inventory by 10%" makes sense on a finance slide. Operationally, it is crude.
Take 10% out of a slow-moving fashion SKU with declining sell-through and you may reduce markdown exposure. Take the same 10% out of a never-out core item that replenishes every week and you may create lost sales almost immediately.
Those are not equivalent decisions.
A more useful definition of right-sized inventory is the minimum productive inventory required to support the service level the business actually needs.
That service-level piece matters. Oracle's retail inventory optimization framework explicitly treats inventory as a trade-off between service level and inventory cost, then translates those decisions into item-location replenishment policies. Higher availability has an inventory cost. The question is where paying that cost is justified.
Different SKUs deserve different service levels
Retailers sometimes talk about service level as if 98% is automatically better than 95%. It is not that simple.
A replenishable black T-shirt that drives repeat traffic probably deserves stronger availability protection than a seasonal fashion top approaching end-of-life. A high-margin core SKU with a reliable rate of sale should not be treated like a long-tail color with plenty of substitutes.
Service levels should reflect some combination of velocity, margin, demand variability, supplier lead time, lifecycle stage, substitutability and strategic importance.
That means some products should carry more safety stock. Others should carry considerably less.
Trying to provide near-perfect availability across every SKU-location combination is expensive. Oracle's optimization methodology makes the same point mathematically: changes in service-level targets can materially change the inventory investment required.
So before asking where inventory can be cut, decide where availability is actually worth protecting.
Find the Inventory Imbalance Hidden Inside Aggregate Numbers
Most inventory problems become clearer when you stop looking at the total.
Company WOS can be useful, but it hides a lot. Category WOS hides even more. The planner usually needs to get down to SKU-location and, in many categories, size-location.
Consider an apparel style showing five WOS at the chain level. That number looks fine. Then you open the size curve.
How to Right-Size Inventory Without Increasing Stockout Risk
XS and XXL have plenty of stock. Medium and large are nearly gone in the highest-volume stores. The style technically has inventory, but the size break means a meaningful portion of demand can no longer be served.
This is why right-sizing needs more than an inventory-value target. Planners should be looking at SKU-location WOS, rate of sale, sell-through versus plan, stock-to-sales, aged inventory, size and color availability, weeks remaining in the selling season and projected ending inventory.
One version of the truth matters here. If those measures are scattered across separate Excel files, planners spend too much of the week reconciling what happened and not enough time deciding what to do next.
A stockout does not always mean the retailer needs more inventory
Imagine a footwear retailer with a sneaker that keeps stocking out in sizes 9 and 10 at several high-volume stores.
The first reaction might be to increase the buy.
But suppose there are plenty of units nationally. Some are sitting in lower-volume stores. Others are concentrated in less productive sizes. A larger style-level PO could actually make the inventory position worse. It adds units without necessarily solving the size-location shortage.
The better question is: where are the units we already own?
Can stock move from slow stores? Is there inventory available at the DC? Should the next allocation favor stores with stronger demand? Is the size curve being allocated based on actual local selling patterns or a chain-average profile?
Connected inventory is valuable precisely because network-level supply can exist while individual nodes remain short. Visibility into what inventory exists and where it sits changes the decision from "buy more" to "position better."
There is a useful real-world proof point. McKinsey describes work with a national retailer of more than 5,000 stores where advanced demand planning and inventory analytics were used to reallocate stock toward higher-demand stores. The retailer achieved $1 billion in inventory savings while customer-service levels increased from 96% to 98%.
Less inventory and better availability can happen together when the starting problem is imbalance.
Protect Availability by Forecasting True Demand and Setting Smarter Safety Stock
You cannot safely reduce inventory without understanding what the buffer is protecting you from.
This is where historical sales can mislead planners.
Suppose a SKU sold 80 units last month. Was demand actually 80 units?
Maybe. Or maybe it was unavailable for six days. Perhaps the two core sizes were broken for half the month. Maybe stores had inventory in the system but not on the shelf. Recorded sales tell you what customers managed to buy. They do not always tell you what customers wanted to buy.
If stockouts repeatedly suppress sales and those sales feed directly into the next forecast, you create an ugly loop:
Lower availability produces lower recorded sales. Lower recorded sales produce a lower forecast. The lower forecast produces a smaller inventory target. Then you stock out again.
A useful demand view should account for baseline demand, recent trends, seasonality, promotions, lifecycle, local differences, demand volatility and periods where availability likely constrained sales.
This is also where forward-looking planning matters more than another spreadsheet reporting last week's misses. Planners need to see projected stockouts and overstock before they occur, while there is still time to change an allocation, transfer inventory or adjust a replenishment decision.
Replace blanket WOS buffers with risk-based inventory targets
A four-WOS rule everywhere is easy to understand. That does not make it good inventory policy.
A replenishable SKU selling at a stable rate with dependable weekly supply does not need the same protection as an item with erratic demand and an unreliable eight-week lead time.
Safety stock should respond to demand variability, lead-time variability, replenishment frequency, target service level, product importance and lifecycle stage.
Oracle's current retail guidance, for example, allows minimum safety-stock units or days to protect against volatile sales and shipping delays. It also describes raising service levels for key in-stock periods such as holidays and reducing them near end-of-life when carrying excess inventory becomes a greater concern.
That last point is worth emphasizing.
Right-sizing does not mean safety stock always goes down.
Some SKUs are carrying buffer they do not need. Some are under-buffered. If a core product has volatile demand, long replenishment lead times and a high cost of being unavailable, increasing its protection may be completely rational even while the company reduces total inventory.
The objective is not less safety stock. It is better safety stock.
Fix Inventory Accuracy and Rebalance Stock Before Cutting Future Buys
Forecasting gets plenty of attention because it is measurable. Execution problems can be less visible and just as damaging.
The system says a store has four units.
Physically, one was stolen, one is sitting in returns, another is somewhere in the backroom, and the last one may or may not be on the selling floor.
From the replenishment system's perspective, the store has stock. From the customer's perspective, it does not.
Phantom inventory, shrink, receiving errors, misplaced units and delayed adjustments all distort replenishment decisions. If the problem persists, planners often compensate by carrying more buffer.
Now the retailer has more inventory without reliably having better availability.
That is why cycle counts and exception-based inventory checks matter. Recurring stockouts where the system still shows on-hand units deserve investigation. So do negative positions, suspiciously static inventory and locations where recorded inventory consistently fails to convert into sales.
McKinsey includes inventory accuracy among the practical levers retailers can use to improve availability and inventory productivity.
Then comes the next question: can existing stock solve the problem before we buy anything else?
A useful decision hierarchy is:
Can existing stock satisfy demand? → Can it be transferred or reallocated? → Are replenishment parameters wrong? → Do we actually need additional inventory?
Oracle's inventory optimization approach includes both purchase-order recommendations and rebalancing transfers, specifically moving unproductive stock toward locations where it has a better chance of selling while considering transfer costs.
For seasonal merchandise, this can matter enormously.
If a product has eight selling weeks remaining and one store has excessive WOS while another is about to stock out, another supplier order may be the wrong answer. By the time it arrives, the full-price selling window could be nearly gone.
Sometimes the inventory you need is already yours. It is just in the wrong place.
Manage Right-Sizing as a Continuous Trade-Off Between Availability, Sell-Through and Markdown Risk
Inventory is not "right-sized" once.
Demand moves. Promotions change velocity. Lead times slip. Size curves develop differently from plan. A new color takes off while another stalls. Products that needed availability protection early in the season eventually become exit inventory.
The target has to move with the business.
A useful inventory-health cadence should look across availability, WOS, sell-through, forecast accuracy, turns, aged inventory, size availability, projected ending inventory and markdown exposure.
No single metric tells the whole story.
High turns can look excellent while customers repeatedly encounter broken size runs. High availability can look excellent while slow stores accumulate excess stock. Strong sell-through at style level can conceal weak colors that are heading toward markdown.
Markdown risk is especially important because excess inventory rarely disappears for free.
McKinsey has pointed out the weakness of one-size-fits-all markdown strategies. Healthy products can be discounted unnecessarily while genuinely weak SKUs do not get enough intervention to reach sell-through targets. The better approach looks at item performance, location and channel differences, forecasts and inventory position.
Inventory planning should work the same way.
This is where predictive inventory tools can earn their keep. At Flagship, the useful part of AI is not producing another black-box number for a planner to accept. It is continuously surfacing where future stockouts, excess WOS, broken size availability and ending-inventory risk are developing, then giving the planner enough context to make a better decision.
The planner still owns the trade-off. The system should make that trade-off visible earlier.
The question is not:
"How much inventory can we cut?"
It is:
"Where can we remove inventory without removing the inventory that protects profitable demand?"
The operating sequence is fairly straightforward:
Measure true demand → expose SKU-location imbalance → validate inventory accuracy → differentiate service levels → right-size safety stock → rebalance existing units → adjust future buys → monitor availability, sell-through and markdown risk.
Less Unproductive Inventory, Not Simply Less Inventory
Lean inventory and high availability are not opposites.
Poorly positioned inventory is the real problem.
Cut every category equally and you may free working capital for a quarter while creating size breaks, lost sales, emergency replenishment and frustrated store teams. Work at the SKU-location level and the picture changes. You can remove slow, misplaced and unnecessary inventory while protecting the units that actually generate sales.
The McKinsey example is a useful benchmark because it breaks the assumption that inventory reduction must damage service. Advanced demand planning, inventory analytics and stock reallocation helped a national retailer generate $1 billion in inventory savings while improving customer-service levels from 96% to 98%.
That is what right-sizing should aim for.
Not simply less inventory. Less unproductive inventory.