The Dead Stock Problem: How Averaging Demand Costs Manufacturers 20% of Revenue
Stocking to the average across a multi-site network guarantees simultaneous stock-outs and dead stock. Classical forecasting cannot fix it. Placement modeling can.
The Dead Stock Problem: How Averaging Demand Costs Manufacturers 20% of Revenue
Dead stock is the mirror image of a stock-out. Both are caused by the same mistake: stocking to the average.
The average demand trap
Every supply chain leader we have worked with has said some version of this out loud:
"We know roughly what each site uses, so we stock to the average. Then half our locations run short while the other half sits on stock nobody will ever use."
The quote is uncomfortable because it is correct. Stocking to the average across a multi-site network guarantees that roughly half the locations will be under-stocked and roughly half will accumulate dead stock. The average is a summary statistic, not a stocking policy.
How averaging creates stock-outs AND dead stock at the same time
Consider a manufacturer with 50 locations serving different customer mixes. Location A serves a high-volume trauma center. Location B serves an outpatient clinic. If both are stocked at the network average for a given SKU:
- Location A stocks out repeatedly. Sales are lost or backordered.
- Location B accumulates units that reach end-of-life on the shelf.
Both problems get booked separately. The stock-outs show up as lost revenue. The dead stock shows up as write-offs. Nobody connects the two because they live in different reports owned by different teams.
Why classical forecasting does not fix this
Classical forecasting improves the accuracy of the average. It does not change the fact that the average is the wrong stocking target for any individual location. A perfect network-level forecast paired with average-based placement will still produce simultaneous stock-outs and dead stock.
The fix is not a better forecast. The fix is location-level placement modeling that respects the demand distribution at each site.
The placement modeling alternative
Placement modeling asks a different question. Instead of "how much will we sell," it asks "where should each unit sit to maximize the probability of consumption before it dies."
The inputs are:
- Location-level consumption history
- SKU shelf life and end-of-life pattern
- Transfer cost between locations
- Service-level target per customer tier
The output is a per-location, per-SKU stocking policy with a confidence band. Monte Carlo simulation is used because deterministic optimization cannot represent the variance that causes the problem in the first place.
Sample math on a 50-location network
Assume a SKU with network demand of 1,000 units per month, average location demand of 20 units, and standard deviation of 12 units across locations.
| Policy | Stock-outs per month | Dead units per quarter |
|---|---|---|
| Stock to average (20 per location) | ~18 events | ~180 units |
| Placement model (variable per location) | ~4 events | ~35 units |
The placement model does not require more total inventory. It requires the same inventory in different places.
Tying back to a real engagement
The US medical device manufacturer described in our inventory case study had a department carrying 20 percent of total company revenue. The initial assessment quantified $19M in annual loss driven primarily by the average-demand trap operating at scale across a national network. A Colorado pilot proved the placement model. National US rollout followed. Within six months, $4M was returned to profit.
Frequently asked questions
Is dead stock the same as slow-moving inventory?
No. Slow-moving inventory still turns eventually. Dead stock will reach end-of-life or obsolescence before consumption and has to be written off or redeployed.
Can better demand forecasting eliminate dead stock?
Partially. Better forecasts reduce the size of the error but do not fix location-level mismatch. Placement modeling is required to actually redistribute stock.
Do we need new inventory management software?
Usually not. Placement modeling produces stocking policies that existing systems can execute. The modeling layer sits on top of the current stack.
How is Monte Carlo simulation different from a standard optimizer?
Standard optimizers assume known demand. Monte Carlo simulates thousands of demand scenarios per location and finds the policy that performs well across the distribution, not just at the mean.
How DATA4AI helps: We quantify the dead stock pool, build the placement model, and pilot it on a single region before national rollout. See our manufacturing practice and inventory optimization case study, or book a discovery call.
Keep reading.
How to Eliminate Distributor Chargebacks in Consigned Medical Device Inventory
Distributor chargebacks quietly leak 3 to 8 percent of revenue in medical device manufacturing. Here are the three root causes and how a unified consumption, billing, and contract platform catches them pre-issue.
Inventory Shrinkage in Multi-Site Consigned Stock: The Five Signals We Look For
Consigned inventory shrinkage is harder to see than owned-inventory shrinkage. The signals live in the data. Here are the five we look for in every first assessment.
From Quarterly to Weekly Demand Forecasting on Microsoft Fabric: A Practitioner Guide
Planners should not be a data pipeline. A two-phase approach returns planner hours in Phase 1 and delivers a consolidated Microsoft Fabric platform with best-fit models and a GenAI layer in Phase 2.
Let’s talk about the second pilot.
Working on a healthcare AI, manufacturing analytics, EHR integration, inventory placement, or demand forecasting project? Book a 20-minute discovery call. We will tell you honestly if we are the right fit.