Stockouts Hide Demand: A Reporting Method for Distributor Replenishment Reviews

When a quiet sales report describes an empty shelf

A distributor reviewing replenishment may reduce a product's target because recent shipments have fallen. That conclusion is risky when the product was unavailable during part of the period. The shipment history records what the business could fulfill. It may omit customers who accepted a substitute, bought elsewhere, delayed their request, or never submitted an order after seeing an unavailable status.

The resulting problem is called censored demand: the observed outcome does not reveal the full quantity customers would have requested under different availability. A stockout can therefore create a feedback loop in which low shipments suggest low demand, lower replenishment follows, and availability remains poor. The existence and size of that loop must be investigated locally; it cannot be inferred from one low-sales week.

This article proposes an evidence layer for a distributor's replenishment review. It does not calculate an optimal reorder point or promise a forecasting improvement. The decision is whether a sales series is sufficiently exposed to stockouts that planners should examine availability and unmet-request evidence before using it as a demand proxy.

What the research supports

Conlon and Mortimer's NBER paper, Demand Estimation Under Incomplete Product Availability, studies how availability information affects demand estimation. Its empirical setting is vending machines, not wholesale distribution. The useful general lesson is that sales observed under changing availability require careful interpretation; the study's measured effects should not be copied as a distributor's expected result.

The Operations Research paper on demand estimation under censoring examines the informational value of stockout timing under stated model assumptions. That supports collecting timing evidence rather than merely an end-of-day inventory balance. It does not establish that a particular mathematical model applies to every distributor, intermittent-demand item, or customer contract.

Our proposed register and examples are an operational adaptation of these research questions. No private sales data were analyzed. All numbers are hypothetical, and no claimed lost-sales amount should be entered into actual demand history without a separate, clearly labeled estimation process approved by the planning owner.

Separate requests, orders, shipments, and availability

Define the series under review. Customer inquiries, accepted order lines, promised quantities, shipped quantities, and invoiced sales describe different stages. A distributor that accepts backorders may preserve some unmet demand in its order book. Another that rejects unavailable products at entry may record no order at all. Comparing their shipment histories without that difference can be misleading.

Availability also has multiple meanings. Physical stock in a building may be reserved, on quality hold, inaccessible to the relevant channel, or unavailable in the customer's required pack size. For the chosen decision, record whether the customer could actually place an eligible order and whether the business could fulfill it under its approved promise rule.

Keep stockout start and end events where reliable. An item with positive opening and closing balances can still be unavailable during a busy interval. Conversely, an item with zero closing stock may have fulfilled all requests that day. End-of-day inventory alone cannot distinguish those patterns. Report uncertainty where the system lacks intermediate observations.

Build a register that preserves evidence and uncertainty

Use a stable item, location, channel, and time interval as the reporting unit. Add eligible operating time, observed unavailable intervals, shipped units, accepted backorders, cancelled requests, substitution evidence, and known data gaps. Preserve the source definitions so a later analyst knows whether unavailable means zero physical stock or a channel-specific selling restriction.

Keep recorded facts separate from derived estimates. A customer note explicitly requesting ten units is evidence of a request, subject to deduplication and order context. A planner's estimate that twenty additional units would have sold is an analytical assumption. Combining both into a single apparently observed number makes the review difficult to reproduce.

A small example shows why ranking can reverse

Consider two hypothetical products over a ten-day review. Product A ships 100 units and is available throughout. Product B ships 70 units but is unavailable for four of the ten days. Ranking only by shipments puts A ahead of B. Dividing B's shipments by its six available days produces about 11.7 units per available day, compared with ten for A.

That calculation raises a question; it does not prove B would have sold 117 units over ten fully available days. The available days may include different weekdays, customer ordering cycles, promotions, or unusually large orders. Stockouts may themselves occur during the busiest periods, making missing demand especially difficult to recover from an average.

A responsible report presents the observed shipments, the availability exposure, and a clearly labeled scenario. It might show what constant available-day demand would imply, then explain why that assumption could fail. The planner can use the discrepancy to prioritize investigation without treating the scenario as a confirmed sales total.

Now suppose 15 units appear as backorders and are shipped after replenishment. Counting both the original unmet request and the later shipment as separate new demand would inflate the series. Link them to the same customer requirement. Timing measures may assign the request to its original date and fulfillment to its later date, but the quantity should not become two independent customer needs.

Handle substitutions and cancellations carefully

Substitution can make the unavailable item appear weak and the replacement item appear strong. Capture an explicit original-item reference where the approved ordering process supports it. Without that reference, do not infer every increase in a neighboring product's sales is caused by the stockout. Price, promotion, seasonality, and changes in customer mix remain possible explanations.

A cancellation reason is also imperfect evidence. A customer may select a broad reason code after several issues, or an administrator may choose the nearest available option. Review a sample against original communications under the organization's privacy rules. Report unknown reasons separately instead of assigning them all to availability.

Preserve repeated inquiries without automatically counting each as a new requirement. One buyer contacting several channels about the same order can create several records. Deduplicate through approved customer and order references where possible, and retain unresolved matches. The report should expose both overcounting and undercounting risks.

Test the method before changing purchasing parameters

Begin with a small group of items whose inventory events and order history can be reconciled. Include a continuously available item, an item with a clear stockout, a backordered item, and a product with substitutions. Ask the planner and operations owner to agree on event meanings and exclusions before reviewing performance rankings.

Compare the report with source transactions for selected intervals. Check whether availability restrictions actually affected the channel under study, whether replenishment reopened ordering when expected, and whether manual orders bypassed the restriction. A precise duration calculated from the wrong event can be more misleading than an explicitly incomplete estimate.

Separate the reporting pilot from the purchasing decision. A finding that shipment history is censored supports collecting better evidence or evaluating an appropriate forecasting model. It does not by itself justify a larger buy. Lead times, service objectives, perishability, substitution, and capital constraints still matter to the authorized planner.

A practical handoff for distribution teams

An administrative support role can maintain availability exceptions, connect backorders to later shipments, and prepare a review packet through distribution reporting support. Inventory records support can help keep the underlying item and restriction records consistent within documented permissions. Neither function should silently overwrite observed demand with a forecast.

The review packet should state the population, interval, source completeness, observed quantities, estimated scenarios, and unresolved questions. Retain the original report when better availability evidence arrives so revisions are explainable. The useful conclusion may simply be that a particular product's shipment history is not a reliable standalone measure of demand.

Where evidence remains incomplete, say so. A well-supported uncertainty is more useful than an invented lost-sales figure. By preserving customer requests, availability exposure, and fulfillment as separate records, a distributor gives its planning team a stronger basis for deciding which items need deeper analysis and which apparent demand declines are credible.

Sources and review date

• Demand Estimation Under Incomplete Product Availability (opens in a new tab). National Bureau of Economic Research. Checked 2026-10-12.
• Demand Estimation and Ordering Under Censoring: Stock-Out Timing Is (Almost) All You Need (opens in a new tab). Operations Research; author copy hosted by National University of Singapore. Checked 2026-10-12.