Stop losing sales to empty shelves: find out-of-stock losses before the weekend does

Where lost sales from stock-outs actually come from

In multi-store retail, stock-outs are rarely caused by the company running out of goods. They are caused by how goods are spread across stores and how slowly that spread is corrected. For distributors, the same problem shows up as order lines shipped short while the item sits in another warehouse.

The most common mechanics look like this:

• Broken size curves: in apparel and footwear, M and L sell out while S and XL remain. The model looks in stock on the total, but the customer who needs an M leaves.
• Allocation by store size, not sell-through: opening quantities follow floor area or last year's turnover, so stores with a different customer mix start with the wrong depth.
• Fixed replenishment calendars: a store gets goods every Monday, even if its best seller sold out on Friday. The busiest days run on an empty shelf.
• Slow, informal inter-store transfers: arranged by phone between store managers, often after the demand peak has passed.
• Short-shipped wholesale orders: a line is cut in one warehouse while another holds enough units, and the customer fills the gap with a competing product.

How to estimate out-of-stock losses in your own business

A first estimate needs no data science team. The formula is: lost sales = days out of stock × normal daily sales when the item is in stock × average selling price. Sum it over store–SKU pairs, apply a conservative capture factor (some customers substitute or come back later), and multiply by gross margin for the profit impact.

One detail matters: measure normal daily sales only on days the item had stock. Averaging over all days lets zero-stock days drag the rate down and hide the problem.

Illustrative example (hypothetical, round numbers): a 20-store chain tracks its top 200 SKUs, or 4,000 store–SKU pairs. Say that on a typical day 200 of those pairs are at zero stock, each would normally sell one unit a day, and the average price is 500 TL. That is 200 × 1 × 500 = 100,000 TL of daily demand not served. Assume only half is truly lost: 50,000 TL a day, roughly 1.5 million TL a month. At a 40% gross margin, that is about 600,000 TL of gross profit per month. Your numbers will differ; the point is that this takes an afternoon with data you already have.

Why ERP reports and spreadsheets miss out-of-stock losses

Classic reporting is built on transactions, and a lost sale is a non-transaction. Several structural reasons keep the problem alive even in well-run companies:

• Censored demand: forecasts learn that an item sells less where it was often missing, so those stores get less stock and run out again.
• Totals hide gaps: a report by model, category or region shows healthy stock while specific sizes or stores are empty.
• Timing: weekly reports arrive after the weekend, when the sale is already gone.
• No owner: stock and sales live in the ERP, but transfer decisions live in phone calls and spreadsheets. Nobody owns today's gap.

How an AI-agent approach recovers lost sales: detect, diagnose, assign, measure

An AI agent turns the calculation above into a daily routine. It detects: on a scheduled run each morning, it compares every store–SKU stock level with recent sell-through and flags items that will run out before the next delivery, plus broken size curves. It diagnoses: is the cause allocation, a late shipment, unusually fast sales, or stock that exists only in the system? It assigns: instead of a report, it proposes a concrete action, such as moving six units of a size from a store with three weeks of cover to one with a day of cover. The regional manager approves or rejects it, and the store receives the task. It measures: it tracks how the transferred units sold and adds up the recovered gross margin.

The data needed is modest: daily sales by store and SKU, stock on hand, open transfers and purchase orders, the product master, and transfer lead times and costs. With Zzeti, these agents run on your own infrastructure, so sales and stock data stay inside your company. Ready MCP connectors read from Nebim V3, Logo or SAP, proposals land in the Actions Inbox for human approval, and Query Bench shows the SQL behind every number, so anyone can check why a transfer was suggested. Role-based access shows each store and region manager only their own scope.

A 4-week pilot plan for reducing stock-outs with inter-store transfers

Start small, with a scope where results are easy to see and compare.

• Week 1: pick one region of 5–10 stores and the top 100–200 SKUs by revenue. Connect sales and stock data and agree what counts as out of stock (zero, or below display minimum).
• Week 2: build the baseline: current stock-out rate and estimated lost sales, with spot checks of flagged items in stores.
• Week 3: start daily transfer proposals in the Actions Inbox, with rules for minimum quantity, maximum distance and a cover floor for the donor store.
• Week 4: compare pilot stores with similar non-pilot stores, review accepted and rejected proposals with reasons, and decide whether to extend.

KPIs to track for out-of-stock and lost sales

• On-shelf availability for top sellers: share of store–SKU pairs with stock, measured daily.
• Estimated lost sales in TL per week, using in-stock-day demand.
• Size-curve completeness for key models, by store.
• Transfer lead time from proposal to approval to shelf.
• Sell-through of transferred units within 14 days.
• Proposal acceptance rate and the most common rejection reasons.
• Transfer cost as a share of recovered gross margin.

You cannot fix lost sales you never measure. Estimate demand on in-stock days, find the stock sitting idle elsewhere, and make sure someone owns each transfer decision every morning.

Stop losing sales to empty shelves

Zzeti spots stock-outs days ahead at store × product level and drops a replenishment or transfer proposal into the right person's Actions Inbox for approval.