Warehouse Shrinkage: Where Your Stock Actually Goes

Shrinkage is the quietest line in an ecommerce profit and loss statement. It never appears as a cost. It appears as a stock adjustment, a write-off at year end, or an availability failure you blame on the forecast. Then you find you have been running at 2 percent of inventory value and the number has been eating a third of your gross margin.

What good looks like

India runs high. Retail shrinkage here has been measured above 3 percent of sales against a global average nearer 2 percent. Warehouse and ecommerce operations should be tighter than a retail floor because there is no open shopping and no customer hand touching stock. A well run own-warehouse D2C operation should hold shrinkage under 0.5 percent of inventory value per year. A disciplined 3PL runs under 0.2 percent. If you are above 1 percent, you probably do not have a theft problem yet. You have a process problem that will become a theft problem.

Convert it to rupees before anyone argues about it. On Rs 4 crore of annual cost of goods, 1 percent is Rs 4 lakh. That funds two more people on receiving and a scanner refresh, with change left over.

The five places stock disappears

Shrinkage is rarely theft first. It is usually a recording failure that later looks like theft.

• Receiving. The supplier short-ships two units in a 200 unit carton and you confirm against the invoice instead of counting. This is the largest source in most Indian operations and the cheapest to fix.
• Picking. Wrong variant picked for a lookalike SKU. The order ships, the customer keeps it, and two SKUs go wrong at once: one short, one long. Variant-heavy apparel and beauty catalogues suffer worst.
• Packing. Multi-unit orders packed short. It surfaces as a customer complaint and a replacement dispatch, with no stock correction recorded anywhere.
• Returns processing. The silent leak. Units come back, get opened, get graded loosely, and never get scanned into a disposition bucket. Sets get cannibalised for a missing component and neither parent nor child is written down.
• Damages. Crushed cartons parked on a pallet in a corner, still live in the system, discovered at the annual count as one ugly number nobody can explain.

Cycle counting that finds root cause

An annual wall-to-wall count tells you the size of the hole. It cannot tell you where the hole is, because by then the trail is cold and the people have moved on. Cycle counting by velocity can.

Run A items monthly, B items quarterly, C items twice a year. Count by location rather than by SKU, so you are auditing a physical place with an owner and a shift attached to it. Count blind: the counter does not see system quantity. A count where the number is visible is a confirmation exercise, not a count.

Then the part everyone skips. Any variance above 1 percent of that location value gets a root cause the same day, by pulling the last thirty days of transactions on that SKU and location. Nine times out of ten you land on a specific GRN, a specific picker or a specific returns batch. Fix that one thing. A cycle count programme without same-day root cause is just a slower annual count.

Track inventory record accuracy at location level as your headline metric. Median warehouses sit near 97 percent. Top quartile holds 99 percent or better. Below 95 percent, stop optimising anything else in the building.

The controls that pay for themselves

Ranked by return per rupee for an Indian ecommerce operation.

• Blind receiving counts against a sealed ASN, with variance raised to the supplier in writing within 24 hours. Costs nothing but discipline and recovers the most.
• Mandatory scan on every returns unit before it moves anywhere, with four fixed disposition codes and no free text. Reverse logistics without a scan gate is an unmetered outflow.
• A locked, camera-covered quarantine cage for damages, counted weekly. Damaged stock that lives loose on the floor eventually becomes somebody’s stock.
• Segregation of duties. The person who counts is not the person who adjusts. The person who adjusts is not the person who approves. This single control removes most internal theft opportunity in a small operation.
• Adjustment threshold approvals. Any single adjustment above a set value, say Rs 5,000, needs a second signature. Losses often hide as a long tail of small, legitimate-looking corrections.

CCTV matters less than people assume. Put cameras on the pack bench and the dock, where disputes happen, and review them only when a variance points there. Cameras nobody watches change nothing.

If you use a 3PL, read the liability clause before you read the rate card. Most cap shrinkage liability at a percentage of inventory value with a threshold below which they owe nothing at all. Negotiate the threshold down and the reporting cadence up. A monthly reconciled stock statement from the 3PL, matched line by line to your ERP, is worth more than any clause you will ever enforce.

None of this is exotic. It is the same five controls, run every week, by someone whose job it is.