Inventory Shrinkage in Singapore Retail: How One SME Found 4.2% Before 11.11

You detect inventory shrinkage by cycle counting your fastest-moving SKUs against system records now, before peak season inflates both your stock value and your error rate — not by waiting for the annual stock-take in January. A Singapore specialty retailer we advised did exactly that in September last year: three weeks, 120 SKUs, no shop closure. The count found a 4.2% variance worth roughly S$88,000 annualised, and four consecutive year-end stock-takes had never surfaced it. The figures below are rounded and the client is anonymised, but the method is reproducible by any SME with a stock system and a barcode scanner.

What Does 4.2% Shrinkage Actually Cost a Singapore SME?

This business turns over about S$4.8 million in cost of goods annually, carries roughly S$860,000 of stock, and lists about 1,400 SKUs. The top 120 SKUs account for 71% of revenue and held S$610,000 of counted value at the time of the exercise.

The variance against system records on those 120 SKUs was S$25,600 — 4.2% of counted value. Extrapolated across the year and the wider catalogue, the leakage sat near S$88,000. On a business making single-digit net margins, that is not a rounding error; it is the difference between funding a hire and deferring one. It was also invisible in the P&L, because closing inventory was being valued off the same system records that were wrong.

Why Did the Annual Stock-Take Never Catch It?

Because an annual count tells you the size of the hole, not where it is. Three structural problems make the January stock-take a poor detection tool:

• It is too late to act on. A variance found in January covers twelve months of causes, including a peak season that ended six weeks earlier. Nobody remembers which pick went wrong in November.
• It is done under time pressure. Counts squeezed into a weekend, often by staff who do not normally handle stock, produce their own errors — which then get posted as adjustments and quietly absorb the real variance.
• The adjustment hides the evidence. Once the system is written down to match the count, the trail is gone. The retailer had been posting year-end adjustments of S$60,000–S$90,000 for three years and treating them as normal.

That last point is the one worth sitting with. A recurring write-down of similar size every year is not shrinkage being managed — it is shrinkage being budgeted for.

How Did They Find the Gap in Three Weeks?

The exercise was deliberately small. Counting 1,400 SKUs would have required closing; counting 120 did not.

1. Rank by movement, not by value. They pulled twelve months of sales lines and ranked SKUs by units moved. Shrinkage concentrates where handling is frequent, not where items are expensive.
2. Freeze the count window per SKU. Each SKU was counted at a fixed point with goods-in and goods-out paused for that bin only — about fifteen minutes each, spread across three weeks of normal trading.
3. Count blind. Counters were not shown the system quantity. This single change matters more than any tooling: when the expected number is on the sheet, people count towards it.
4. Classify every variance before adjusting anything. Each gap was traced back through goods-received notes, delivery orders and picking records before a single adjustment was posted.

Step four is where the value was. The count produced a number; the classification produced a fix list.

What Was Actually Causing the Loss?

Of the S$25,600 variance, the causes broke down roughly as follows:

• 58% — mis-picked bundles. Promotional bundles assembled by hand during high-volume periods were being shipped with the wrong variant. Customers who received a more expensive variant rarely complained, so the error only ever showed up as stock loss.
• 24% — unclaimed supplier short-shipments. Goods-received notes were being signed against the delivery order rather than checked against the purchase order. Short deliveries entered the system at ordered quantity and were never claimed.
• 11% — damage never written off. Damaged stock was set aside in a back room and forgotten rather than recorded.
• 7% — unexplained. Genuine theft or loss, and by far the smallest slice.

This ordering is common and counter-intuitive. Most owners assume shrinkage means theft and reach for cameras. In this case, 82% of the loss came from two process failures that cost nothing to fix — and a quarter of it was money the suppliers legitimately owed back.

What Changed, and What Should You Do Before 11.11?

Three changes went in before the November peak: scan-on-pick for all bundled orders, a three-way match between purchase order, delivery order and goods-received note at receiving, and a weekly rolling cycle count of the top 40 SKUs. The following quarter's variance on the same SKU set came in at 0.9%. Separately, S$9,400 in supplier credits was recovered on documented short-shipments from the prior six months.

If you sell physical goods and 11.11 is a material week for you, the sequence is straightforward. Rank your SKUs by movement this week. Blind-count the top 40 next week. Classify every variance before posting an adjustment. You will either confirm your records are sound — useful information heading into your heaviest trading period — or find a number that changes what you do in November. Either way, it is a far better use of three weeks in September than a weekend in January.

Frequently Asked Questions

How often should a Singapore SME cycle count?

Weekly for your A-class SKUs (roughly the top 10% by movement), monthly for B-class, and annually for the long tail. The point is frequency where handling is heaviest, not uniform coverage.

Does cycle counting replace the annual stock-take?

No. You still need a full count for statutory reporting and closing inventory valuation. Cycle counting changes what that count finds — a business counting continuously should reach year-end with a variance in the low single digits rather than a five-figure surprise.

What system do we need before this is worth doing?

A stock system that records movements per SKU and barcode scanning at pick and receive. If you are still tracking stock in a spreadsheet, fix that first — you cannot detect a variance against records that are manually maintained by the same people handling the goods.