Cycle count vs physical count: what each catches, and what each costs

Eighteen hours and a lost trading day, twice a year. That's a wall-to-wall count of a 900-SKU store at a two-person pace of 30 SKUs an hour, plus setup and reconciliation. The same store running a 45-SKU cycle count every week spends 1.5 hours a week — about 75 hours a year — touches every SKU every 20 weeks, and never closes the doors. Both schedules are legitimate. The decision is which errors you need caught fast, and whether you can afford to pause — and most stores under $2M land on the hybrid: cycle counts monthly-or-better on the SKUs that pay the bills, plus one full count a year for the baseline.

Two definitions, thirty seconds each

A cycle count is a small subset of SKUs counted on a schedule — today the bestsellers, next week the watchlist — while the store keeps operating. The counted SKUs get verified; everything else waits its turn.

A physical count (wall-to-wall, full stocktake) is every SKU at a location counted in one event, usually with selling paused, typically around fiscal year-end. Nothing is trusted; everything is touched once.

They're not competitors so much as different instruments: one is a thermometer you use weekly, the other is the annual X-ray.

The comparison, side by side

Read the hours row honestly: cycle counting costs more total hours for the same coverage. What it buys with the extra hours is freshness — each error is caught days after it happens instead of months, while the receiving slip or the damage bin that caused it is still identifiable. The physical count buys completeness and a number your accountant, insurer, or lender will accept without footnotes.

What each catches that the other doesn't

The physical count sees aggregate shrinkage. Steady, systematic loss — a pocket here, a "borrowed" case there — spreads across many SKUs, and a rotating subset can miss it for months. Only a wall-to-wall adds up the whole floor at once and makes the total undeniable. It also sweeps the places cycle counts never visit: the returns bin, display units, the sample shelf, the box of "I'll deal with these later." And when the number has to be defensible — a lender package, an insurance proof-of-loss, the fiscal-year valuation — a dated full count is what carries it. The accounting side of that event — the year-end count as an accounting event — is its own subject.

The cycle count sees causes. A variance caught on Tuesday points at Monday's receiving or last week's damage log; the same variance found at a year-end count points at eleven months of possibilities, which is another way of saying none. Fast feedback is also what keeps reorder inputs trustworthy between annual resets — if your buying math multiplies phantom quantities for ten months a year, the year-end correction arrives ten months late.

Accuracy per hour: where the hours earn the most

An hour of counting is not fungible, and this is where tiering earns its keep. One hour, two destinations (illustrative numbers):

• Twenty bestseller SKUs — average retail $34, average on-hand 40 units. You find 3 variances averaging 5 units each: 15 units × $34 ≈ $510 estimated of misrecorded retail value, on SKUs where velocity will act on every phantom unit within days.
• The same hour on tail SKUs — average retail $11, slow movers. You find 1 variance of 2 units: ≈ $22 estimated.

Same hour, roughly 23× the error exposure covered. Your catalog's numbers will differ — these are illustrative — but the shape won't: the hour goes where velocity multiplies drift. Counting the tail has value (the annual event exists for exactly that), but spreading it evenly across the year is the most expensive way to buy the least accuracy.

Five decision rules

1. Never counted properly? Run one wall-to-wall first. Cycle counting on top of an unverified baseline just polishes guesses; the full count sets the zero point everything else measures against.
2. Peak season inside eight weeks? Tighten cycle counts on bestsellers and don't schedule a shutdown. Losing a trading day to a stocktake in your hottest month is a self-inflicted wound.
3. A lender, insurer, or your taxes need a dated, defensible number? Physical count, aligned to fiscal year-end. No accumulation of cycle counts substitutes for it in their eyes.
4. Shrinkage suspected but unproven? Physical count. Aggregate variances across many SKUs are the signature systematic loss leaves, and the rotating subset may never land on the right aisle. If the pattern points at a person rather than a process, the response side is a different playbook.
5. Solo operator under ~2,000 SKUs? The hybrid is the steady state: weekly or monthly cycles on the tiers that matter, one annual stocktake to reset and reconcile. Most small stores converge there eventually; the only real decision is which end to start from.

Running either in Shopify

Shopify's native counting doesn't force the choice. You create a stock count from Inventory > Stock counts by choosing a location and the products to include — two SKUs or two thousand — staff perform the count with a barcode scanner or by entering quantities manually, and a manager reviews and approves the results before your inventory levels change (Shopify Help Center, "Stock counts," accessed November 2026). At POS Pro locations, staff can perform counts right in the POS app. Approving a count applies the difference and writes each change into the adjustment history, so the audit trail survives the count. A partial count and a wall-to-wall are the same tool with different product selections — which means the hours-and-closure trade-off, not the software, is what actually decides your cadence. The step-by-step mechanics of running counts natively are covered in depth separately, and every variance a count surfaces deserves the investigation workflow for inventory discrepancies before it becomes a permanent shrug.

Common questions

Is a cycle count a full inventory count? No. A cycle count verifies a chosen subset of SKUs on a rotating schedule while trading continues; a physical count verifies every SKU at a location in one event, usually with selling paused. At a 45-SKUs-per-week pace on a 900-SKU store, one full pass takes about 20 weeks — completeness on demand is what only the physical count offers.

Which is more accurate, a cycle count or a physical count? They're accurate in different ways. A physical count is complete — nothing is left unverified — but its findings are months old by the time you reconcile. A recently counted SKU's cycle number is fresher and closer to the shelf. For a defensible valuation, physical wins; for keeping daily reorder math honest, cycles win.

How often do small retailers do a full physical count? Typically once a year, aligned to fiscal year-end, because the annual stocktake feeds accounting and tax valuation. Some add a second event before or after peak season. Between events, cycle counts carry the accuracy load — which is exactly the hybrid pattern most small stores settle into.

Does a cycle count require closing the store? No. Cycle counts run on a subset of SKUs while the store operates — count during a quiet hour or between picks, and native Shopify counts let you pick the products per session. Only wall-to-wall counts usually demand closed doors, because every unit at the location has to be touched at once.

Can I run both without doubling the work? Yes, and the two reinforce each other. The annual physical count sets the baseline; cycle counts maintain it week to week; the next physical count then validates the whole program — if your variance rate shrinks year over year, the cycle counting is working. The hours don't stack; they mostly trade against each other.

Next step

If you've never run a verified count, schedule one wall-to-wall — a quiet Sunday beats a lost trading day — and let it set your baseline. From the following week, start the rolling routine: your ten fastest SKUs every week, the middle tier monthly. How to size that cadence to the hours you actually have is its own decision, laid out in how often you should cycle count — and when a full count's variances don't add up to error, the investigation workflow will tell you what you're actually looking at.