Cycle counting vs the annual stocktake: a practical guide for UK warehouses
Why the once-a-year stocktake keeps failing multi-channel sellers, how ABC cycle counting works in practice, and a rollout plan that gets you to reliable stock accuracy without closing the warehouse.
The annual stocktake is one of those rituals that survives on habit. Close the warehouse for a day or two, count everything wall to wall, feed the corrections into the system, and enjoy roughly accurate stock figures — for about six weeks. Then picking errors, unrecorded damages, supplier short-shipments and mis-scans quietly drift the numbers again, and you spend the next ten months making decisions on data you know is wrong.
For a multi-channel seller the cost of that drift is specific and measurable: every phantom unit — stock the system thinks you have but you don't — is a future oversell. On a marketplace, an oversell isn't just a refund; it's a cancellation defect against your seller account, and defects compound into lower visibility and, eventually, account restrictions. Stock accuracy isn't a bookkeeping preference. On eBay, Amazon, TikTok Shop and Temu, it's account health.
Cycle counting is the alternative: instead of counting everything once a year, you count a small slice of the warehouse continuously — a few SKUs or locations per day, chosen deliberately — so the whole catalogue gets verified on a rolling schedule and the important SKUs get verified often.
Why wall-to-wall stocktakes underperform
- The accuracy decays immediately. A stocktake is a snapshot. The error sources it corrected — mispicks, unlogged breakage, receiving mistakes — resume the next morning.
- Counting under time pressure produces bad counts. A warehouse team racing to finish 4,000 SKUs before Monday makes exactly the kinds of errors the count was meant to find. Recount discrepancies on stocktake day are notoriously common.
- It closes the operation. For most sellers that means lost dispatch days, which in practice pushes the count to the quietest weekend of the year — usually eight-plus months away from peak, which is when accuracy matters most.
- It finds errors without explaining them. Twelve months of accumulated discrepancies arrive as one anonymous adjustment. You'll never trace which process caused what, so nothing gets fixed at the source.
None of this means skipping the year-end count if your accountant or auditor requires one. It means the annual count should be a financial formality that confirms what your system already knows — not the mechanism by which your system finds out.
How cycle counting works: ABC in practice
The standard approach is ABC classification, which ranks SKUs by how much their accuracy matters — typically by picking frequency or by stock value, and for marketplace sellers picking frequency is usually the right lens, because fast-moving SKUs are where oversells happen.
| Class | What it covers (typical) | Count frequency |
|---|---|---|
| A | Top ~20% of SKUs, ~80% of picks | Every 4–8 weeks |
| B | Next ~30% of SKUs | Quarterly |
| C | The slow-moving tail | Once or twice a year |
The daily workload is deliberately small. A 3,000-SKU catalogue on the schedule above works out to roughly 15–25 SKU counts per day — one person, well under an hour, without stopping dispatch. That's the entire point: accuracy becomes a small daily habit rather than an annual crisis.
Two refinements worth adopting early:
- Event-triggered counts. Count a location whenever something suspicious happens there: a picker reports a shortage, an oversell occurs, a return is restocked, or the system shows negative stock. These "exception counts" catch problems at the moment they're diagnosable.
- Zero-stock verification. When the system says a location just hit zero, verifying is nearly free — confirming an empty bin takes seconds and catches phantom stock before it can oversell.
Counting blind
One rule separates useful counts from theatre: the counter must not see the expected quantity before counting. Show someone "system says 47" and they will find 47. Blind counts — where the person scans the location, counts what's physically there and enters the number — are the only counts that reliably surface real discrepancies. If your stock control system supports a blind count mode with a barcode scanner, use it; the scan-first workflow also eliminates the "counted the right stock into the wrong SKU" class of error.
What to do with discrepancies
The count is the cheap part. The value is in the follow-up:
- Adjust promptly, with an audit trail. Every correction should record who counted, when, and the before/after quantity — so trust in the numbers is inspectable, not assumed.
- Track accuracy as a KPI. The standard measure is the percentage of counted locations that matched the system exactly. Below ~95%, stock-driven errors are costing you daily; well-run operations sustain 98%+.
- Chase root causes by pattern. Discrepancies concentrated in one product category suggest a receiving problem; concentrated in one aisle, a slotting or labelling problem; spread thinly everywhere, a picking discipline problem. Cycle counting turns stock errors from an anonymous annual lump into a diagnosable weekly signal — this is the single biggest advantage over the stocktake.
A four-week rollout
- Week 1 — classify. Rank SKUs by picks over the last 90 days, assign A/B/C, and generate the count calendar. If Q4 matters to you, weight the classification toward your expected peak sellers, not just the summer's.
- Week 2 — pilot on A items only. One counter, 30 minutes a day, blind counts. Expect the accuracy number to be uncomfortable at first; that's the baseline, not the verdict.
- Week 3 — add exception and zero-stock triggers. These generate few counts but catch the highest-value errors.
- Week 4 — full schedule, publish the KPI. Put the weekly accuracy percentage somewhere the whole team sees it. Accuracy improves fastest when pickers know it's being measured and can see the trend.
From then on it's maintenance: re-run the ABC classification quarterly (product velocity changes), and keep the January returns wave on an exception-count footing, because returned-stock restocking is one of the most error-prone flows in the building.
The short version
An annual stocktake tells you how wrong your stock was, once a year, after the damage is done. Cycle counting keeps you within a percent or two of reality all year, costs less labour in total, never closes the warehouse, and — because discrepancies surface within days of the process that caused them — actually fixes the causes. For a multi-channel seller whose marketplace accounts depend on not overselling, it isn't really optional.
MaxInvent tracks live stock across every connected channel with a full adjustment audit trail — see stock control, or try the oversell risk calculator to estimate what phantom stock is costing you today.