Is cycle counting better than an annual stocktake?

For accuracy, yes — errors are caught within weeks instead of compounding for a year, and counting in small batches without shutting down operations produces more careful counts. Many businesses still run a year-end full count for accounting purposes, but with good cycle counting it becomes a confirmation rather than a discovery exercise.

How often should each item be cycle counted?

Base it on ABC classification: count A items (your high-value core) monthly or at least quarterly, B items quarterly to twice a year, and C items once or twice a year. Also trigger an immediate count whenever a discrepancy signal appears — a pick short, a negative stock reading, or an unexplained channel mismatch.

Do I still need a year-end stocktake for HMRC?

You need an accurate stock valuation at your accounting year end. A full count is the traditional way to get it, but a well-run perpetual inventory system with documented cycle counts is widely accepted — speak to your accountant about what evidence they want. Many sellers keep a lightweight year-end count as a belt-and-braces check.

Cycle Counting vs Full Stocktake: Which Is Right for Your Business?

Q: What is cycle counting?

Cycle counting is counting a small portion of your inventory on a rotating schedule — a few SKUs or locations each week — instead of counting everything at once in an annual stocktake. Over a cycle, everything gets counted, with high-value fast-moving items counted more often than the slow tail.

The annual stocktake is where stock accuracy goes to be discovered dead. Errors made in February sit undetected until December, quietly causing oversells, phantom stock and bad reorders all year. Cycle counting — little and often, weighted toward the stock that matters — fixes the problem at its root.

The two approaches

A full stocktake counts everything at once — typically annually, often at the financial year end, usually with operations paused. It produces a complete snapshot, then goes stale from the moment it finishes.

Cycle counting counts a small slice of inventory on a rotating schedule — ten SKUs every Tuesday morning, a shelf bay a day, one category a week. Nothing shuts down; over the cycle everything gets counted; the stock that matters most gets counted most often.

	Full stocktake	Cycle counting
Frequency	Once or twice a year	Continuous — weekly/daily slices
Error detection lag	Up to 12 months	Weeks, days for A items
Operational disruption	Shutdown or overtime weekend	None — part of the routine
Count quality	Rushed, fatigued, error-prone	Small batches, careful counts
Root-cause finding	Impossible — too much time passed	Realistic — recent transactions traceable
Accounting year-end	The traditional evidence	Accepted with documented process (ask your accountant)

Why the lag is the killer

A stock record error is not a passive fact — it actively causes damage every day it exists:

Record says 5, shelf has 0 (phantom stock): every channel keeps selling it. Oversells, cancellations, defects, ranking damage — for months.
Record says 0, shelf has 5: the listing shows out of stock, the money sits invisible on the shelf, and you may reorder stock you already own.

Annual counting means each error does an average of six months of damage before discovery. Counting your important SKUs monthly caps that at two weeks. The value of counting isn't the number — it's the shortened life of every error. For multi-channel sellers the stakes are doubled, because one wrong number is broadcast to every marketplace you sell on.

Building the schedule: ABC-based cycle counting

Don't count everything equally — weight by value, exactly as in ABC analysis:

A items (top ~80% of value): count monthly, quarterly at minimum
B items: count quarterly to twice a year
C items: count annually — the year-end check covers them

Worked example: 400 SKUs, one person, 20 minutes a week

Say your catalogue splits 80 A / 120 B / 200 C:

A items monthly: 80 counts/month ≈ 20 per week
B items quarterly: 120 ÷ 13 weeks ≈ 9 per week
C items annually: 200 ÷ 52 ≈ 4 per week

Thirty-three SKU counts a week — with a barcode scanner and locations that are actually organised, that is 15–20 minutes every Tuesday morning. In exchange: no annual shutdown, and your best sellers are never more than a month from a verified count.

Count by location, not by list

Walking a shelf bay and counting everything in it beats hunting 33 SKUs scattered across the room. Structure the rotation by location where you can, and let the ABC weighting decide which locations rotate fastest — most sellers naturally shelve fast movers together anyway.

Exception counts: the queue-jumpers

The schedule handles routine drift. Some events warrant an immediate count of that SKU, same day:

A pick comes up short — the picker went for 3 and found 2
Stock goes negative in the system (always means an earlier error)
A customer reports a wrong or missing item
A channel mismatch appears — marketplace quantity disagrees with your system
After a return is processed — returns handling is a top source of record drift; our returns inventory guide covers why

Exception counting is the highest-value counting you will ever do, because the trigger has already told you where an error probably lives.

Counting while trading

The classic objection: how do you count stock that is being sold in real time? Practical answers:

Count at quiet hours — first thing in the morning, before marketplace dispatch cut-offs generate pick activity.
Count, then reconcile against transactions — if the system says 10, you count 9, and one sold while you were counting, there is no discrepancy. Software that timestamps counts against live transactions does this automatically.
Pause picks for the bay being counted, not the warehouse — a 20-minute local freeze nobody notices.

What to do with every discrepancy

A discrepancy is data, not just a correction:

Recount — a third of discrepancies are counting errors.
Adjust the record with a reason code (damage, theft, mis-pick, receiving error, unknown). The reason codes are the point — they tell you which process leaks.
Look for the root cause while the trail is warm — check recent orders, returns and transfers for that SKU. With monthly counting there are maybe 30 transactions to scan; with annual counting there are 400 and you won't bother.
Track your accuracy rate — counts-correct ÷ counts-done. Above 97% you can trust your system; below 90% your processes (receiving, picking, returns) need fixing before any clever forecasting is worth doing.

Keep the year-end count — but make it boring

Most accountants still want year-end stock evidence, and a full count remains the cleanest form of it (a documented perpetual-inventory process with cycle count records is increasingly accepted — ask yours). The difference cycle counting makes is that the year-end count becomes boring: instead of discovering a year of surprises, you confirm what the system already knew. That's the goal. Our stocktake guide covers running the full count efficiently when you do it.

Tooling

Cycle counting on paper works; cycle counting with a scanner sticks. Salync's stocktake mode lets you scan a barcode with your phone camera or a USB scanner, shows the expected quantity, records the count with an audit trail, and adjusts stock across every connected channel instantly — so a Tuesday-morning count directly protects your eBay and Shopify listings by lunchtime. Free up to 50 SKUs.

Frequently asked questions

What is cycle counting?

Counting a small, rotating slice of your inventory continuously — weighted so high-value fast movers are counted monthly and the slow tail annually — instead of counting everything once a year.

Is it better than an annual stocktake?

For accuracy, clearly: errors get caught in weeks, not months, and small careful counts beat one rushed marathon. Keep a year-end count for accounting evidence if your accountant wants one — cycle counting makes it a formality.

How often should each item be counted?

A items monthly, B items quarterly, C items annually — plus an immediate exception count whenever a discrepancy signal appears (short pick, negative stock, channel mismatch, processed return).

What accuracy should I aim for?

97%+ of counts matching the record. Below ~90%, fix the processes causing the drift — receiving, picking and returns — before investing in anything more sophisticated.