The cycle count playbook: weekly vs monthly vs continuous (a decision matrix)

The first annual stocktake I ever ran ended at 4am on a Sunday with two pickers crying and a finance director who hadn't slept. We'd shut the floor for 36 hours. We had eight people walking aisles with clipboards. The variance on the final reconciliation was 11% by value — enough to materially restate the year-end accounts. Nobody could tell me where it had come from. Was it shrinkage? Mis-picks? Pallets we'd booked in three months ago and never put away properly? All of the above, probably. The honest answer was: we had no idea, because the last time we'd counted anything was 11 months earlier and the trail had gone cold.

That stocktake was the last one we ever did. Two months later we'd binned the annual count and were running cycle counts five days a week. Our reported accuracy was at 99.4% inside a quarter, and the auditor — who had been visibly worried about us — signed off with no qualifications the following year.

This piece isn't another general "you should cycle count" article. There are enough of those. The piece most guides miss is the one operators actually need: which cadence to use for your warehouse. Weekly, monthly, or continuous ABC-tiered. They're not interchangeable, and picking the wrong one is how cycle count programmes quietly die six months in.

Why cycle counts beat annual stocktakes

The annual stocktake is a financial control disguised as an operational one. Your auditor wants a number at year-end. They don't care how you got it. The problem is the warehouse pays the price for that financial control — a shut floor, overtime, exhausted staff, and a single bulk number that tells you the cumulative drift over twelve months without telling you a single thing about why.

Cycle counts flip the trade. You count a small slice of the warehouse every day, every week, or continuously. Three things follow from that:

1. Variances are catchable. A 2-pallet discrepancy found within five days of receipt can usually be reconstructed from CCTV, receipts, and pick history. The same variance found 11 months later is just a write-off.
2. The floor never stops. Counting half a zone at 6am on a Tuesday costs you almost nothing operationally. Stopping the whole site for a weekend costs you a week of throughput.
3. You get a real accuracy signal. Not a year-end snapshot but a rolling KPI you can manage to.

The catch — and this is the bit cycle count consultants gloss over — is that running cycle counts properly is harder than running an annual count. An annual count is brute force. Cycle counting is a discipline. If you pick the wrong cadence, you either burn out your team (too aggressive) or kid yourself about your accuracy (too lax).

The three cadence options

Weekly

Every SKU touched at least once a week. Usually run as small rolling counts — a zone or aisle per day, hitting the whole warehouse in five working days.

Best for: small warehouses (under 500 SKUs), high-value/high-shrinkage operations (electronics, alcohol, pharma), or any site recovering from a known accuracy crisis.

Watch out for: count fatigue. Walking the same aisles every five days gets old fast, and counters start "remembering" what was there last week instead of counting what's there today. Weekly cadence needs the strictest discipline on who counts what (more on that below).

Monthly

Every SKU counted at least once a month, usually broken into daily slots — count roughly 1/22nd of the warehouse each working day. This is the safe default and the one most 3PLs land on.

Best for: mid-size operations (500–5,000 SKUs), mixed-velocity inventory, contract warehouses with standard customer audit requirements.

Watch out for: cadence drift. "Monthly" sounds gentle, so when a busy week hits, the counts get skipped "just this once". Six weeks later you've got a backlog and the programme's effectively annual again. Monthly cadence dies from neglect more often than from any operational problem.

Continuous (ABC-tiered)

Not every SKU on the same cadence. Instead, SKUs are bucketed by value or velocity and each bucket gets its own frequency.

A typical tier scheme:

• A items (top 20% by value or velocity): counted every 30 days.
• B items (next 30%): counted every 60–90 days.
• C items (bottom 50%): counted every 180 days.

Counts run continuously through the working day — the system schedules them, the floor executes them in pockets of dead time, and the dashboard reports a rolling accuracy figure.

Best for: large warehouses (5,000+ SKUs), long-tail inventory, mature operations with the WMS to drive task lists.

Watch out for: the C-tier blind spot. If your bottom 50% only gets counted twice a year, slow-moving shrinkage can hide in there for months. Most continuous programmes need a "lookback" rule that pulls any zero-on-hand or write-off SKU into the next-day count regardless of tier.

The decision matrix

There's no universally correct cadence. The right answer falls out of five inputs: SKU count, value per SKU, your target accuracy, the headcount you can spare, and what your customer contracts demand.

| Factor | Weekly | Monthly | Continuous (ABC) | |---|---|---|---| | SKU count | Under 500 | 500 – 5,000 | 5,000+ | | Avg value per SKU | High (£500+) or high-shrink | Mixed | Long tail, mostly low-value | | Target accuracy | 99.5%+ | 99.0% | 99.0% on A, 97% on C | | Headcount required | 1–2 dedicated counters | 0.5 FTE equivalent | Driven by WMS, ~0.3 FTE | | Customer audit needs | High (pharma, regulated) | Standard 3PL contracts | Enterprise audits with tiered SLAs | | Time to first accuracy signal | 1 week | 4 weeks | Immediate (rolling KPI) | | Recovery from a bad count | Fast — re-count next week | Slow — month-long gap | Fast — re-queue same day | | Risk if neglected | High burn-out | Cadence drift | Tier blind spots |

A few patterns I see repeatedly when operators sit down with this matrix:

• A small 3PL with 200 SKUs and a couple of high-value clients thinks they need continuous because it sounds modern. They don't. Weekly is genuinely simpler and gives them a tighter accuracy signal.
• A growing fulfilment operation with 3,000 SKUs wants to leap from annual straight to continuous. They almost always fail. Monthly first, prove the discipline, then tier when the SKU count climbs past 5,000.
• A large contract warehouse with 12,000 SKUs runs monthly and wonders why it costs them two full-time counters. ABC-tiering would cut that to one, with better A-item accuracy than they have now.

The playbook for each

Weekly playbook

• Count groups by zone. Five zones, one zone per day. Everyone knows which day their zone gets counted.
• Tolerance threshold: zero. Any variance triggers a recount the same day by a different counter. At weekly cadence you're trying to catch things small, so you don't tolerate "rounding".
• Who counts: not the picker who picked it, not the manager who owns the KPI. A second-shift colleague or a dedicated cycle-count role. This is the single most-violated rule in the industry.
• When: before shift start, 6:00–7:30am. Floor is empty, no concurrent movement, counts can be reconciled before pickers go live.
• Documentation: signed count sheet (paper or RF device), recount sheet if triggered, variance log with a one-line reason. Filed for 12 months.

Monthly playbook

• Count groups by velocity bucket and zone. Run the top-velocity SKUs in week 1, medium in weeks 2–3, slow movers in week 4. Don't try to count the whole warehouse in alphabetical aisle order — you'll miss the items that actually move.
• Tolerance threshold: 1% by value, 1 unit by quantity, whichever is tighter. Anything outside triggers a recount within 24 hours. Most operators set this too loose and end up "accepting" 2–3% variances that compound.
• Who counts: rotating from the inbound team, never the outbound pickers. Inbound staff have less incentive to "explain away" a missing pallet because they didn't put it away — they just counted it.
• When: twice-daily slots, 6:30am and 1:30pm. Splitting it means a sick day doesn't wipe out a whole day's coverage.
• Documentation: WMS-driven count tasks with electronic signature, exception report generated weekly, monthly accuracy summary to operations manager and finance.

Continuous (ABC) playbook

• Count groups by tier, scheduled by the WMS. Don't hand-pick what to count. The system picks based on last-count-date, value, and any flags (zero stock alert, recent write-off, customer audit request).
• Tolerance thresholds vary by tier. A items: zero tolerance, immediate recount. B items: 1% tolerance. C items: 2% tolerance but with a strict cap that any cumulative variance over 5% across the bin triggers a full bin recount.
• Who counts: floor team during downtime windows the WMS identifies. The key is the system must route counts to counters who didn't pick that SKU in the last 7 days. This is impossible to enforce manually at scale.
• When: rolling, throughout the day, fitted around pick waves. The system delays a count if a pick is queued for that location.
• Documentation: all electronic. Daily accuracy dashboard, weekly tier-by-tier report, monthly trending, quarterly review of tier definitions (some SKUs move tier).

Common pitfalls

Counting at the wrong time. The single biggest mistake. Counting an aisle while pickers are working it gives you a number that's wrong before you've keyed it in. The variance you find is then half real shrinkage, half concurrent movement, and you can't separate the two. Off-peak windows only. No exceptions.

Accepting your own variance reports. "Probably a mis-pick last week, write it off." That sentence has caused more inventory drift than every thief who ever walked through a warehouse. Every variance needs a named cause: receipt error, putaway error, pick error, damage, shrinkage. "Unknown" is a category — but if more than 10% of your variances are "unknown" you don't have a counting problem, you have a process problem.

Letting the manager who owns accuracy choose what to count. Self-marking homework. The count schedule has to come from the system or from someone whose bonus doesn't depend on the result.

Counting in pallet whole numbers when you store in units. Or vice versa. Whatever your inventory unit of measure is, the count has to match it. The number of cycle counts I've seen sign off "1 pallet present" when the system said "47 cases" — and the pallet was actually 44 cases — is depressing.

Treating the count as the goal. The count isn't the point. The point is the variance investigation that follows. A warehouse that counts religiously but never investigates the why is just generating paperwork. The KPI that matters isn't "counts completed" — it's "variances investigated and root-caused within 5 working days".

Skipping the recount. First count finds a variance; you key it in; you adjust the system. Wrong. Every variance needs a recount by a different counter before you adjust. The number of times the recount comes back matching the original system figure — and the first count was just wrong — would surprise you.

Where to start

If you're still on an annual stocktake, move to monthly. Don't try to leap to continuous; you'll fail at the discipline before you fail at the tooling. If you're already running monthly and it's drifting, audit the who counts what rule first — that's where 80% of cycle count programmes silently break. If you're past 5,000 SKUs and your monthly programme is eating two FTEs, you're ready to tier.

The cadence matters, but the cadence isn't a strategy on its own. It's the surface of a deeper question: do you trust your inventory number on any given Tuesday afternoon, without warning? If the answer is no — regardless of what your year-end stocktake says — your cadence is wrong or your discipline is.

Cycle counting done well is one of the few warehouse practices where the software actually carries the weight. Schedule generation, tier definitions, off-peak windowing, counter-routing that respects "not the picker who picked it" — these are all things a good WMS does in its sleep and a clipboard never will. Loaditude's inventory tracking treats cycle counting as a first-class workflow with built-in ABC tiering and counter-routing rules, and it slots into the wider warehouse operations toolkit so the count schedule and the pick schedule are aware of each other.

The annual stocktake survives because finance hasn't seen anything better. Once they see a rolling 99.4% accuracy KPI with a defendable audit trail, they stop asking for the weekend shutdown. That's the goal — not the count itself, but the conversation it lets you have with the people who used to make you count.