A method organised around practical checks.
Each step gathers the facts to establish before moving to the next.
Cycle counting is the most widely recommended and most frequently abandoned initiative in logistics. Almost everyone has tried it. Few still practise it a year after launch.
This guide starts from that observation rather than from theory. It describes what must be decided before starting, in what order, and why unsuccessful initiatives fail.
What cycle counting replaces, and what it does not
A complete annual stocktake fulfils an accounting obligation: establishing the value of stock on a given date. It does that well.
What it does not do is maintain daily reliability. Between two annual stocktakes, stock drifts, and no one knows until the next count.
Cycle counting meets another need: detecting variances close to their source. It does not necessarily remove the need for an annual stocktake — that depends on your accounting context and the demonstrated quality of your process. This must be decided with your statutory auditor, not in a guide.
PART ONE — THE SIX PRIOR DECISIONS
They can be made in a two-hour meeting, and they determine the initiative’s success.
Decision 1 — Who counts
Three options, with different consequences.
---------------------- ----------------------- ----------------------- Option Benefit Risk
Operations teams, No cost, knowledge Counting is skipped on during their time of the floor busy days
A dedicated part-time Regularity, growing Cost, dependence on person expertise one person
A separate team Independence from Cost, lack of floor the person who made knowledge the error ----------------------- ----------------------- -----------------------
The second option is the most sustainable, for a simple reason: regularity survives better when it is not competing with daily urgency.
One rule must be decided explicitly: is the person who counts also the person who puts stock away? Independence improves count reliability, but costs time in travel and location knowledge.
Decision 2 — How often
This question is answered by product class, not globally.
A distribution that works in most operations:
---------------------- ----------------------- ----------------------- Class Criterion Frequency
A High value or high Monthly turnover
B Intermediate Quarterly
C Low value and low Half-yearly or turnover annually ----------------------- ----------------------- -----------------------
The classification criterion depends on your issue. If the concern is financial, classify by value. If it is customer service, classify by dispatch frequency. The two classifications produce very different lists.
Do not forget products with particular risks — attractive to thieves, fragile, short-dated or regulated. They deserve their own frequency, regardless of value.
Decision 3 — How to count
Blind. The counter does not see the expected quantity. This is non-negotiable: a counter who sees ‘48’ and counts 47 will count again until they find 48, or record 48. This is not dishonesty; it is how human attention works.
By location, not by item. Count everything in one place. Counting by item requires visiting all its locations, which takes longer and misses misplaced products.
With a second count when there is a variance. Any variance above a defined threshold triggers a recount by another person before any adjustment. Set the threshold by class.
Decision 4 — Who adjusts
This is the most important and most frequently forgotten decision.
The counter must never be able to make the adjustment. Otherwise the variance disappears without a trace and the process loses its purpose: it corrects the symptom without ever revealing the cause.
Write three lines:
AUTOMATIC ADJUSTMENT if variance < [threshold] → recorded, without approval
APPROVAL [role] if variance between [threshold 1] and [threshold 2]
APPROVAL [management] above [threshold 2]
Thresholds are expressed in value, not quantity. Three missing units do not mean the same for a ten-dirham product and a five-thousand-dirham product.
Decision 5 — Which reason
An adjustment without a reason is lost information.
A closed list of six to eight reasons is sufficient and must be adapted to your operation. A common starting point:
picking error
receiving error
unreported breakage or damage
misplaced product (found elsewhere)
data-entry error
theft or disappearance
unit-of-measure variance (case / unit)
cause not identified
The final line is essential, and its proportion is your maturity indicator. At launch, it will account for most variances. If it has not fallen after six months, the reasons are not being entered seriously.
Decision 6 — Who reviews, and when
This is where the initiative’s survival is decided.
Cycle counting that produces adjustments but is never reviewed becomes a correction routine. It consumes time and changes nothing. Teams understand this within a few months, and the initiative dies by itself.
Set the ritual from the outset: one hour a month, one named person, three questions.
Which locations produced recurring variances?
Which reasons dominate, and in which families?
What corrective action will we decide by next month?
Without this monthly hour, everything else is counting. With it, this is an improvement initiative.
PART TWO — THE TWELVE-WEEK START-UP PLAN
Weeks 1 and 2 — Prepare
☐ Make the six decisions above and write them on one page
☐ Classify items (A / B / C + risk class)
☐ Check that every location has a readable identifier
☐ Set adjustment thresholds by value
☐ Establish the closed list of reasons
The third point blocks more initiatives than people imagine. Counting by location assumes that locations are identified and their identifiers can be read from the aisle. If not, this is the first task — and it is inexpensive.
Weeks 3 and 4 — Measure the baseline
☐ Count a representative sample: around one hundred locations across all classes
☐ Record the observed accuracy, by number of correct locations and by value
☐ Archive this figure
This baseline will enable you to demonstrate progress in a year. Without it, everyone will feel that ‘things are better’ and no one will be able to prove it.
Weeks 5 to 8 — Trial period
☐ Count according to plan, without making automatic adjustments
☐ Have all adjustments approved, regardless of amount
☐ Record the practical difficulties encountered
This phase reveals methodological problems before they become habits: locations that cannot be found, ambiguous units of measure, unidentified products and underestimated counting time.
Weeks 9 to 12 — Steady state
☐ Apply automatic-adjustment thresholds
☐ Hold the first monthly review
☐ Decide and implement a first corrective action
☐ Communicate the result to the teams
The last point is often neglected. The people who count need to know why it matters. A variance handled, a cause identified and an action taken give meaning to a repetitive task.
PART THREE — WHY INITIATIVES FAIL
The first enemy: workload
A busy operation abandons counting. That is rational: shipping takes priority over counting.
Two safeguards:
A sustainable volume. Ten locations a day maintained all year are better than fifty for six weeks. Calculate the volume from time actually available, not theoretical time.
Planned catch-up. If one week is missed, it is caught up according to a written rule, not the goodwill of the moment.
The second enemy: no consequences
If counting changes nothing visible, counting stops.
The monthly review is not a refinement: it is the mechanism that keeps the initiative alive. One visible corrective action per quarter is enough to sustain engagement.
The third enemy: looking for someone to blame
As soon as variances are used to identify who is responsible, two things happen: reported variances fall, and actual variances do not move. Stock is quietly put back in order before the counter arrives.
The rule that protects the initiative: variances reveal causes, not names. People management happens elsewhere, using other tools.
The fourth enemy: too much precision at the start
Some people want highly detailed reasons, thresholds by family and cross-classifications. The system becomes cumbersome before it has proved anything.
Start simply: three classes, six reasons, two thresholds. Refine them once the initiative is established — meaning once it has lasted a year.
PART FOUR — THE TWO INDICATORS TO TRACK
Not ten. Two.
Accuracy by location. The number of locations counted correctly divided by the number of locations counted. Tracked monthly and compared with the archived baseline.
Detection time. The difference between the probable date of the incorrect movement and the date of the count that revealed it. This indicator shows whether your counting plan is well targeted — and it is the only one that measures the real contribution of cycle counting compared with annual stocktaking.
A third indicator to review quarterly: the proportion of variances without an identified cause. It must fall. If it remains above half after six months, entering reasons has become mechanical and adds nothing.
What can honestly be expected
Do not expect a miracle in three months. A cycle-counting initiative takes effect over a long cycle for a logical reason: it does not correct stock; it reveals causes. Those causes are then corrected one by one, and each correction takes time to show in the figures.
What you will notice fairly quickly — from the first few months — is a better understanding of your own processes. Variances tell the story of where your operation goes off track. This is often the most useful benefit, and no one highlights it.
