How to run a stocktake that tells you something

A stocktake is a full count of everything you hold, on one day, against the number the system shows. Done well, it ends with causes, not corrections.

By Andres Rodriguez Rey Updated November 3, 2026 5 min read

Sunday morning, fulfilment paused, four people with clipboards. The system says 1,240 of your best seller; the shelf says 1,187. A stocktake is how you learn that 53 are missing. What you do with that number is the whole point.

This page covers how to prepare and run the count, what to do with the variances it finds, and when a rolling count should replace it. The rolling version is cycle counting; in the US the full count is usually called a physical inventory count, and the method is the same.

What is a stocktake, and how is it different from a cycle count?

A stocktake counts everything you hold at one moment and compares it to the record. A cycle count counts part of the catalogue on a rolling schedule and never stops the business. The stocktake gives you a true baseline; cycle counts keep it true.

The full count costs a shutdown and a weekend. Its value is that every line is measured on the same day, so the variances can be read together and a pattern shows: one zone off, one product family short, one process that leaks. A cycle count cannot give you that picture, because its lines are measured weeks apart.

That is why most small brands keep one full count a year even after cycle counts are running. The annual count is for the books and for the pattern; the cycle count is for catching the next error before a customer does.

How do you prepare and run the count?

Pick a quiet window, stop receipts and shipments for its length, and clear the floor of anything half done. Print count sheets by location with the quantity column blank, assign pairs to zones, and have one person count while the other writes. Finish a zone before moving on and mark it done.

Before the count
Do thisWhy
Freeze receipts and shipmentsA unit that moves mid-count is counted twice or not at all
Clear staging, returns and the pack benchHalf-done work is where most variances hide
Print sheets by location, quantity blankCounters find stock where it is, and cannot copy the expected number
Pair counters on high-value zonesOne counts, one writes, both initial the line
Name a referee for recountsA variance is recounted before it is argued about

Blind sheets are the one rule people argue with and the one that matters most. A counter who can see the expected number tends to find it. Keep the system quantity at the desk, compare there, and let the floor report what it sees.

Count in units even when stock sits in cases, and convert once at the desk. Damaged or on-hold stock goes on its own line and is labelled, not folded into the good count, because it is the first thing the variance list will ask about.

What do you do with the variances?

Sort variances by value, not by count, and start at the top. For each one, ask what process could have produced it: a receipt booked twice, a return put back in one system only, a bundle whose parts did not move, a mis-pick. Adjust the record only after the cause is named, and write the cause next to the adjustment.

Reading the pattern
What the variances look likeLikely cause
Many SKUs a little shortMis-picks or shrinkage; check the pick process first
A few SKUs badly overA receipt booked twice, or a return restocked in one system only
Bundle parts short, the bundle notComponents did not move when the bundle sold
One zone off, the rest cleanA counting error; recount before anything else

The log of causes is the product of the day. Corrections make the number right for a week; the causes tell you which process to change so that next year's count is smaller. Keep it to four columns, SKU, expected, found, cause, and read it once more a month later.

When should a stocktake give way to cycle counts?

When the full count's variances are small and the causes repeat. At that point a rolling cycle count on the SKUs that carry the value and move the most catches the same errors sooner and without a shutdown. Keep one full count a year for the books; let cycle counts do the operating job.

Which SKUs earn a weekly count, and how to tier the rest, is in cycle counting cadence. The stocktake you just ran is its starting data: the SKUs that drifted most are the ones that go on the weekly list first.

What should you do next?

Before

  • Pick the window and freeze receipts and shipments.
  • Clear staging, returns and the pack bench.
  • Print blind sheets by location and assign pairs.

During

  • Count units where they sit; both counters initial each line.
  • Mark each zone done before moving on.

After

  • Sort variances by value and name a cause for each.
  • Adjust only after the cause is written down; keep the log.

Frequently asked questions

What is a stocktake?

A complete physical count of everything a business holds at one point in time, compared line by line to the quantity its records show. The variances between the two are what the exercise is for.

How often should a small brand do a stocktake?

Once a year as a full count, with cycle counts in between. If the full count finds variances worth more than a few percent of stock value, run a second full count six months later and fix the causes in between.

Should the web store be paused during a stocktake?

Pause fulfilment, not necessarily sales. Orders can keep arriving as long as nothing is picked, packed or received during the window; they ship after the count against the corrected numbers.

Count once, then keep the number true

Start with the free reconciliation scorecard, then keep one count per location for Shopify and Square in a 45-day trial.

Start your 45-day trial — no card