DEV Community

Hive80-lab
Hive80-lab

Posted on Originally published at hive80-lab.github.io

Stocktake Checklist: Count What You Have, Not What the System Says You Have

A stocktake is the only instrument that measures the gap between what your shelves hold and what your system claims — and that gap is where the margin goes. Most small businesses measure it once a year and receive the answer as a shrug: shrinkage, call it 2%. The 2% isn't a weather condition. It's an invoice — every dollar of it went somewhere specific, and the only way to make it speak is to measure it while the trail is warm.

The quick version: run it blind, count what's expensive first, recount before you adjust, and never move a number without a reason code attached.

The five-step stocktake ritual

  1. Count blind. The counter never sees the system number — print count sheets without the on-hand column. A counter who knows the system says 12 will find 12. Blind counting is the whole discipline; every other rule is scaffolding for it.
  2. Freeze the floor, or count in zones. Small business that can close for two hours? Count everything at once. Can't stop? Rolling zone counts on a calendar — stockroom this week, fridges next. Never count a zone mid-delivery: receiving and counting the same shelves in the same hour manufactures variances out of thin air.
  3. Count value-first. The top 20 SKUs by value get counted to the unit, every cycle, by two people on different shifts. The cheap stuff gets tolerance bands — nobody recounts a $3 item to resolve an 11-cent gap.
  4. Recount before you adjust. Every above-band variance gets a second pair of hands: different person, different shift, blind again. Half of all variances die right here — the first count was wrong, and no adjustment was ever needed. The rule prevents the most expensive kind of accounting: fixing the books to match a bad count.
  5. Adjust with a reason code, or not at all. Receiving error, spoilage, breakage, comping, theft-suspected, data entry, supplier short. A number without a reason is a decision postponed — and a postponed decision is a pattern that gets to repeat all year.

The rules that make the count honest

  • The system number is a claim, not a fact. It's the residue of every uncounted delivery and unlogged breakage since the last count. The shelf is the fact. The count makes the claim match the fact — never the reverse.
  • Count units you sell, not units you filed. The system holds coffee in kilos, the shelf holds bags, the invoice holds cases. One counting unit per SKU, printed on the sheet, converted in exactly one place. Unit-of-measure mismatch is the quietest variance factory in small business.
  • Variance gets a threshold and a band. ±2% on the $500 line gets investigated same-day; ±10% on the $3 line gets noted and released. Bands are where the counting effort goes instead.
  • Shrinkage is a story, not a subtotal. Every unexplained dollar gets a hypothesis and a fix within the month. A hypothesis you can't test in a month is an excuse with a schedule.
  • The count feeds the close. Reason-coded adjustments land in the monthly close as COGS you can defend, not a memo you shrug at. And the delivery receiving checklist is the upstream half of the same ledger: a docket counted at the door is a variance that never gets born.

The five traps

  • The annual-only stocktake — twelve months of drift measured at once, every trail cold, the causing habit long since 364 repetitions deep.
  • Counting from memory — "I know we have 12" is the sound of a blind count dying. Memory is the system's advocate in the room; the count exists to outvote it.
  • The tidy-up adjustment — the gap rounded to zero, the books balanced, the cause never named. That's not an accounting entry, it's an alibi, renewed every count.
  • Counting in sales units, adjusting in purchase units — the count says 3 bottles, the system adjusts 3 cases, and next quarter the books announce 33 phantom bottles.
  • The pre-announced count — everyone knows stocktake is Friday, so Thursday night is unusually tidy. Surprise spot counts on the top-value SKUs are the honest version: cheap, frequent, impossible to rehearse for.

Cycle counts for businesses that can't stop

  • ABC the shelf: A items (top ~20% of SKUs carrying ~80% of value) count weekly; B monthly; C quarterly with bands. Twenty SKUs is an hour with a blind sheet.
  • The daily ten-minute count: top five value items, every trading day, before open. Keeps two counters calibrated and catches big leaks while they're small.
  • Tie the count to receiving: every counted docket becomes tomorrow's count-sheet input. Where the two disciplines touch, variances get caught in days instead of quarters.
  • Surprise one zone a month: unannounced, top-value zone, thirty minutes. The month it finds nothing is a good month twice over.

Worked example: the bottle shop's $6,200 answer

A suburban bottle shop replaced its annual count with a rolling weekly zone count — blind sheets, recount before adjust, reason codes on every line. Two weeks running, the spirits zone showed the same variance: three bottles of the same $105 whisky, entering nowhere in the system. The reason codes pointed one direction, a camera angle confirmed it, and the leak closed in a fortnight — roughly $6,200 a year, found by an hour a week that had previously been spent on nothing.

The shop up the road does the annual count every January and budgets 2% for shrinkage, because it has always been 2%. It isn't a budget. It's a leak with a subscription.


The full page — with the worked detail, the variance-band tables, and the kit block — lives at the canonical source: Stocktake Checklist on HIVE80lab ops notes. Related reading: the delivery receiving checklist (where the count starts), the monthly close checklist (where the adjustments land), and the vendor escalation ladder (when the reason code keeps reading supplier short).

Top comments (0)