Stock count variance: when to recount, adjust or leave it

Recount a line when the gap is large in units or in money, not when it crosses one percentage. A workable rule, and it is ours, so change the numbers: recount at 3 units (if that is also 5% of the expected quantity) or $25. Investigate before you adjust at 25 units or $100. Everything smaller gets adjusted and left alone.

Halfway through a stocktake (or stock take, if that is your word for it), the question is which of the differences deserve another hour. A rule you write down before the count answers it. The steps around it are in our stocktake guide.

What a variance is, and which number it uses#

A variance is the counted number minus the expected number. Counted 5, expected 6, variance minus 1. Counted 15, expected 12, variance plus 3.

The expected number has to be something you can name. If you compare with the wrong Shopify state, you invent a gap that does not exist. Our article on why Shopify does not match the shelf explains on hand, available and committed. Pick one, write it on the count sheet, and use it for every line.

A variance in units says how far off the shelf is. A variance in money says what that costs you. You need both:

Money gap = units gap x unit cost

Use your cost, not your price. If a variant has no cost, the money gap is a guess, and a price will overstate it.

Why a percentage rule alone misleads#

Shopify's own guide to counting with POS gives a simple recount rule. It says to recount a product physically when the variance is more than 5% (Shopify blog, 1 Apr 2026, checked 15 Oct 2026). That is a fair starting point, and the guide does not claim it is the only way. Here is where a percentage on its own fails.

Take a mug with 6 expected and 5 counted. That is one unit, worth $4.10 at cost.

Variance = 5 - 6 = -1
Percent  = 1 / 6 = 16.7%

Sixteen percent looks alarming. It sends someone back to the shelf, walking past other work, to settle a question worth $4.10.

Now a tote bag with 400 expected and 380 counted, at $2.50 each.

Variance = 380 - 400 = -20
Percent  = 20 / 400 = 5.0%
Money    = 20 x 2.50 = $50.00

Exactly 5.0% is not more than 5%, so Shopify's rule does not flag it. Yet it is twenty units and fifty dollars, three times the money of the lines the percent rule did flag. Small quantities swing the percent wildly. Large quantities hide big gaps inside small percents.

A flat unit limit has its own hole. Take 1,000 screws expected and 985 counted, at $0.04.

Variance = 985 - 1,000 = -15
Money    = 15 x 0.04 = $0.60

Fifteen units is a lot to a flat limit of 3, and sixty cents to your business. Money is the fix for cheap items with big quantities. Units fix expensive items with tiny ones. Use both.

A two-part rule: units and money#

Write two limits for each level of response. A line trips a level if it passes either row.

The 5% condition on the unit row is what keeps the screws out. Without it, 15 screws would trip the 3 unit limit. With it, 15 is 1.5% of 1,000 and passes.

The levels mean this:

• Below both limits: apply the counted number and move on. No recount.
• Recount: a second person counts the line again before you apply it.
• Investigate: if the recount agrees with the first count, find the cause before you adjust. Do not write the gap off first.

Worked example: six lines, one policy#

Say a count comes back with these six lines.

Now run each line through the rule.

• A and C: adjust, no recount. Mug: 1 unit and $4.10, under both limits. Screws: $0.60, and 15 units is 1.5% of expected, under the 5% condition. Together they are worth $4.10 + $0.60 = $4.70.
• B: recount. $50.00 is over $25 and under $100. 20 units is 5.0% but under 25 units.
• D: recount. One unit, but $48.00 is over $25. A percent rule would also catch it, at 33.3%, but only because the stock is small.
• F: recount. The money is only $15.00. The units trip it: 3 units, and 3 / 12 = 25%, over 5%. An overage deserves a second look as much as a shortage does, because it can mean a delivery that nobody booked.
• E: investigate. $180.00 is over $100. Recount it, and if the second count agrees, find the cause.

That is 2 lines adjusted, 3 recounted, 1 investigated. Under a 5% rule, A, D, E and F would be flagged and B and C would not. The two-part rule skips A, sends you to B and leaves E as the one that gets real attention.

Net result if every count stands:

Units: -1 - 20 - 15 - 1 - 3 + 3 = -37
Money: -4.10 - 50.00 - 0.60 - 48.00 - 180.00 + 15.00 = -267.70

Net numbers cancel overages against shortages. Read them last.

How to recount#

Send a different person. The first counter has the first number in their head, and so does anyone who has seen the screen. A recount by the same person is likely to repeat the first count, right or wrong.

Count the whole location of that item, not only the shelf slot. A unit missing from the shelf may sit in the back room, in an open box, or on a pack station.

Do not show the first number or the system number. If your tool puts the expected quantity next to the line, cover it or count on paper. A recount is the hard case, because the line already carries both numbers. Have the second counter write their figure down first, then type it in.

Then compare:

• The recount equals the first count. The gap is probably real. Apply the number or investigate, by your level.
• The recount equals the expected number. The first count was a mistake. Keep the recount.
• The recount is a third number. Count a third time, with a third person if you can.

Say the recount of D finds 3 headphones, one in a box that the first counter missed. D's gap is now 0 and drops out. The recount cost a few minutes and prevented a $48.00 write-down.

After a recount: adjust, investigate or leave it#

When a recount confirms a gap at the investigate level, read through the causes in our article on why stock does not match the shelf. Start with the first three: an unbooked delivery, a return never put back, damaged stock never adjusted.

Say the jacket gap of 3 stands. A look at the history shows three returns last month marked as refunded without restock, but the jackets went back on the rack. Then the physical stock is right, the record was wrong, and the adjustment is a correction, not a loss.

Skip the investigation if the gap is below the investigate limit. Apply it. Write one line on why you think it happened, even if the line is only "probably a miscount". An adjustment with no note is how the same gap returns next quarter.

A common mistake: a team writes off every difference straight away because the count is already late. The shortages vanish from the books, and nobody learns that the same product leaks every month. A note costs ten seconds, and it is the only record of why the number moved, which is the first thing anyone asks when the same gap shows up again.

When a recount keeps changing#

Counts that move on every pass point at the method, not the stock. Check these:

1. Cases counted as singles on one pass and the reverse on another.
2. The same product stored in two places, with one place missed each time.
3. Items that are hard to see: deep boxes, small parts, anything stacked in a way that hides a layer.
4. A counter working fast because the shop opens in ten minutes.

If one product shows a different number on three passes, set a rule for how you count it. For example: "count by sealed carton, then add loose units". Write it next to the shelf.

Sales during a long count are a different issue. A sale moves the expected number under you. Our article on counting while your store keeps selling covers it.

Repeat offenders#

Keep a short log: product, date, gap in units, gap in money, cause. After three counts you will see which lines are always off, and by how much. A line that is off every time by a similar amount has a cause that is not random: a supplier that ships short, a bin that gets raided, a barcode on two variants.

Do not set the limits once and forget them. If almost every line goes to a recount, the limits are too tight for your counters, or something upstream is leaking. If none does for several counts in a row, check that nobody is counting from the screen.

Take it to the next count#

Write the table above on one page with your own numbers. Put it where the counters will see it. After the count, go down the Off expected list once, mark each line as adjust, recount or investigate, and do the recounts before you apply anything. Our inventory count page shows what that review step looks like.

Frequently asked questions#

How much stock count variance is acceptable?

No single number fits every store. Shopify's POS counting guide uses a 5% recount trigger, but a percent alone misleads on small and large quantities. Set two limits you can defend, a unit gap and a money gap. In our example policy, a gap under 3 units and under $25 is adjusted without a recount.

Do I need to recount every difference?

No. Recounting every line costs hours and mostly confirms small gaps. Recount the lines that trip your unit or money limit, and adjust the rest. In our six-line example, 2 lines worth $4.70 together skip the recount.

What is a blind recount?

A blind recount is a second count by a different person who has not seen the first number or the system number. It removes the pull to agree with a figure already on the screen. Use it for lines that tripped a limit, not for the whole shelf.

What if the count changes every time I recount a line?

Then the stock is not the problem, the method is. Look for mixed units (cases against singles), stock in two places, items that are hard to see, or a counter who is rushed. Fix that, then count the line once more.