The Number the Counter Is Not Shown
A blind count hides the expected quantity from the person counting. Doing that properly means hiding three figures rather than one, hiding them in the data rather than on the screen, and hiding the same item everywhere else in the product for as long as the session is live.
If the sheet says the shelf should hold forty-one, the shelf will hold forty-one. This is not dishonesty. It is what happens to anybody counting a hundred and twenty lines before lunch.
A blind count is a count where the person doing it does not know what they are supposed to find. It is the single cheapest improvement available to stock accuracy, and it works for an unglamorous reason: an expected figure on the sheet turns counting into confirming, and confirming is much faster and much less accurate.
Hiding one number is not hiding the number
The obvious implementation hides the expected quantity column and leaves the rest of the sheet alone. It does not work, and the reason is arithmetic: variance is counted minus expected. Print the variance beside the count and the expected figure has been published in a different font.
So the decision here is made once for the whole group rather than per column. Expected quantity, variance and anything either can be recomputed from are withheld together, or none of them are. There is no configuration that hides one and shows another, because that configuration would be a way to believe you were running a blind count while running an ordinary one.
A hidden column beside a printed subtraction is not a hidden column.
Hidden in the data, not on the screen
The second decision matters more than it sounds. The figures are blanked on the record before it is handed to anything that draws it, rather than being skipped by the template. That is the difference between a screen that declines to show a number and a system that declines to send it.
The consequence is that the interface behaves the same way as the page. A counter using a phone, a tablet, an offline client or anything else reading the same endpoints receives a line with those figures absent and a flag saying they are hidden — not a zero, which a client would render as a real quantity, and not the true figure with an instruction not to display it.
The flag matters as much as the blank
A client that receives an empty field has to guess whether the number is zero, unknown or withheld, and all three render differently to a human. Sending an explicit "hidden" alongside the blank lets the counting screen say so plainly. It is a small piece of protocol design that removes an entire class of misread sheets.
The item is hidden everywhere, not just on the sheet
This is the part that distinguishes a blind count from a blind count sheet. While a blind session is live, the items in it have their stock figures withheld from the people counting across the rest of the product — the item list, the item detail page, the stock-by-location view, on the web and through the interface.
Without that, the blind count lasts exactly as long as it takes somebody to open the catalogue in a second tab. With it, the sheet and the system agree, and the count is genuinely blind for the person doing it.
| Session status | Withholding active? |
|---|---|
| Open | Yes — the session exists and lines are being prepared. |
| Counting | Yes — the obvious one. |
| Submitted | Yes — counts are in and awaiting review; revealing now would let a counter revise. |
| Approval required | Yes — variances are being weighed. |
| Adjustment pending | No — deliberately released. |
| Adjusted or cancelled | No — the session is finished. |
The one that looks like an oversight is deliberate and worth explaining. Once variances are approved, they are raised as adjustments rather than written straight into stock, and the session waits for those adjustments to be applied. Holding the freeze during that wait would block the very stock movements that applying the adjustments requires — the session would be waiting on a correction it was itself preventing. Releasing is safe because what is pending is a difference rather than a replacement: an unrelated movement in the meantime still nets out correctly.
Revealing is an act, not a permission
Some people are allowed to see the expected figures during a live blind session: administrators, and anyone holding the permission to apply or approve adjustments. That is the right group, because those are the people who have to weigh a variance rather than produce one.
But being allowed to reveal does not mean seeing. A user who can reveal is still withheld from until they actually do it — revealing is a separate, deliberate act rather than a property of the account. So a supervisor who counts a section themselves counts it blind, like everybody else, unless they choose otherwise and that choice is recorded.
The unassigned lines follow the same logic from the other direction. A line with nobody assigned to it is withheld from the session's creator and from ordinary inventory users, and is not withheld from the reveal group — because for them, the withholding is opt-out rather than opt-in.
Two signatures, two different claims
A count session carries two distinct signatures and it is worth being precise about what each one certifies. The counter signs that the sheet they filled in is complete. The reviewer signs the variances before stock is rewritten. Those are different assertions by different people at different moments, and collapsing them into a single "approved" would lose the only evidence that the person who found the discrepancy was not the person who decided what it meant.
Five questions to ask about a blind count
What exactly is hidden?
A good answer sounds like
Expected, variance and anything derivable.
What ours actually is
All three as one group. There is no setting that hides the expected figure and prints the variance.
Where is it hidden?
A good answer sounds like
In the data.
What ours actually is
On the record before it reaches any renderer, so the interface withholds it identically to the page, with an explicit hidden flag rather than a zero.
Can a counter look the item up elsewhere?
A good answer sounds like
No, while the session is live.
What ours actually is
No. The item's stock figures are withheld from that user across the catalogue, the item page and the stock views, on web and interface, for the four statuses in which a session is active.
Who can see the expected figures?
A good answer sounds like
A named group, by deliberate act.
What ours actually is
Administrators and holders of the adjust or approve adjustment permissions — and only once they have actually revealed, not by default.
Who signs what?
A good answer sounds like
Two signatures, two claims.
What ours actually is
The counter certifies the sheet is complete; the reviewer certifies the variances before stock changes.
What AWRA OpsHub does today
- A per-session blind flag, so blind and open counts can both be run and each one is a deliberate choice.
- Expected quantity, variance and every figure derivable from them withheld as one group, with no setting that separates them.
- Masking applied to the record rather than at the render point, so the interface and the page withhold identically and an offline client cannot read what the screen declines to draw.
- An explicit hidden flag alongside the blanked figures, so a client can say "hidden" rather than draw a convincing zero.
- The same withholding applied across the item catalogue, the item detail page and the stock-by-location views for the duration of a live session, on web and interface.
- A reveal restricted to administrators and holders of the adjust or approve adjustment permissions, and treated as a deliberate act rather than as a property of the account.
- Four session statuses that hold the withholding, and a fifth — awaiting adjustments — that deliberately releases it so applying an approved correction is not blocked by the session that raised it.
- Two distinct signatures on a session: the counter certifying the sheet is complete, and the reviewer certifying the variances before stock is rewritten.
More we can add to your workspace
- A record of each reveal, naming who revealed which items and when, so a deliberate act leaves a deliberate trail.
- A second-count requirement above a variance threshold, sending a line back to be counted again by a different person before it can be approved.
- Blind recounts of only the disputed lines, so a session with six variances out of four hundred does not need the whole sheet recounted.
- A counter accuracy measure over time, built from the variances each counter produced, as an input to who counts what.
- A withheld-item indicator in the catalogue, telling a counter that a figure is hidden because they are counting it rather than leaving a blank they have to interpret.
- Scheduled recurring count sessions generated on a cycle, so cycle counting is a rhythm rather than something somebody remembers to start.
Where we point you to a specialist
- We will keep the three figures inseparable. An option to hide the expected quantity while printing the variance would let an organisation believe it was running a blind count when it was not, and a control people wrongly believe in is worse than one they know they have yet to put in place.
- Whether your stocktake procedure satisfies your auditor is a question for your auditor. We will produce blind sheets, the signatures, the variances and the adjustments raised from them; whether that constitutes sufficient evidence in your sector is theirs to say.
- We will not let a count write stock directly. Variances become adjustments that somebody approves, because a count that silently rewrites a balance removes the review step at exactly the moment a review is most valuable.
A reveal audit trail is the smallest piece here and the one with the clearest value: the act is already deliberate and already permission-gated, so recording it is a row rather than a redesign.
From a blind sheet to a counting programme
The withholding is thorough. What organisations ask for next is the process around it — recounts, cadence, and knowing who reveals what.
A reveal trail
Who revealed which items, on which session, and when — the audit half of an act that is already deliberate.
Threshold-driven recounts
A variance above a line you set sends that item back to be counted again, by somebody else, before it can be approved.
Scheduled cycle counts
Sessions generated on a cycle by class or by location, so counting is a rhythm rather than a memory.
We publish scope, not dates.
Scope a counting programmeRun one section blind and compare
The argument for blind counting is not theoretical and it does not need a pilot programme. Count one aisle blind and one aisle with the sheet filled in, and the variance distribution will make the case in an afternoon.
Talk through stocktakesFrequently asked questions
Why hide the variance as well as the expected quantity?
Because variance is counted minus expected, so anybody who can see their own count and the variance can recover the expected figure by subtraction. Hiding one and showing the other is the most common way a blind count stops being blind, and it is invisible from the screen because the hidden column really is hidden. The three figures are treated as one group for exactly this reason.
Can a counter just look the item up in the catalogue?
Not while the session is live. The items in a blind session have their stock figures withheld from the people counting them across the item list, the item page and the stock-by-location views, and the same restriction applies through the interface. Without that, the blind count would last as long as it took somebody to open a second tab.
Who is allowed to see the expected figures?
Administrators and users holding the permission to apply or approve adjustments. Even for them it is not automatic — they are withheld from until they deliberately reveal, so a supervisor counting a section counts it blind unless they choose otherwise.
Why is the withholding released while adjustments are pending?
Because holding it would block the stock movements that applying those adjustments requires, and the session would be waiting on a correction it was itself preventing. It is safe to release because what is pending is a difference rather than a replacement — an unrelated movement in the meantime still nets out correctly.
Does a count write stock directly?
No. Approved variances are raised as adjustments and stock changes when those are applied. That keeps the review step where it belongs and is why the session has a status meaning "counted, approved, waiting for the corrections to land".
Is blind counting the right default?
For a full stocktake, almost always. For a quick spot check where somebody is verifying a specific suspicion, an open count is a reasonable tool and pretending otherwise would be dogma. It is a per-session flag because it is a choice about what this particular count is for.
