Opening Stock Balances at Go-Live
The number you type into a new system on day one is the number every future variance is measured against. Type it carelessly and you have not saved a week — you have bought a permanent argument about whether the system or the store is lying. How to count into a system properly, and why the count has to be blind.
There is a moment in every inventory rollout where someone suggests exporting the current stock figures from the old spreadsheet and loading them in, because the store is busy and a full count would cost a Saturday. It sounds like pragmatism. What it actually does is inherit every error in the old records and then attach a new system's authority to them, so that six months later the storekeeper is being asked to explain a discrepancy that was already there before he touched anything.
The opening count is the only moment you will ever get where the system and reality can be made to agree by definition. Spend the Saturday.
Why the opening count has to be blind
A blind count is one where the counter cannot see the expected quantity. It sounds like distrust and it is not — it is a correction for a well-documented human tendency. Show someone that the system expects 48 and they will count 48. Not dishonestly; they will count a row of tins, get 47, assume they miscounted, and count again until the numbers agree. This happens to careful people. It happens most to careful people, because they assume the discrepancy is theirs.
On an opening count the stakes are higher than usual, because there is nothing behind the figure. Every later count is checked against a system balance built from recorded movements. The first one has no such backstop — whatever the counter writes down becomes the truth, permanently, and the only defence against a lazy or anchored count is that nobody could see what they were supposed to find.
-
Freeze movement, genuinely
No receiving, no issuing, no sales from the counted area for the duration. In practice that means a Saturday, or after close on a Friday, or one warehouse at a time. A count taken while stock moves is not a count; it is a sample from a moving population, and it will differ from itself.
-
Count by location, not by item list
Walking the shelf and recording what is there finds the items your list has forgotten. Walking the list and looking for each item finds nothing you did not already know about, and takes twice as long.
-
Two counters on high-value lines
Independently, without conferring, on anything above a threshold you set — in most Kenyan SMEs somewhere around KES 20,000 of line value. Where the two disagree, count a third time. Where they agree, stop.
-
Record damage and expiry as themselves
Broken, expired and unsellable stock is not stock. Counting it in gives you an opening figure that includes goods you can never sell, and the write-off then looks like a loss you caused rather than one you inherited.
-
Enter it the same day
Count sheets that sleep overnight get tidied. The person tidying them is trying to help and is destroying the only clean baseline you will ever have.
The opening count is the only stock figure in the history of your business that you can make true by fiat. Every subsequent one has to be earned.
What the count is worth in shillings
It is worth doing the arithmetic, because "a Saturday" sounds expensive until you compare it to the alternative, which is not zero — it is a year of low-confidence figures.
A 1,200-line store, counted properly against copied figures
The 3–8% range is what businesses of this size typically find the first time anyone counts everything on the same day. If your first count comes back at under 1% variance, do not celebrate — check whether it was blind, because that is the number an anchored count produces.
Costing the opening stock, which is a separate decision
A quantity without a value is half a record. The opening cost you assign determines your opening inventory value, your first gross margin figures, and the point from which weighted average cost starts moving. Three defensible answers, in order of preference:
| Basis | When it is right | The catch |
|---|---|---|
| The last price you actually paid | Almost always. It is recent, defensible and you have the invoice. | Needs the invoice found for each line. Slow on a long tail. |
| A weighted average of the last few purchases | Volatile imported goods where one recent invoice is unrepresentative. | More arithmetic, and you have to decide "a few". |
| Current supplier list price | Only for lines where you genuinely cannot find what you paid. | Overstates inventory value and flatters your first margin. |
Note what is not on that list: selling price. It appears on this list in real projects surprisingly often, usually because it was the only number to hand. Valuing stock at what you hope to sell it for overstates assets and understates margin on every subsequent sale, and an auditor will find it immediately. Our own costing basis is weighted average, with the buying price used where an item has never been costed — the reasoning is in FIFO versus weighted average.
What the system does for you here, and what it does not
The opening count, layer by layer
Blind counting
Set on the count plan and the session. Expected quantities are withheld from the counter and disclosure is gated by a separate permission, so it is an enforced control rather than a discipline someone remembers.
Count plans with a defined scope
By warehouse, location or category, so you can count one area at a time instead of freezing the whole business.
Variance captured per line
Counted against expected, line by line, with the difference recorded rather than silently absorbed.
Which lines get two counters
A policy you set and enforce. There is no dual-count requirement in the system, so this lives in your instructions to the team.
Count scheduling
A plan defines scope, not recurrence. There is nothing that raises next quarter's count on its own — the rhythm is a diary entry.
Count accuracy as a KPI
Variance is captured per line but nothing aggregates it into an accuracy percentage, a trend, or a scoreboard by counter or location. You can compute it from the data; the system will not compute it for you.
Recount of a single disputed line
A disputed line needs another count session covering it. There is no per-line recount step.
For an opening count the two absences barely matter — you are running one session, once, and you will read the variance yourself. They matter from the second count onwards, which is the argument in cycle counting versus the annual stocktake.
On the financial side, there is no equivalent
Stock has a count. Debtors, creditors and bank balances do not, and there is no manual journal entry and no working opening-balances importer — so your opening financial position has to arrive as actual documents. That is a different job with a different owner, and it is covered in setting up the chart of accounts and opening balances.
Our take
Count it, blind, in one frozen window, by location rather than by list, with two counters on the high-value lines and damage recorded separately. Value it at what you last paid. Enter it the same day. This is one bad Saturday that removes an entire category of argument from your next two years, and it is the single implementation task we would never let a client negotiate away.
Set up the count before you set up the system
Count plans by warehouse, location or category, blind by default with disclosure behind its own permission, and variance recorded line by line so the opening position is evidence rather than an assertion.
See plans & pricingFrequently asked questions
Can we just import our current stock figures instead of counting?
You can type them in as an adjustment, but you should not, and there is no importer for stock quantities precisely because a bulk load would bypass the count and approval path that makes a figure trustworthy. Copied figures inherit every error already in the old records and then borrow the new system's authority for them — which is how a storekeeper ends up defending a discrepancy that predates his involvement.
What is a blind count and why does it matter?
A count where the counter cannot see the expected quantity. It corrects for anchoring: shown that the system expects 48, a careful person who counts 47 will assume they miscounted and count again until it agrees. On an opening count there is no system balance behind the figure at all, so the only protection against an anchored count is that nobody could see what they were supposed to find. Blind counting is set on the plan and the session here, with disclosure gated by its own permission.
How much variance should we expect on a first count?
Three to eight percent of stock value is normal for a business counting everything on the same day for the first time. A first count that comes back under one percent is more likely to indicate an anchored count than an accurate store — check whether it was blind before treating it as good news.
What cost should we put on opening stock?
The last price you actually paid, in almost every case — recent, defensible, and you have the invoice. A weighted average of recent purchases is better for volatile imported goods. Supplier list price only where you genuinely cannot find what you paid, and never selling price, which overstates your assets and understates margin on every later sale.
Do we have to freeze the whole warehouse at once?
No. Count plans scope to a warehouse, a location or a category, so you can freeze and count one area at a time. What you cannot do is count an area while stock is still moving through it — that produces a figure that differs from itself depending when you looked.
Does the system tell us our count accuracy?
It records variance line by line, but it does not aggregate that into an accuracy percentage, a trend or a scoreboard by counter or location. For an opening count that is fine, since you read the variance yourself. From the second count onward it means the accuracy metric is something you compute, not something you open.
