Physical stock against book stock

Book stock is what your records say you have. Physical stock is what is on the ground. Reconciliation is comparing them, and it is the only thing that keeps a store's numbers honest over time.

Everybody knows the exercise. Rather fewer get anything useful out of it, because the exercise is usually treated as a way of producing a corrected number rather than as a way of producing information.

The arithmetic#

For one material, over one period:

Opening stock, plus receipts, less issues, equals closing book stock.

Count the material. The difference between that count and the closing book figure is the variance.

That is all the arithmetic there is. The difficulty is entirely in what the variance means and what you do with it.

Variance is a residual, and residuals absorb everything#

The variance is not a measurement of pilferage, or of wastage, or of anything else in particular. It is whatever is left over after every recorded event has been accounted for, which means it contains every unrecorded event of every kind:

• Issues that were never written down.
• Receipts posted at a wrong unit.
• Material returned to store and never entered.
• Material transferred to another site with no paperwork.
• Counting error, this time or last time.
• Damage and handling loss.
• Actual loss.

Six of those seven are recording problems. The seventh is the one everybody jumps to. If your issue register is incomplete — and on most sites it is, for the reasons set out in running a site store — then your variance is dominated by the first item on that list and tells you almost nothing about the last one.

So the first question about any variance is not where did it go. It is how good are the records this variance was computed from.

Count small, count often#

The standard practice is a full stocktake at long intervals. It is the worst possible design.

A quarterly count produces one large variance covering three months of transactions, thousands of entries and several changes of personnel. It is unexplainable by construction. So it gets written off with a generic reason, and the exercise teaches nobody anything.

Counting a few materials every week — perpetual inventory, cycle counting, whatever your organisation calls it — produces small variances covering a small number of entries. Those are traceable. Somebody can actually walk back through a week and find the issue that was not recorded.

The other advantage is unpredictability. A count everybody knows the date of is a count people prepare for.

Rank what to count by value at risk rather than by item count: quantity times rate times how easily the material moves. The top few materials on a construction site usually carry most of the exposure, and counting them monthly while counting the tail annually is a far better use of the same hours than counting everything quarterly.

The adjustment is an entry, not an edit#

When the count and the book disagree, the book has to move. How that happens decides whether the exercise was worth doing.

The wrong way is to correct the stock figure to match the count. The number changes, the difference disappears, and next quarter nobody can tell whether this material has needed adjusting every quarter for two years.

The right way is a dated adjustment entry, of a stated quantity, with a stated reason, attributed to a person, sitting in the ledger alongside every other movement. The balance ends up the same. The history does not.

This is the store's version of a rule we hold everywhere in our own systems: a correction is an entry, never an overwrite. A record that can be silently revised is a record that answers today's question and no other.

The pattern is the finding#

One variance is noise. Materials get miscounted, entries get missed, and a single number in a single month means very little.

What is worth acting on:

The same material, the same direction, repeatedly. That is a process fault — a conversion factor, a habitual unrecorded issue, a supplier who consistently delivers at one end of a tolerance.

Variance that appears after a change. A new storekeeper, a new site layout, a new material with an unfamiliar unit. The change is the explanation and the fix is training or design, not investigation.

Variance concentrated in one location while the same material is clean elsewhere. That narrows the question to a place and a group of people, which is the only situation in which the security conversation is well founded.

A variance that suddenly stops. Worth as much attention as one that starts.

None of those is visible in a single reconciliation. They are all visible in a year of small ones, which is the argument for frequency stated a different way.

Two things to stop doing#

Stop netting variances across materials. A shortfall of cement and a surplus of sand do not cancel. They are two separate facts and adding them produces a number with no meaning, which will nonetheless be reported.

Stop netting across periods. A shortfall one month and a surplus the next usually means a timing problem — an issue recorded in the wrong period, or a receipt posted late. That is worth knowing, and it disappears entirely if you only look at the cumulative figure.

What a good reconciliation produces#

Not a corrected balance. You would have got that from any adjustment.

A good reconciliation produces a short list of specific, dated, explained differences, and a shorter list of differences nobody can explain. The second list is the output, and it is one of the things that has to be settled before the books close — month-end at a builder's accounts desk. If it is empty every month, either your store is exceptionally well run or your counts are being reconciled to the book rather than the other way round, and it is worth knowing which.

The short version#

Book minus physical is a residual, and a residual contains every unrecorded event, not just the interesting one.

Count small, count often, adjust by entry rather than by edit, keep materials and periods separate, and look for repetition rather than magnitude.

The variance is a number. The pattern is the information.