Running a site store so the numbers survive

A site store is a warehouse with none of a warehouse's advantages. It is temporary, often outdoors, staffed by one person who is also doing three other jobs, and its contents are removed by people in a hurry who consider paperwork an obstacle to actual work.

Most advice about store management assumes conditions that a site does not have. What follows is the part that survives.

Four things, in order of importance#

Everything in is recorded at the moment it comes in. Not that evening, not from the pile of challans. The record has to be made by the person who counted, while they are counting.

Everything out is recorded against something. A project, a location, a work item, a person. Material that leaves the store to no destination is material that has left your knowledge, and no amount of stocktaking afterwards will tell you where it went.

Every material has exactly one unit. One. Not "bags, or tonnes when it comes loose".

Corrections are entries, not edits. A number that can be quietly changed is a number nobody can rely on.

Get those four and the store is manageable. Miss any one and you are running a guessing operation with a ledger attached.

Why the issue side is the one that fails#

Receipts get recorded because there is a lorry, a driver, a challan and somebody waiting for a signature. The event has ceremony. It is hard to miss — provided the barrier knew the load was coming, which is the subject of why a lorry surprises the gate.

Issues have none of that. A carpenter needs six sheets of ply and takes six sheets of ply. Somebody needs another two bags of cement at four in the afternoon. There is no external party, no paper, and nobody standing there waiting. So the issue register gets written up "later", and later is where site records go to die.

The consequence is systematic and always in the same direction: book stock is higher than physical stock, because receipts are complete and issues are not. Then a stocktake shows a shortfall, and the shortfall gets called wastage or pilferage, when a large part of it is simply issues that were never written down.

If you are going to be strict about one half of a store, be strict about the half that leaves.

Make the issue cheap to record#

Two things make an issue register work.

It must be doable by the person taking the material, where they are standing. A register kept in a cabin that requires a walk is a register filled in from memory at the end of the day. A phone in the storekeeper's hand is worth more than a beautifully designed book on a desk.

It must accept the way people already talk. People do not say "issue quantity six, unit number, material plywood 12mm". They say "6 ply gaya, second floor". A record that demands a form's vocabulary gets a form's compliance rate.

We learnt this the hard way in our own tools. Our software recognised a photograph from site as a delivery only if the caption contained one of four words — challan, bilty, GRN, delivery. Measured against four hundred captions the site team had actually sent, those four words matched exactly two. What people really wrote was the material and a quantity, or the name of the vehicle that brought it, or a plain list of items. Reading those three shapes instead took the match rate from two to thirty-four out of the same four hundred.

The vocabulary was not the site's problem. It was ours. There is a longer piece about that in the words the site already uses.

One unit per material, and mean it#

A store that records cement in bags on Monday and in tonnes on Thursday does not have a stock figure. It has two stock figures added together.

Choose the unit that the material is physically handled in, not the one it is bought in or the one it is consumed in. Cement in bags, because bags are what get carried. Steel in kilograms if it is weighed and in pieces if it is counted, but never both, and if that means a conversion at the receipt stage then do the conversion at the receipt stage where somebody can check it.

Conversions belong at the boundary, once, with the factor written down. A conversion applied ad hoc, differently, by whoever is entering that day, is the most expensive kind of error because it is invisible in every individual record and only shows up as an unexplained gap in the total. It has its own piece: the unit of measure.

The correction problem#

The storekeeper types the wrong number. It will happen. Not occasionally — regularly, because the conditions are a phone, a lorry and the end of a long afternoon.

What matters is what happens next. If there is no way to reverse an entry, the correction goes on paper, into a note, and into the head of the one person who remembers which entry was the bad one. That is not a hypothetical; it is what we found in our own system, where the code to reverse a stock entry had been written and had never been given a button.

A reversal has to do four things or it makes matters worse:

• Write a reversing entry, not an edit, so the history stays readable.
• Be refused if the stock has since gone out. Reversing a receipt after most of it has been issued drives the balance below zero, silently, which is not a state a physical store can be in.
• Be restricted to people entitled to do it, which is not the same as "whoever's name is on the row".
• Carry a reason into the record. A reversal with no stated reason is an unexplained hole, and unexplained holes are what a stocktake turns into an argument.

Stocktaking without theatre#

Count often, count small, count unannounced. A full stocktake once a quarter produces one enormous unexplainable variance. Counting three materials a week produces small variances you can still trace to a week's worth of entries.

And when book and physical disagree, resist the urge to make the book match the floor with a single adjusting entry. Adjust, yes — but record the adjustment as its own dated entry with its own reason, so that next quarter somebody can see whether the same material keeps needing adjusting. That pattern is the actual finding. The number itself is not. There is more on this in physical stock against book stock.

The short version#

A site store fails on the way out, not the way in.

Make issues as easy to record as receipts are, fix one unit per material, never edit a number that has already been read by somebody, and count small and often.

The rest is shelving.