A material balance is the first manufacturing improvement project

Two decisions determine whether a balance is useful or merely arithmetic. The first is the boundary. A balance drawn around a single production line will show different losses from one drawn around the whole site including warehousing. Neither is wrong, but mixing them is. State whether you are balancing a line, a department, a site or a product family, and keep it consistent between periods.

The second is the basis. Material can be measured wet or dry, and the difference is water. A process that buys material at, say, 8 per cent moisture and sells a product at 2 per cent will show an apparent gain if inputs are recorded wet and outputs dry. That is not efficiency; it is a units error. Where moisture matters, fix the basis explicitly — usually dry tonnes — and convert consistently.

The same discipline applies to units. Record whether figures are in tonnes, kilograms or units, and whether a tonne means a metric tonne. A balance that mixes units will not close, and the resulting confusion is often what convinces a team that the exercise is not worth doing.

The most common reason a first material balance fails to close is that it ignores inventory. Material purchased in a period is not the same as material consumed in that period — the difference is the change in stock of raw material, work in progress and finished goods.

The correct relationship is that opening stock plus purchases minus closing stock gives material available for consumption. Skipping that step produces an apparent loss in a period when the business was building inventory, and an apparent gain when it was drawing down. Both are artefacts, and both destroy confidence in the exercise.

This is why a frozen closing rule matters. Stock must be counted consistently — same time of day, same cutoff for goods in transit, same treatment of material issued to the line but not yet processed. A balance built on inconsistent counts is worse than no balance, because it produces confident numbers that are wrong.

Scrap is usually reported at material cost, which understates it. Material that has been through several process steps carries the labour, energy and machine time already invested in it. A tonne of scrap arising at the end of a long process route destroys more value than a tonne arising at the first cut.

A more useful view is to value scrap at the cost of the furthest process step it reached, then subtract any recovery value. That reframing often changes which loss a factory chooses to attack first, because it reveals that a small tonnage arising late in the route can outweigh a larger tonnage arising early.

It also connects the material balance to the financial accounts. The balance explains the volume; the cost model explains why that volume matters. Together they give a factory a defensible prioritisation, which is what the exercise is for.

A residual tells you that material is unaccounted for. It does not tell you why. Reason codes do. A short, closed list — process loss, contamination, changeover purge, handling damage, measurement error, unknown — is enough to start. The discipline is that every reconciliation line above the agreed tolerance must carry a code, and that 'unknown' is permitted but tracked.

Owning the balance is a role, not a project. A monthly balance compiled by whoever has spare capacity will not survive a busy quarter. A named individual, a fixed weekly slot of perhaps thirty minutes, and a one-page output that goes to the operations lead is what sustains it.

Once the balance is stable, its outputs become the specification for anything else you buy. You will know which flows need better instrumentation, which losses need reasons and which period comparisons are meaningful — and you will be able to tell a software vendor what you need measured, rather than accepting what they happen to measure.