Material Flow Analysis for Manufacturers

Manufacturers know what they buy in currency and what they ship in units. Very few can say what they buy in tonnes of material, where it goes, and how much leaves as something other than product.

Material flow analysis answers that, and the reason it is worth doing is that mass and money diverge. Scrap is cheap per tonne, so a substantial material loss can be financially invisible while being the largest circularity problem in the plant.

What is material flow analysis?

The mass conservation principle is what gives the method its power. Unlike most reporting, an MFA has an internal consistency check: if inputs do not equal outputs plus accumulation, something is missing and you know it before anyone else does.

Step 1 — Define the system boundary

Decide what is inside and outside, in space and in time. A single plant over one calendar year is the usual first choice, because it matches the period your other records already cover.

Resist the temptation to include the supply chain in a first analysis. Extending the boundary upstream multiplies the data problem and delays the finding, and the internal picture is usually where the surprises are anyway.

Step 2 — Identify the flows

The fourth and fifth rows are where first analyses fail. Internally recovered material is often not recorded because it never becomes waste, and unrecorded losses are the residual that the mass balance eventually exposes.

Step 3 — Convert everything to mass

Purchase records are in currency, units, litres, metres and sheets. All of it has to become mass, which is the least intellectually interesting and most time-consuming part of the exercise.

Bills of materials give unit masses for products. Supplier specifications give densities and areal weights. Waste transfer notes are usually already in mass because disposal is priced that way, which makes them the most reliable dataset in the whole analysis.

Step 4 — Run the mass balance

Total inputs should equal total outputs plus the change in stock. They will not, on a first attempt, and the size of the gap is the most informative number the analysis produces.

1. A gap under a few percent is usually measurement error and rounding.
2. A gap of five to ten percent usually means an unrecorded flow, often internal recovery.
3. A larger gap usually means a boundary problem or a missing waste stream entirely.
4. Inputs exceeding outputs consistently means accumulation nobody has accounted for.

Closing the gap is the analysis. Each iteration finds a flow that was not being recorded, and the flows that were not being recorded are almost always the ones nobody was managing.

Step 5 — Interpret in mass, not money

The point of the exercise is to see the picture at a different scale from cost accounting. A material representing two percent of spend can be twenty percent of mass, and the circularity intervention that matters is on the mass figure.

This is also why an MFA frequently changes priorities. Teams arrive expecting the answer to concern their most expensive input and find it concerns packaging, process water or an offcut stream nobody had costed because it was disposed of cheaply.

How does this connect to reporting?

ESRS E5 under Directive (EU) 2022/2464 requires disclosure of resource inflows and outflows, which is an MFA presented in a reporting format. Organisations that have run one find the disclosure straightforward; those that have not are usually assembling it for the first time under deadline.

The same data supports circularity indicators, recycled content verification and waste reduction targets, which is the argument for treating the analysis as infrastructure rather than as a one-off study.

Where product data helps

The input side of an MFA is a bill of materials problem. Where composition is held per product with accurate masses, converting production volumes into material inputs is arithmetic rather than estimation.

Where it is not, the conversion relies on assumed unit masses and average compositions, and the resulting uncertainty is frequently larger than the effects being measured. That is the practical link between passport data quality and any material accounting built on it.

How often should it be repeated?

Annually, once the first analysis has closed. The first is a project because the flows have to be discovered; subsequent years are a reporting exercise because the flow map already exists and only the quantities change.

Repeating it is what turns the analysis from a snapshot into a trend, and trends are what reveal whether an intervention worked. A single year tells you where the losses are; three years tell you whether anything you did about them had an effect.

It also catches drift. Process changes, supplier substitutions and new product lines all move material flows, and a flow map that is not refreshed silently describes a factory that no longer exists.

What to do with the result

Rank the losses by mass and ask, for each, whether it is avoidable in design, recoverable in process, or genuinely unavoidable. Those three categories lead to different owners: design engineering, manufacturing engineering, and nobody respectively.

The third category deserves scrutiny, because unavoidable frequently means unexamined. A yield loss accepted for years because it was always that way is the most common finding of a first analysis, and it is usually the cheapest to address once someone owns it.

Common mistakes

• Including only materials that appear in the product, and excluding process consumables.
• Counting internally recycled material twice, once as input and once as recovery.
• Using supplier specification masses rather than actual delivered masses.
• Extending the boundary upstream before the internal balance closes.
• Reporting the result without stating the boundary, which makes it incomparable.

The second is the most common and the hardest to spot, because regrind genuinely is both an output of one step and an input to another. Deciding once whether internal loops are counted, and stating it, prevents an analysis that quietly double-counts.