Skip to content

MANUFACTURING

When you can't prove what the product costs to make...

A standard cost set once drifts from the real one, because ingredients, scrap and yield all moved and none of them fed back.

CK
Caybo Kotzé 13 November 2026 · 7 min read

In July 2026 the FAO Food Price Index stood at 131.1 points, just 1.0% higher than the same month a year before. Read only the headline and a manufacturer might reasonably conclude that ingredient costs had barely moved this year. Underneath that calm number, vegetable oils were at 195.7, the highest reading since June 2022, and dairy was 24.8% lower than a year earlier. Two ingredients in the same basket, moving in opposite directions by double digits, and the single number a purchasing director quotes to the board is the average of both, which describes neither.

That gap between the headline and the ingredient is where a great deal of manufacturing margin quietly disappears, and it disappears without anybody lying about it. A mid-market manufacturer knows its revenue to the cent, every day, because the accounting system posts it in real time as the sale happens. The same manufacturer rarely knows what a single case of finished product actually cost to make on the day it was made, because that figure depends on a bill of materials, a scrap rate, a labour rate and an overhead allocation, moving at four different speeds.

The standard cost that stopped being true

Call a standard cost what it actually is: a bet, made once, that the relationship between ingredients, waste and price will hold roughly steady until somebody revisits it. For a long stretch of the last two decades that bet was reasonable, because input costs mostly moved together, a rising or falling tide that one blended assumption could describe well enough for pricing purposes. It is not describing anything well now.

Farm Credit Canada reported egg processor base prices up 63% from the start of 2024, in the same period that sugar prices were expected to fall roughly 1.0% in 2025 and a further 5.6% in 2026. A recipe weighted towards eggs and a recipe weighted towards sugar are now on entirely different cost trajectories, and a single standard cost applied to both is wrong for at least one of them by construction.

The Institute for Supply Management's prices index has held above 70 for six consecutive months this year, and Deloitte's manufacturing outlook has US manufacturers expecting input costs to rise 5.4% on average over the next twelve months. Averages are the problem here, not the solution: a blended 5.4% assumption applied across a hundred SKUs is wrong for most of them, wrong in a different direction for each, and it stays wrong until somebody sits down with a spreadsheet and recalculates by hand, typically at month end, by which point three more weeks of drift have already happened underneath the number everyone is still pricing against.

Why discounting evenly is the real symptom

Here is the part that gets missed in most conversations about thin margins. The consequence of a stale standard cost is not that margins are uniformly thin, because they are not; some products in a typical range are genuinely profitable and some are quietly being sold at a loss, and a cost that has not moved since the last review cannot tell anyone which is which.

What actually happens instead is that discounting gets applied evenly across the whole range, because evenly is the only defensible policy available to a sales team with no product-level signal to work from. Give every account manager the same discount authority and nobody can be blamed for a specific bad call, even though the same percentage discount lands completely differently on a high-margin line and a low-margin one.

Argue it from the sales team's incentives and the pattern stops looking like carelessness and starts looking like the predictable output of the system it sits inside. A rep is measured on volume and on the relationship, not on a margin figure that finance will not hand over with any confidence, because finance does not trust the number either and updates it too rarely to defend it in a live pricing conversation. Given a stale number on one side and an impatient customer on the other, the rational choice is to discount and let the standard cost absorb the consequence later, in a report nobody reads until the quarter has closed. Nobody inside that system is behaving badly; the incentive is set entirely by the fact that nobody has a trustworthy number at the moment the discount gets granted.

The obvious objection is that a monthly recalculation is close enough, that finance already reconciles standard cost against actual cost at period end and adjusts pricing guidance from there. It would be close enough if the gap moved in one direction, at one speed, across the whole catalogue, the way it roughly did when input costs still rose and fell together. It does not do that anymore. A monthly number is wrong for most of the month it covers, and, per the FAO figures above, it is wrong by a different amount and in a different direction depending on which components a given product actually contains, which means the correction carries the same flaw as the number it is correcting.

What a case actually costs, calculated instead of remembered

Miridia treats cost as a calculation rather than a number somebody types into a field once a quarter. The bill of materials itself stores no prices at all; it stores a structure, a list of component products, the quantity of each one required, and a scrap factor set per line rather than blended across an entire product range.

When a cost is requested, Miridia walks that structure and reads each component's purchase price from its own price history as it stood on the date in question, so a case made in March is costed against March's component prices and the same case made in June is costed against June's, rather than a single figure carried forward between reviews.

Labour, freight and overhead are added as production cost lines that belong to the assembly itself rather than to any one component, each fixed per unit, per cent or per hour, with a general ledger account code attached so the figure reconciles cleanly through the accounting sync instead of needing to be rebuilt by finance from scratch every time somebody asks. A price rise from a single supplier ripples immediately into every product that uses that component, rather than sitting unnoticed until the next scheduled review finds it by accident. Multi-level assemblies explode through their nested components the same way, so a finished good built from a sub-assembly built from raw materials still resolves to one number, not three spreadsheets stitched together by hand.

None of this stays theoretical once production actually happens on the floor. A work order snapshots the bill of materials at the moment it is created, so a recipe revision made on Thursday cannot quietly rewrite the cost of a batch that started on Tuesday, and a bill of materials itself carries revision numbers, effective dates and expiry dates, so a recipe change is a dated event rather than a silent overwrite of history.

Completing the work order posts both sides of the transaction automatically: components consumed out of inventory, finished goods produced into it, valued at cost of goods using weighted average cost with a price-history fallback where the weighted average is not yet established. Roll those work orders up by SKU and by product line and profitability reporting stops being a quarterly project assembled under deadline pressure, and becomes a query anyone can run on a Tuesday afternoon.

The honesty clause: this only runs on receipt discipline

None of this is a free calculation, and presenting it as one would be dishonest. Dated cost calculation is only as current as the price history it reads from, and that price history is built, line by line, from goods received notes posted when stock actually arrives and from supplier invoices captured when they actually land in the system, rather than when somebody eventually gets round to them.

A goods received note posts straight to the inventory movement ledger and updates weighted average cost automatically, which is exactly what makes the whole mechanism work, and it is also exactly the point at which the mechanism depends on a person doing their part of the job on time rather than in a batch at the end of the week.

The Hackett Group found procurement teams expecting total workload to rise 9.8% against only a 1.0% increase in staff, a productivity gap that has to be absorbed somewhere in the process, and the somewhere is very often the speed at which a supplier invoice gets keyed into the system.

A manufacturer that runs three weeks behind on receipt costing will get a cost figure that is calculated correctly and still three weeks late, which is a smaller problem than a cost that was never updated at all, but it is not no problem, and we would rather say that plainly than let the phrase "dated cost calculation" imply a currency it cannot deliver on its own. Automating the calculation raises the value of getting the receiving right; it does not remove the need to do it.

Where this argument stops working

Four situations where the argument above does not apply to your business, and where a different system is the honest recommendation rather than this one.

  • Your compliance obligation is lot and batch traceability. Miridia's movement ledger and BOM versioning record what moved, when, in what quantity and under which document, but they do not assign a traceability lot code or track which lot went into which finished unit. A business that must produce lot-level genealogy for a regulator needs a system built for that specific obligation, not this one.
  • Your production yields several valuable outputs from one run in a fixed ratio. An abattoir or a refinery needs multi-output work orders and joint cost allocation across those simultaneous outputs. Miridia's work orders are single-output by design; shared upstream production has to be modelled as an intermediate product, which covers a decoupled process cleanly but does not solve a genuinely joint one.
  • You need cost allocated per operation from work-centre rates. Labour and overhead inside Miridia are cost lines carried on the bill of materials, applied per unit, per cent or per hour. There is no routing of parts between machines at their own rates, so if your cost accuracy problem is really a shop-floor scheduling and capacity problem wearing a costing costume, this does not solve it.
  • You need the financial core replaced, not fed. Miridia integrates with accounting platforms and orchestrates the postings between them; it is not a general ledger and it does not perform statutory consolidation. If the underlying project is genuinely an ERP replacement, say so at the outset, because the scope of that decision is different from anything argued above.

Disclosure

We sell Miridia, and the bill of materials, the work orders and the dated cost calculation described above are a real, shipped product, not a hypothetical one built for this article. Read everything above with that fact in view, and weigh the specific claims rather than the general argument, because the specific claims are the ones we are willing to be checked on and corrected against.

If you cannot currently say what your best-selling case actually cost to make last month, that question is worth an hour of anyone's time before anything else on this page is. Book a discovery call and bring one bill of materials with its last few supplier invoices; we will calculate the case cost against your own numbers on the call, not a demonstration dataset built to look good.

Sources