Quality departments ask for resources in engineering units and are answered in financial ones. A coordinate measuring machine, a calibration contract and a training day appear on the profit and loss account as pure cost, with no revenue line attached. Present the same request against a stack of invoices for external sorting, premium freight and customer chargebacks, and it becomes something else entirely: a proposal to recover cash that is already leaving the business.
Timing is the deeper problem. Prevention spend lands in this year's budget, whilst the failures it prevents would have landed in later years, scattered across scrap, expedited freight, containment and warranty provisions. A trained financial controller sees that asymmetry immediately. Our task is to bridge it: show the historical cash from the ledgers, model the forward cash the control interrupts, and discount both streams so the comparison is honest rather than rhetorical.
I have watched sound cases die in budget meetings because they were argued in defect counts, audit findings and customer scorecards. Nobody in the room disputed the engineering; they simply had no way to price it. Everything that follows is about building that price, from sources the finance function already trusts.
Start with the ledger you already have
Before modelling anything, extract the actual failure spend from the accounts: scrap tickets raised in the ERP, rework labour bookings, third-party sorting invoices, premium freight, deductions on the customer portal, and warranty claims split into parts and dealer labour. Map each item to its general ledger account so the figures reconcile to the trial balance. This baseline is not your number argued against theirs; it is their number, taken from their own books, which is exactly why it holds up.
Engineer the baseline properly, because scrap is not uniform. A die casting rejected for porosity at the final leak test carries the casting, every machining cycle, deburring, washing, handling and the overhead absorbed to that point. Rejected at the first operation, it is little more than raw material. Value each scrap ticket at the routing step where it was raised rather than at a flat average, or you will understate precisely the failures that matter most.
Then sort the ledger twice: once by failure mode and once by part number. In most plants a small set of part numbers carries the bulk of the sorting and freight invoices, and one or two failure modes — a plating adhesion problem, a heat-treatment hardness escape, a recurring torque-strip complaint — dominate the warranty entries. That concentration tells you where a marginal unit of prevention buys the most recovery.
Model avoidance as a priced counterfactual
Cost avoidance is the difference between two futures: one with the control in place, one without. Finance distrusts the concept, because the benefit never shows up as cash in — only as cash that never leaves. The remedy is discipline about anchors. Every avoidance claim in the budget case should trace back to an event that actually happened, with invoices attached, rather than to an industry benchmark nobody in the room can verify.
Take the last three escapes of a given type and total them: the sorting house invoices, the air freight, the chargeback, the quarantine handling, the engineer's time on site at the customer. That sum is the realised cost of a single escape, and it needs no adjustment to be credible. Round nothing, dress nothing up; the audit trail is the argument. Average across the events if their sizes differ, and keep the range visible rather than hiding it in a point estimate.
Frequency comes next, and it must come from your own escape history rather than a textbook multiplier. If a hardness escape has reached the customer twice in the past five years, model it at that observed rate and state the source openly. Then apply an effectiveness assumption honestly: no control catches every instance, so haircut the benefit accordingly. A modest claim with visible arithmetic survives scrutiny; a heroic one is dismissed whole, and it drags the rest of the case down with it.
Building the avoidance claim
- 01Extract failure spendPull scrap, sorting, freight, chargebacks and warranty from the ERP and reconcile to the trial balance.
- 02Value realised escapesTotal the last three events of each type from invoices, keeping the range visible.
- 03Set observed frequencyUse your own escape history, stated with its source, not a benchmark multiplier.
- 04Apply effectiveness haircutReduce the benefit for controls that cannot catch every instance.
- 05Discount both streamsCompare prevention spend and avoided failure cash on finance's own terms.
What the same defect costs at each detection point
A defect has no fixed price; it is priced by where it is caught. Porosity found at the machine is one scrapped casting and a few operator minutes. The same porosity at final leak test is a finished component with full conversion value written off. Caught at the customer's goods-inward inspection, it becomes containment: a third-party sorting team billed by the hour, quarantine of every pallet on site and in transit, expedited replacement stock and a chargeback on the next statement.
In the field the categories multiply again: the replacement part, dealer labour at published rates, the investigation, the travel, and — where the affected batch cannot be bounded — a stock sweep across territories. Every category has an invoice somewhere in your files. Build a simple table of detection point against cost categories triggered, populated from those invoices. The progression argues itself, because each column represents cash the business demonstrably paid.
That table is the engine of the prevention argument. Any proposed spend — a leak test cell, incoming hardness checks on safety fasteners, a supplier surveillance visit — can be placed against it: which detection point it moves a defect from, and which it moves it to. Shifting detection from the customer's dock back to your own final test is a transfer between two priced columns, and transfers can be summed.
Where the money sits
Prevention payback that stands up on paper
For capital items — an additional measuring machine, a vision system, an extra test cell — present payback within the depreciation life, using the conservative avoidance case as the benefit stream. For recurring spend — calibration contracts, supplier visits at the source, corrosion cycle testing before design freeze — show recurring benefit against recurring cost. A controller will not fund an annual programme out of a one-off saving, and pretending otherwise squanders the credibility you have built.
Choose the worked examples deliberately. Incoming hardness verification on safety-critical fasteners costs little to run and interrupts a failure mode — hydrogen embrittlement after plating, a missed temper — whose escape cost is severe and whose liability exposure is worse. A second shift on the measuring machine converts sampled inspection into full coverage of a critical characteristic, and the case rests on the escapes your own record shows slipping past sampling. Durability rigs catch fatigue cracks whilst a fix is still a drawing change rather than a field campaign.
Discount the benefit stream as finance would, and get there first. Where the improvement releases warranty provisions, say it in balance-sheet language: lower accruals mean freed working capital. That single sentence, delivered correctly, earns more ground in a budget review than a wall of defect-rate trend charts, because it speaks to cash and risk rather than to programme performance.
Never present prevention as free. The new machine needs calibration, maintenance, programmed inspection cycles and people to run it; the training day needs backfill labour. Net these running costs against the benefit inside the model itself. Controllers extend real trust to cases that carry their own costs on the same page as their benefits.
Stress-testing before the controller does
Assume the baseline will be attacked. The most common strike is double counting: some scrap already sits inside standard cost, so presenting it as recoverable spend overstates the case. Pre-empt this by splitting the ledger into spend absorbed in standards and spend outside them. External sorting invoices and premium freight almost never touch standard cost, which makes them the safest pillars of the whole argument.
Frequency claims drawn from a single bad year are another exposed flank, so present a run-rate across several years and flag the anomalies openly. Watch also for rework labour valued at full rate when those operators would otherwise have been idle. Concede the small points before they are raised; a case that argues against itself on the minor items keeps its standing on the major ones.
The strongest line in any budget review is arithmetic, not projection: this programme breaks even if it prevents one contained escape per year, at the invoice value shown in the appendix.
Alongside the break-even statement, show the do-nothing case — the trend in failure spend and the inspection capacity implied by the volume forecast — because standing still has a price too. Controllers respect a proposal that prices its own alternative honestly, and the do-nothing trend is usually the easiest number in the pack to defend, since it is simply the past projected forward with no assumptions attached.
What actually goes into the room
One page carries the decision: the ledger summary by cost category, the ask split into capital and operating, payback on conservative assumptions, the break-even statement and the sensitivity table. The appendices carry the evidence — sample invoices, the ledger mapping, the assumption log with a named source against each figure. Detail is not weakness; it is what separates a proposal from a wish.
Offer a feedback loop before it is requested: quarterly reporting of realised failure spend against the modelled case. The first budget built this way is an experiment; the second is a track record, and the third is a habit the finance function defends on your behalf. Credibility compounds across cycles precisely because so few quality functions ever close the loop.
Throughout, argue in their grammar: margin protected, provisions released, payback within the depreciation life, working capital freed. Keep parts per million and audit scores for the engineering reviews where they belong. The budget meeting is not won by enthusiasm or by risk warnings alone; it is won by a priced counterfactual whose arithmetic survives the most suspicious person in the room.