Every plant has a rush order story. Ours arrived on a Tuesday, flagged red in the ERP, offering a premium price for a compressed lead time on a machined housing assembly the customer needed airfreighted by Friday. Commercial looked at the margin, sales looked at the customer, and nobody looked at the second shift that would have to run it on machines that had never processed that part in that sequence. We accepted it.

The premium covered expedited freight and a shift premium, according to the quote. The quote had no line for scrap, rework, a containment inspection at the customer's dock, or the engineer who spent most of a week writing the corrective action. When I pulled those costs together, they exceeded the premium. That is the quiet arithmetic of schedule pressure: revenue arrives visibly, defects arrive invisibly, and no accounting system is built to marry the two.

The argument here is not that rush orders are forbidden. Some are legitimate: a launch escalation, a line-down at a strategic customer. The argument is that if you cannot measure the defect cost of schedule pressure, you will keep taking orders that cost more than they earn, and you will take them confidently, with a signed quote in hand.

Where Overtime Shows Up in the Product

The correlation between extended overtime and defects is real, but it is not mystical fatigue arithmetic. It shows up in specific, nameable failure modes. The first is missed in-process checks. When a cell is behind, operators skip the interim verification steps that feel optional: first-off inspection after a tool change, periodic bore gauge checks on a turning cell, torque verification on the fifth fastener of a pattern. I have audited weekend runs where the first-off was signed off for an entire batch on the strength of one measurement taken two hours earlier, before a carbide insert was swapped.

The second failure mode is substitution. Under pressure, people reach for the available rather than the correct: a gauge that is due for calibration but still on the rack, a slightly different torque driver because the right one is in use on the day shift, a bin of fasteners that looks identical to the one on the print. None of these decisions is documented at the time, which is exactly why the escape later becomes expensive to trace. The paper trail says the right things happened; the physical part says otherwise.

Third comes training debt. A rush order on a Saturday is staffed by whoever volunteered or whoever could be pressed into it, and that person is frequently not the operator who normally runs the operation. Setup errors climb: wrong offsets, wrong fixture orientation, a deburr operation performed before rather than after a cleaning step. Each produces defects at a rate the normal shift does not, and weekend quality data, if you keep it separately, will show it plainly.

The defect is rarely decided at the moment it is found; it was decided shifts earlier, in the checks that were skipped to make the date.
The defect is rarely decided at the moment it is found; it was decided shifts earlier, in the checks that were skipped to make the date.

What to Measure Before You Can Argue

You cannot manage this without separating the data. The single most useful step I ever took was asking production control to tag every work order with a schedule-pressure flag: expedite requested, overtime required, or normal flow. From there the comparison writes itself: defect rate per thousand pieces on flagged versus unflagged orders, scrap value, rework hours, and customer escapes, all sorted by the same flag.

Tagging alone is not enough; the measurement must be honest about attribution. A defect found at final inspection on a rush order may have originated in a machining operation that ran normally, or vice versa. Attribute the defect to the work order and the shift pattern together, which means labour reporting must capture which shift performed which operation. Most modern MES systems already do this; the failure is that quality teams rarely ask for the join. Ask for it.

Beyond rates, measure the escape ratio, internal finds versus external finds, separately for expedited and normal orders. The most damaging signature of a rushed order is not a higher defect rate but a shifted one: the same defects, found later, after value has been added and freight has been paid. An escape that triggers sorting at a customer's receiving dock costs an order of magnitude more than the same defect caught at the machine, and expedite conditions reliably degrade the internal catch rate because inspection is itself compressed.

The measurement set

DPMODefects per thousand, flagged vs normalThe core comparison; normalise by volume
2xRework hours per orderFlagged orders consume disproportionate touch time
10xExternal vs internal find costAn escape at the customer dock costs roughly an order of magnitude more
3Flag categoriesExpedite requested, overtime required, normal flow
Four numbers, split by schedule-pressure flag, are enough to start the argument with finance.

The Hidden Channels of Cost

The visible costs, scrap, rework hours and freight, are the easy part. The expensive channels are quieter. Containment is one: once a customer finds an escape from an expedited shipment, they will often demand one hundred per cent inspection on subsequent lots, sometimes for months. That inspection labour is charged back or absorbed, and rarely attributed to the original order. Sort a few thousand parts at a customer's site and the premium on the rush order that caused it becomes a rounding error.

Then there is the equipment toll. Compressed schedules tempt supervisors into running machines harder: shorter or skipped preventive maintenance windows, tooling pushed past its change interval because the change costs twenty minutes nobody has. A marginal insert run to failure does not produce one bad part; it can score a bore across a whole fixture load before anyone notices the surface finish change under shop lighting. Tooling overrun is the most under-recorded defect source in expedited production, because the scrap it creates gets booked against the wrong cause.

Capacity theft is the last channel and the least discussed. Every hour of engineering, quality and skilled trades time spent firefighting a rushed order is an hour not spent on the work that prevents the next fire. I have watched a launch readiness review get postponed twice to support expedited shipments, and the defects that escaped the postponed review cost more than both rush orders combined. You will never see that on a quote, and you will never see it on a variance report either, which is why it survives.

Revenue arrives visibly, defects arrive invisibly, and no accounting system is built to marry the two.

Building an Expedite Policy With Teeth

Once you can measure, you can govern. A workable expedite policy has three gates. The first is a feasibility review genuinely empowered to say no: not a rubber stamp from the planner who will be blamed either way, but a documented check that qualified operators are available for the required operations, that the inspection points in the routing can physically be performed at the compressed rate, and that tooling and gauge capacity exist. If any of those fail, the honest answer is a longer lead time or a reduced quantity, not a hope.

The second gate is pricing that includes the measured defect cost. When tagged data shows that expedited orders of a given product family historically generate a known amount of scrap, rework and containment, that amount belongs in the quote as a line item, not buried in overhead. This changes the conversation with sales: the premium stops being margin illusion and starts reflecting reality. On more than one occasion, presenting the true cost internally led the commercial team to renegotiate quantity or date rather than the plant eating the difference.

The third gate is a sunset rule. Expedites authorised as exceptions have a habit of becoming routine, because the customer learns the date is negotiable and the plant learns the red flag works. Review expedite volume monthly, by customer and by cause. If a customer's demand pattern generates chronic rushes, the correct response is a capacity or stocking discussion, not another year of Friday airfreight.

The three-gate expedite decision

  1. 01Feasibility reviewOperators, inspection points and tooling verified as available at the compressed rate
  2. 02Priced defect costHistorical scrap, rework and containment for the product family loaded into the quote
  3. 03Sunset trackingMonthly review of expedite volume by customer and cause; chronic patterns trigger a capacity discussion
Each gate is a documented check; the order only proceeds if all three pass, and the pricing gate carries the measured defect cost.

Making the Case in Money

The final piece is translation. When I presented our first schedule-pressure analysis, the scatter of defect rates meant nothing to the finance director until it was converted into currency: scrap value plus rework labour plus containment costs plus freight recoveries, summed by flag. The expedited population carried a disproportionate share of total quality cost despite being a small share of volume. That is the sentence that gets attention, and you can only write it honestly if the tagging discipline was in place long enough to be statistically respectable.

Present it with humility, though. Correlation is not causation, and some rush orders are rushed precisely because they are already abnormal: engineering changes, new part numbers, difficult customers. Control for this by comparing expedited orders only against normal orders of the same part family in the same period. In my experience the effect survives that control. It is not the part difficulty driving the defect rate; it is the compression.

Then close the loop on decisions. Track, for a defined period, the margin on expedited orders against their fully loaded quality cost, and put that figure in front of whoever authorises the next one. After we did this, the approval rate for expedite requests fell, not because a policy forbade them, but because the people signing them could finally see what they were signing. A few orders still go through, and rightly so. The ones that remain are the ones that actually earn their keep.

Where to Start on Monday

Do not begin with the policy; begin with the flag. A schedule-pressure tag on the work order is a one-line change in production control's routine, and without it every subsequent argument is anecdote. Give it three to six months of clean data before you present anything, and resist the temptation to extrapolate from the first bad weekend.

In parallel, walk your own weekend data. Pull the quality records for the last ten overtime runs and compare them against normal flow on the same part families. In most plants the pattern appears within the first afternoon of looking: the same defect codes, at higher rates, with a shifted internal-to-external find ratio. That half-day of analysis is what earns you the credibility to ask for the systemic tagging.

Finally, choose the audience before the chart. The engineering conclusion is a defect-rate comparison; the management conclusion is a currency figure per flag. Bring both, but lead with the second. Across two decades in automotive and aerospace, I have never seen a rush-order argument won on defect rates alone; it is won when the premium and the bill appear in the same column of the same table.