How to set tolerance thresholds for MRO consumables

Executive Summary

Most MRO tolerance failures are not a math problem. They are a design problem: one flat percentage applied across a catalog that ranges from a $2 fastener to a $4,000 pump seal. A single rule cannot serve both, so it ends up serving neither, clearing real overcharges on the high end and flooding review with immaterial noise on the low end.

The fix is to tier the threshold by unit price and by SKU volatility, not to pick a better single number. A fastener that moves a few cents needs a wide percentage band and a low dollar floor. A capital-adjacent repair part needs a tight percentage band because a few percent on a high-dollar line is real money.

What changes this is treating the threshold as a control you maintain, not a setting you configure once. Price files move, commodity inputs move, and a threshold set against last year's catalog is already stale.

1. What is a tolerance threshold in MRO invoice review?

A tolerance threshold is the variance between an invoice's unit price and the contracted or last-approved price that a control allows before flagging the line for review. Below the threshold, the line clears automatically. Above it, someone checks whether the difference is a legitimate cost change or an overcharge. The threshold is a decision rule, not a report.

In MRO and Class C categories, the threshold usually needs two components rather than one: a percentage variance and a minimum dollar amount. A percentage alone breaks down at the extremes of a catalog that spans fasteners, safety supplies, and repair parts priced two or three orders of magnitude apart.

A 5% variance on a $1.50 item is 7.5 cents, immaterial no matter how consistently it recurs. The same 5% on a $2,000 part is $100, worth a look every time. Pairing the percentage with a dollar floor and a dollar ceiling keeps the rule meaningful across the whole catalog instead of just the middle of it.

The rule also needs a direction. Most controls only flag overages, price above contract or last-approved cost. Some AP teams also flag unexplained decreases, since a price drop that never reached the rate card is a rebate or credit opportunity, not just a pricing curiosity.

Without an explicit rule, AP teams default to checking everything or checking nothing. Neither is a threshold. Neither is a control.

2. How do you choose between a dollar tolerance and a percentage tolerance?

Use both, applied together rather than as alternatives. The percentage catches proportional overcharges on mid- and high-value lines. The dollar floor stops the percentage rule from flagging trivial variances on low-cost items, and a dollar ceiling stops a wide percentage from letting a large-dollar error pass unreviewed. Neither rule alone covers a Class C catalog.

A percentage-only rule fails at both ends of the price range. On a $0.40 washer, even a 50% swing is 20 cents, not worth an AP reviewer's time. On a $5,000 gearbox component, a percentage band wide enough to absorb normal cost noise can let a four-figure overcharge pass without a flag.

A dollar-only rule has the opposite problem. A flat $10 threshold is too tight for high-value parts, where legitimate cost movement of $10 happens constantly, and too loose for the bulk of Class C spend, where $10 is often the entire line value.

The combination works because each rule covers the other's blind spot. State the logic as: flag if variance exceeds X% AND variance exceeds $Y in absolute terms. Both conditions have to be true. That keeps a 40% swing on a 60-cent item off the exception queue while still catching a 3% swing on a $3,000 item.

3. Should MRO and Class C consumables use the same threshold?

No. MRO parts and Class C consumables behave differently enough to warrant separate bands. MRO parts are lower volume, higher unit value, and tied to specific equipment, so price changes are traceable to a cause. Class C items are high volume, low unit value, and substituted often, so the threshold has to tolerate more routine noise per line.

Treating the two categories as one line item type is a common source of both missed drift and wasted review time. A repair part for a specific machine has a contract price, a defined vendor, and a clear reason for any change: a supplier cost increase, a substitution, a different pack size. Variance there is worth investigating individually.

Class C consumables (gloves, fasteners, tape, general shop supplies) are ordered in volume, often from a catalog that gets partial updates throughout the year. A single SKU might show small variances constantly as pack sizes or suppliers shift within an approved list.

A workable split: tighter percentage bands and a lower dollar floor for MRO repair parts, where each line is traceable to a specific asset or work order. Wider percentage bands and a higher dollar floor for Class C consumables, where the volume is too high to review line by line and the goal is catching only the lines large enough to matter in aggregate.

• MRO repair parts: Lower volume, higher unit value, tied to a specific asset. Tighter percentage band, lower dollar floor, individual review.
• Class C consumables: High volume, low unit value, frequent substitution. Wider percentage band, higher dollar floor, review in aggregate.
• Capital-adjacent parts: Occasional high-dollar items inside an MRO catalog. Treat with a dollar ceiling regardless of percentage, since even a small percentage swing is material.

4. What data do you need before setting a threshold?

You need three inputs: the contracted or last-approved unit price for each SKU, a rolling price history long enough to see normal variance, and the unit of measure and pack size for each line. Without the last input, a threshold check will flag pack-size changes as price changes and generate false exceptions.

A threshold set from a single reference price is fragile. Vendors substitute pack sizes, switch between each-priced and case-priced billing, and update SKU numbers without notice. If the control compares invoice price to a static reference without normalizing for unit of measure, most of its flags will be unit mismatches, not real drift.

A rolling price history, even just the last several invoices for a SKU, tells you what normal variance already looks like before you set a band around it. A threshold set purely from the contract, with no reference to what has actually been invoiced, tends to either mismatch what the vendor's own system does or ignore drift that has already crept in.

The vendor master and price file are the other half of this. A duplicate vendor record or an outdated price file entry produces variance that looks like an overcharge but is really a data hygiene problem. Cleaning that layer first makes the threshold results trustworthy.

5. How does price file volatility affect the threshold?

A threshold set once against a snapshot of input costs goes stale as those costs move. Machinery and equipment input costs, tracked by the US Bureau of Labor Standards, rose 5.6% year over year as of July 2026 (BLS Producer Price Index, series WPU114, read 2026-09-05), which is enough movement to shift what a normal invoice variance looks like within a single year.

A tolerance band calibrated when input costs were flat will generate false positives once costs start moving, or worse, will get widened in response and stop catching real overcharges. The band needs a periodic reset against current cost trends, not a one-time calibration.

This does not mean tying the threshold directly to a commodity index. Few MRO contracts price that way. It means recognizing that a rising input cost environment raises the baseline of what counts as a normal increase, and a static threshold set during a calmer period will misclassify legitimate cost pass-through as an exception, training reviewers to ignore the exception queue.

The practical response is a scheduled review of the threshold itself, tied to the same cadence as a broader price file governance process, rather than leaving it fixed indefinitely.

6. How often should the threshold be reviewed and adjusted?

Review the threshold on a fixed schedule, at minimum quarterly, and any time a major vendor renegotiation or price file update happens. A threshold is a control setting, not a one-time configuration, and leaving it untouched for a year is the most common way it stops catching what it was built to catch.

A quarterly review checks three things: whether the exception volume is manageable for the AP team reviewing it, whether the exceptions being generated are turning into real findings or being dismissed as noise, and whether the underlying price file has been updated since the threshold was last set.

If the exception rate is climbing and the dismissal rate is also climbing, the threshold is too tight for current conditions and needs widening, or the underlying price file needs updating before the threshold is touched. If exceptions are rare but real findings keep surfacing through other channels, like a rebate reconciliation or a vendor statement review, the threshold is too loose.

Tying the review to the same calendar as a broader margin drift review keeps it from being forgotten. A threshold nobody revisits eventually measures nothing but its own age.

7. What should happen when an invoice line falls outside tolerance?

An exception should route to a defined reviewer, with the contracted price, the invoiced price, and the SKU's recent price history attached, and a decision should be logged either way: approved as legitimate, corrected, or escalated to the vendor for credit. An exception with no logged resolution is not a control, it is a queue.

The most common failure after a threshold flags a line is not a wrong decision, it is no decision. The line sits in a queue, the invoice still gets paid on schedule because payment terms do not wait for review, and the exception is never resolved either way.

A workable process separates the payment decision from the resolution decision. The invoice can be paid on terms while the exception is investigated separately, with the outcome, correct, error, or credit due, recorded against that vendor and SKU. That record is what makes the next quarter's threshold review meaningful, since it shows whether flagged lines are turning into real recoveries.

For a repeated variance on the same SKU, the resolution should reach back to the price file itself, correcting the reference price so the same error does not generate a new exception every cycle.

For the wider pattern this sits inside, start with the margin drift guide.

8. Frequently Asked Questions (People Also Ask)

What percentage tolerance is standard for MRO invoices?

There is no single standard percentage across companies, since a workable band depends on your catalog's price spread and your AP team's review capacity. Set it from your own price history and unit value tiers rather than adopting a generic figure, and pair any percentage with a dollar floor and ceiling.

Should Class C items even be reviewed line by line?

Generally no. Class C volume is too high for individual line review to be practical. Set a wider tolerance band with a higher dollar floor so the control only surfaces variances large enough to matter in aggregate, and review those in batches rather than line by line.

How do pack size changes affect a tolerance threshold?

A pack size change looks identical to a price change if the control compares invoice price without normalizing for unit of measure. Normalize every SKU to a per-unit price before applying the threshold, or the exception queue fills with false flags that have nothing to do with pricing drift.

Can a tolerance threshold be automated inside an ERP?

Most ERPs support a basic percentage or dollar tolerance on three-way matching, but that check typically compares invoice to purchase order, not invoice to the underlying contract or rate card terms. A threshold built only on the PO comparison will miss drift that was already baked into the PO.

What is a reasonable dollar floor for a Class C tolerance?

There is no fixed figure that applies across companies, since it depends on your catalog's typical unit price and your team's tolerance for reviewing small-dollar lines. Set the floor by looking at your own price history: high enough to exclude routine cents-level noise, low enough to still catch recurring small overcharges that add up.

Does a tolerance threshold replace a periodic audit?

No. A threshold is a forward-looking control that catches new invoices going out of bounds. It does not surface drift already embedded in invoices paid before the threshold existed, which is what a periodic audit is for.

Who should own the tolerance threshold, AP or procurement?

Both need a role. AP typically owns the mechanics of the check and the exception queue, while procurement or the category owner should own the reference price the check is comparing against, since they hold the contract and vendor relationship.

What happens if the threshold is set too tight?

AP review time gets consumed clearing immaterial variances, and reviewers start dismissing exceptions without real investigation because the volume is unmanageable. A threshold that generates more noise than signal trains the team to ignore it, which defeats the control.