A three-way match can pass when the invoice and receipt agree with a PO that contains an outdated commercial price. The next check is whether the PO reflects the terms that apply to that material, quantity, entity, and transaction date.
That does not mean three-way matching is broken. It means a matching result answers a specific question about the records being compared. A review of the agreement answers a different question: were the right terms carried into those records?
Download the contract-to-PO review template (Excel) to test that question on a sample of orders.
Define what your matching control checks
Start with your actual configuration. In Dynamics 365 Finance, Microsoft describes three-way matching as comparing invoice prices with PO prices, within configured tolerances, and invoice quantities with corresponding product receipts. Other controls and configurations can add checks. Read the official matching overview before interpreting a pass or fail in that product.
Ask your systems owner what sources feed the PO price, which discounts and charges are compared, and where overrides are recorded. Your ERP may already manage purchase agreements, effective dates, and automated price selection. The practical question is whether the relevant agreement and amendments reached those controls correctly.
Write down what a pass proves in your environment. Do not label it “contract compliant” unless the applicable agreement was actually part of the validated chain.
Look for breaks between an agreement and the order
Useful review candidates have a reason to be checked. For example:
- A price amendment became effective while open orders still used an older schedule.
- A material number changed, and the replacement record did not carry the agreed terms.
- A supplier acknowledged a different price and the buyer's decision was never recorded.
- A quantity break depends on cumulative purchases, while the order uses a single-order price.
- A surcharge was retired or changed, but remained in a purchasing record.
- A blanket agreement was applied to a buying entity or site outside its scope.
These are investigation hypotheses, not automatic claims. An agreement might explicitly grandfather open orders. A tier might apply only to future purchases after a threshold. A quoted discount might depend on a condition the order did not meet.
Establish which terms apply
For each sampled PO line, answer five questions:
- Is this the covered material, specification, and supplier entity?
- Which date determines eligibility: order, shipment, delivery, or another agreed event?
- Does the ordered or accumulated quantity meet the specified condition?
- Is the price on the same currency, unit, discount, and charge basis?
- Is there an approved amendment or exception for this order?
Attach the evidence behind every answer. If the agreement wording is unclear, route that uncertainty to the authorised commercial owner. An extraction from a PDF is not a decision about conflicting terms.
Preserve the issued PO revision. Quietly overwriting the original price before documenting the issue destroys the explanation for why the invoice passed and makes later supplier discussions harder.
A worked example of a valid match and a questionable price
Illustrative example: a manufacturer orders 2,000 housings. The issued PO shows $4.80 per housing; the supplier invoices $4.80; receiving records 2,000 housings. Those records agree.
An approved amendment lists $4.65 for orders placed from September 1. This PO was issued September 3, for the covered material and buying entity. There is no recorded exception. The difference to investigate is $300: 2,000 × $0.15.
Before making a claim, the buyer checks whether the amendment was accepted, whether its date rule covers the order, and whether another commercial change applies. If those checks confirm the price, the supplier receives a request referencing the amendment and affected invoice line.
Now change one fact: the amendment applies to new orders from September 10. The same September 3 PO may be correctly priced. A comparison that ignores the effective-date rule would produce a false claim with convincing arithmetic.
Review a focused sample before expanding
Begin with a bounded population such as orders released around a known price change. Record the selection rule, date range, included supplier accounts, and excluded records. Include both matching passes and failures where relevant, so the review is not limited to the exceptions the existing control already found.
For each result, use a plain disposition: supported difference, valid price, missing evidence, or commercial review required. Count unique affected invoice lines, and check previous corrections before adding amounts. Never extrapolate a few findings to annual savings without disclosing and validating the assumptions.
The historical review guide explains how to use past documents to test a workflow. A focused review can also show where master data or approval handoffs need repair.
Separate a matching failure from a source-data failure
A matching failure says that the compared records disagree under the configured rule. A source-data failure can exist even when those records agree. Keep those categories separate during review because their remedies differ.
For a matching failure, the investigation might reconcile an invoice to a receipt or an approved PO amendment. For a source-data failure, the reviewer must trace why a governing commercial change did not reach the ordering record. Reprocessing the same three documents will not answer a question about a fourth document that was never included.
Do not assume every agreement belongs directly in an automated matching rule. Some conditions require a commercial judgment, such as whether an order-specific exception was accepted. Establish which conditions can be represented reliably in the current system and which need an evidence-backed review. Existing ERP capabilities may already cover much of the requirement; the task is to test how your implementation behaves.
Use a small test set with known outcomes
Prepare several controlled examples before reviewing a large transaction population. Include a valid old price, a valid new price, an unclear amendment, a unit conversion that should produce no difference, and a previously corrected invoice. Have the commercial owner record the expected disposition for each example before comparing results.
A useful test asks whether the workflow finds the right answer and explains it. If two prices normalize to the same value, there should be no price claim. If evidence is missing, the outcome should say what is missing. If a correction already exists, it should be linked rather than counted again.
Use copies or an isolated review process so the exercise does not change issued orders, invoices, or supplier communications. Any later production update should have its own approved instruction. This is a process-validation exercise, not permission to rewrite historical purchasing records.
Reproduce a boundary-date review
Consider an illustrative agreement that changes the price from $10 to $9.60 for orders placed on or after October 1. It explicitly preserves the old price for earlier orders. Four invoice lines each cover 250 units, and each invoice matches its PO.
- Order A was placed September 30 at $10. The price is valid under the stated rule, even if the goods arrive in October.
- Order B was placed October 2 at $10. Its potential difference is 250 × $0.40 = $100.
- Order C was placed October 3 at $9.60. No price difference is present.
- Order D was placed October 4 at $10, but the file includes a possible order-specific exception that has not been verified. Its $100 difference remains under commercial review.
The review has four lines, one supported $100 difference, two valid prices, and one unresolved $100 question. It does not establish $200 of recoverable money. If B already received a verified $40 correction, its outstanding amount is $60. Keep the original difference and prior correction visible.
Now change the agreement's trigger from order date to receipt date. The classifications may change. That is why the trigger must be written into the review record rather than inferred from a file name such as “October prices.” The supplier price-list change guide explains how to preserve that rule when updating purchasing data.
Report findings without overstating the sample
A targeted sample around a price amendment is deliberately enriched for possible problems. Its findings cannot be treated as a representative percentage of annual spend. Report the reason for selection, included suppliers and materials, dates, number of lines, unresolved evidence, and amounts by currency.
Keep separate counts for supported differences, valid prices, and unresolved reviews. If a later commercial decision changes a finding, update the result while preserving the original classification and reason. A smaller corrected exposure figure is a better result than an inflated number nobody can reproduce.
Track the time required to establish the applicable terms and the recurring point of failure: amendment intake, material mapping, effective-date setup, approval, or manual override. This locates repair work. It does not justify attributing every price difference to the ERP, the supplier, or a buyer without following the evidence.
Correct the record and follow the supplier outcome
A useful finding produces two actions. First, resolve the affected invoice using the price discrepancy process. Second, assign the underlying price-record or amendment handoff correction to its authorised owner.
Keep those actions linked. Closing the invoice issue without fixing the source can leave the next order exposed. Updating a price record without resolving the earlier invoice leaves the original difference outstanding.
Mandel's invoice discrepancy work connects the applicable terms, supplier follow-up, and updated records. The useful outcome is a supported resolution with a traceable decision, not a longer list of mismatches.
