CurcleVisit Website →
← Back to blog

Operations first purchase order matching: fix PO acknowledgements

September 21, 2026
Operations first purchase order matching: fix PO acknowledgements

Purchase order matching compares an invoice against its supporting documents and, when configured correctly, lets your accounts payable team auto-approve clean invoices while routing only genuine exceptions for review. It relies on comparing what was ordered, what arrived and what was billed. Done well, it delivers two things AP leaders care about most: fewer payment errors and a hard line against duplicate or fraudulent invoices.

TL;DR:

  • Automated matching engine handles high invoice volumes quickly by extracting data via OCR, vendor portals, or EDI, and compares it to purchase orders and receipts.
  • Proper configuration of tolerances—using a mix of fixed and percentage limits—is essential for balancing automation efficiency and error detection.
  • Exceptions should be triaged by root cause and value, with clear ownership and SLAs, to prevent backlog and ensure timely resolution.
  • Using a connected system with integrated job, stock, and invoice data reduces false positives and streamlines discrepancy investigation.
  • Starting with clean master data and piloting on messy suppliers helps optimize rules and achieve reliable first-pass auto-approval rates.
Curcle
Bring Operations Into One System
Curcle connects jobs, stock, customers, compliance and invoicing, helping service teams work from the same operational information.
Explore Curcle

Table of Contents

What is purchase order matching in the procure-to-pay cycle?

Purchase order matching sits near the end of procure-to-pay, the moment a supplier's invoice arrives and someone has to decide whether it gets paid, queried, or blocked. It compares four possible documents: the purchase order (what was agreed), the goods receipt note (what physically arrived), any inspection or quality sign-off (for regulated or high-value items), and the invoice itself (what the supplier is billing for).

Each document has an owner. Procurement raises and owns the PO. The warehouse or receiving team logs the goods receipt. Quality or compliance staff handle inspection sign-off where it applies. AP owns the invoice and, ultimately, the matching decision. When these teams work from separate spreadsheets or disconnected systems, matching becomes a manual chase for missing paperwork rather than a control.

The mismatches AP teams see most often follow a predictable pattern:

  • Price variance: the invoice bills a different unit price than the PO states, often because a supplier increased rates without formal notice.
  • Quantity variance: the invoice charges for more units than the goods receipt confirms were delivered.
  • Missing PO reference: the invoice arrives with no PO number at all, usually from a maverick purchase outside the approved process.
  • Duplicate invoice: the same invoice number, or a near-identical one, has already been processed against the same PO.
  • Tax or freight discrepancy: additional charges appear on the invoice that were never itemised on the original order.

Each of these is a signal, not necessarily a problem. A well-run matching process treats them as data points to investigate quickly, not automatic grounds for rejection.

What are the three types of purchase order matching?

Most AP teams choose between two-way, three-way and four-way matching depending on what they are buying and how much risk a bad invoice represents. None of these is universally "better". The right choice depends on spend category.

  1. Two-way matching compares the invoice directly to the purchase order, checking price and quantity ordered against price and quantity billed. It works well for services, subscriptions and low-risk indirect spend, where there's no physical goods receipt to check against. The trade-off is speed for scrutiny: a two-way match can't catch the invoice for goods that never actually arrived.

  2. Three-way matching adds the goods receipt note, comparing PO, receipt and invoice together. This is the standard approach for physical goods, because it confirms the invoice reflects what actually turned up on the loading dock, not just what was ordered. Corpay's guide to three-way matching describes it as a strong defence against duplicate payments and billing fraud, and for most goods-based businesses it's the sensible default rather than the exception.

  3. Four-way matching adds an inspection or quality acceptance step on top of the three-way check. It's reserved for high-value, regulated or safety-critical purchases, where receiving the correct quantity isn't enough. A batch of components might arrive complete and on time but fail a quality check, and only a four-way process catches that before payment goes out.

The trade-offs are real. Two-way matching is fast but blind to delivery failures. Three-way matching catches most fraud and delivery discrepancies but adds a dependency on timely, accurate goods receipting, which many warehouses still do late or inconsistently. Four-way matching gives the tightest control but adds a manual inspection step that slows down payment, so it only makes sense where the risk of a bad unit reaching production or a client site outweighs the delay.

A pragmatic rule many AP teams use: apply two-way matching below a set spend threshold, three-way matching for anything involving physical goods above it, and reserve four-way matching for a short list of regulated categories, such as safety equipment, pharmaceuticals or engineering components, where a quality failure carries real consequences.

How does automated purchase order matching actually work?

Automation doesn't remove the matching logic above, it just executes it at speed, across thousands of invoices, without a human manually pulling three documents and comparing them line by line.

Invoices reach the matching engine through a few common capture routes:

  • Optical character recognition (OCR) reads scanned or emailed PDF invoices and extracts header and line data automatically.
  • Vendor portals let suppliers submit invoices directly into a structured format, skipping OCR entirely.
  • EDI (electronic data interchange) connects large suppliers' systems directly to the buyer's ERP for high-volume, recurring purchases.

Once captured, the invoice data needs a PO and, for three-way and four-way matches, a goods receipt record pulled from the ERP or procurement system. This is where header versus line-level matching becomes a genuine design decision rather than a technical footnote. Header-level matching checks totals: does the invoice total fall within tolerance of the PO total? It's faster to configure and more forgiving of messy line data, which makes it a sensible fallback when invoice line detail is poor, as Signup Software's implementation guide notes. Line-level matching checks every individual item, quantity and price, which catches far more granular discrepancies but demands accurate, well-structured line data from both the PO and the invoice.

Partial shipments and instalment invoicing need a different logic entirely: life-to-date matching. Instead of comparing a single invoice to a single PO, the system tracks cumulative invoiced amounts against the PO's total value over its life. This stops the system flagging a false exception every time a supplier bills incrementally against a long-running contract or a multi-delivery order, a scenario that trips up plenty of poorly configured systems.

When an invoice, or a line within it, falls inside the configured tolerances on price, quantity and any tax or freight variance, the system triggers straight-through processing: it posts the invoice for payment with no human touch. Anything outside tolerance drops into an exception queue instead. That queue, not the matching engine itself, is usually where AP teams spend most of their time.

What matching rules and tolerances should you configure?

Tolerance settings are where most PO matching implementations succeed or fail, and it's worth treating rule configuration as an ongoing discipline rather than a one-off setup task.

Rules can be applied at header, line or vendor level, and separately as individual rules (each invoice checked in isolation) or life-to-date rules (cumulative checks against the PO's running total). SAP's configuration guidance for Concur groups these into rule sets that determine whether an invoice auto-approves or gets blocked for review, and whether a variance type triggers a hard stop or a soft warning.

Tolerance types generally fall into three categories:

  • Absolute tolerances: a fixed amount, such as £10 or £25 variance allowed regardless of invoice size.
  • Percentage tolerances: a proportional allowance, such as 2% or 5% of line or invoice value, which scales sensibly across both small and large purchases.
  • Asymmetric tolerances: different thresholds for overbilling versus underbilling, since a supplier charging less than agreed rarely needs the same scrutiny as one charging more.

Many teams combine both: a percentage tolerance with an absolute cap, so a 5% variance on a £200,000 order doesn't wave through a £10,000 discrepancy unchecked.

Getting this balance wrong in either direction causes real damage. Set tolerances too loose and genuine billing errors slip through. Set them too tight and you flood the exception queue with trivial rounding differences and freight surcharges, which is a documented failure mode: overly tight thresholds routinely turn automation into extra manual work rather than less. Currency conversion rounding, tax calculation differences and freight charges added after the PO was raised are the three most common sources of nuisance exceptions, and each deserves its own tolerance line rather than being lumped into a single blanket rule.

The right message lets whoever picks up the exception resolve it in seconds instead of reopening three documents to work out what went wrong.*

How should exceptions be triaged and approved?

An exception queue without clear ownership rules becomes a dumping ground nobody feels responsible for clearing. The fix is triaging by variance type and value from the moment an exception is raised, not after it has sat unresolved for a week.

  1. Sort by variance type first. Price variances, quantity variances and missing-PO invoices need different people and different questions, so route them into separate queues rather than one undifferentiated pile.
  2. Assign ownership by root cause. Price discrepancies usually belong with procurement, since they likely negotiated the original rate. Quantity discrepancies belong with the receiving team, who can confirm what actually arrived. Missing-PO invoices go straight to the original requester to establish whether the purchase was ever authorised.
  3. Set value-based SLA targets. A £50 variance on a routine supply order can wait two or three days for resolution. A £5,000 variance on a capital purchase needs same-day attention and probably a named escalation contact.
  4. Escalate on age, not just value. Any exception open beyond its SLA, regardless of size, should escalate automatically to a supervisor rather than quietly ageing in the queue.
  5. Capture resolution metadata every time. Log why each exception occurred and how it was resolved. That history is what turns next quarter's tolerance review from guesswork into an evidence-based adjustment.

The goal isn't zero exceptions. A healthy exception rate means the system is catching real discrepancies rather than letting them pass. The goal is that every exception has an obvious owner and a realistic clock running against it, which is usually the difference between a two-day resolution and a three-week one.

How do you roll out automated purchase order matching?

Most successful rollouts move through four stages rather than attempting a full switch in one go.

Prepare. Clean up master data before anything else: standardise vendor names, SKU codes and units of measure, since inconsistent data is the single biggest cause of false-positive exceptions once automation goes live. Decide your PO acknowledgement policy now, and map out exactly who owns each document in the process, from PO issue through to goods receipt.

Configure. Choose header or line-level matching per category, based on how reliable your line-item data actually is rather than which option sounds more rigorous. Build your rule groups and tolerances, starting conservative. Connect your capture method, whether OCR, vendor portal or EDI, and confirm it talks cleanly to your ERP or job management system.

  • Start with one or two supplier categories, not your entire vendor list.
  • Measure first-pass match rate and exception volume weekly during the pilot.
  • Keep a manual fallback for anything the pilot flags as ambiguous.

Pilot. Run the configured rules against a representative slice of suppliers and spend categories before scaling. Evidence on invoice reconciliation points to piloting tolerances on a representative supplier subset and tracking first-pass match rate against exception cost to find the workable middle ground, rather than guessing at thresholds from a spreadsheet exercise.

Scale. Once the pilot's numbers hold up, widen supplier coverage, add life-to-date rules for contracts with partial deliveries, and build measurement into a recurring review rather than a one-off audit.

Pro Tip: Run your pilot on a supplier group with genuinely messy invoicing habits, not your cleanest vendor. If the rules survive contact with your worst supplier, they'll hold up everywhere else.

What best practices keep purchase order matching effective?

The technology behind matching gets most of the attention, but the controls sitting upstream of it decide whether that technology actually works.

Treat the PO acknowledgement as a genuine control point, not paperwork. When a supplier confirms an order, validate the acknowledged price, quantity, currency and delivery date against the original PO before goods ever ship. AuraVMS's acknowledgement checklist treats the acknowledgement effectively as a contract amendment, catching price or scope drift weeks before it would otherwise surface as an invoice discrepancy.

Purchase order acknowledgement validation flow

Master data hygiene matters more than most rule configuration decisions. Inconsistent vendor records, duplicate SKUs and mismatched units of measure generate a steady stream of exceptions that have nothing to do with genuine billing errors, and poor upstream data quality is consistently cited as the main cause of false-positive exceptions once automated matching is live.

Track a small set of KPIs rather than a dashboard nobody checks:

  • First-pass match rate: the percentage of invoices that clear automatically with no human intervention.
  • Exception rate: the proportion of invoices requiring manual review.
  • Time to resolution: average days from exception raised to cleared.
  • Cost per exception: an estimate of staff time spent resolving the average flagged invoice.

Automating everything isn't the goal. High-value purchases, new suppliers with no invoicing history, and genuinely ambiguous variances still warrant a human look, and forcing them through straight-through processing just to hit an automation percentage tends to create the exact fraud exposure matching exists to prevent.

How Curcle approaches purchase order matching from field experience

Curcle grew out of a real UK service and engineering business, not a whiteboard exercise in software design, and that history shapes how it treats PO matching. When jobs, stock, purchase orders and invoicing live in separate systems, exceptions multiply for reasons that have nothing to do with genuine billing errors, they're simply data that never reconciled in the first place.

Some platforms keep jobs, stock, purchase orders and invoicing in a single connected system, so a PO raised against a job links directly to the goods or parts recorded against that job and the invoice that eventually follows. That single source of truth reduces the false-positive exceptions that come from chasing paperwork across disconnected tools, and it keeps an audit trail ready for export when compliance or finance need to review a decision.

[Author bio, client case studies and measured customer outcomes to be added.]

What actually makes purchase order matching work

Most AP teams overinvest in tolerance precision and underinvest in the boring upstream discipline: PO acknowledgement checks, consistent supplier master data, and someone actually reading the exception queue every day rather than letting it become a monthly clean-up job.

If you're implementing or improving matching this quarter, start with three things: clean your PO and vendor master data before touching any tolerance settings, pilot automation on a genuinely messy supplier group rather than your best-behaved one, and track first-pass match rate from week one so you have evidence, not opinion, when you adjust the rules.

— Luke Herridge

See how Curcle keeps PO matching under control

Curcle gives operations and finance teams a single connected view of jobs, stock, purchase orders and invoicing, so an exception doesn't mean chasing three systems for the missing piece. Because job costs, stock movements and PO data sit in one place rather than scattered across a job management tool, a spreadsheet and separate accounting software, discrepancies get caught with the context already attached, not reconstructed after the fact.

Curcle

If invoice queries are eating hours every week in your AP team, that's usually a sign the upstream data, not the matching logic, needs attention. Curcle's features page covers how job, stock and invoice data connect in practice, and the pricing page lists current plans, starting with Starter from £99 per month, through to Enterprise on request. Book a look at the product tour to see how it fits your current PO process before you commit to anything.

Sources

For teams building or refining matching rules, SAP's Concur configuration guide is worth working through directly, particularly for life-to-date rule setup. Corpay's guide to three-way matching explains the fraud-prevention case for goods-based businesses in more depth than most vendor pages attempt. Amazon Business's guide to invoice matching covers the shift from reactive reconciliation to proactive control and why upstream data quality determines most outcomes. For acknowledgement controls specifically, AuraVMS's checklist lists the fields worth validating before goods ever ship.

FAQ

Should a purchase order always match an invoice?

Not always, but any variance should be understood before payment, not ignored. Invoices within your configured tolerance can auto-approve safely; anything outside tolerance needs a documented reason before it clears.

What happens if a PO does not match an invoice?

The invoice drops into an exception queue rather than being paid automatically, and gets routed to whoever owns that variance type, typically procurement for price issues or receiving for quantity issues. It stays unpaid until the discrepancy is resolved or formally approved as an exception.

How do you match purchase orders with invoices?

Compare the invoice against the PO (two-way), or the PO plus goods receipt (three-way) for physical goods, checking price, quantity and any tax or freight charges against agreed tolerances. Systems described in SAP's matching configuration guide automate this comparison and flag anything outside tolerance for review.

What if the PO quantity doesn't match the vendor invoice?

A quantity mismatch usually means the invoice is billing for more, or fewer, units than the goods receipt confirmed arrived. It should route to the team that logged the receipt to confirm the actual delivered quantity before the invoice is approved or queried with the supplier.

What is the difference between two-way and three-way matching?

Two-way matching checks invoice against PO only, which suits services and low-risk purchases. Three-way matching adds the goods receipt note and is the standard for physical goods, since it confirms what was billed reflects what actually arrived, as Corpay's guide explains.