What is a Purchase Order? The Control Finance Teams Lose First as They Scale

A purchase order is a formal document a business sends to a vendor to authorize a purchase before it happens. It spells out what's being bought, how much, at what price, and when it should arrive or be delivered. Once a vendor accepts it, the purchase order becomes a binding commitment on both sides, the buyer agrees to pay under the stated terms, and the seller agrees to deliver what was ordered. That's the textbook definition, and if your company is past the ten-person stage, you already know it. What's less obvious is how quietly a working PO process falls apart as a company grows, and how much money slips through that gap before anyone notices. 

What's actually on a purchase order 

A PO usually isn't a single line. It's a small package of information built to remove ambiguity from a purchase before money moves. At minimum, that includes: 

• A PO number, so every downstream document (receipt, invoice, payment) can reference the same transaction 

• The vendor's name and the buyer's shipping or billing details 

• A description of what's being bought, including item, quantity, and unit price 

• The total committed amount 

• Delivery or service dates 

• Payment terms 

That last part is the piece finance teams care about most. Once a vendor accepts a PO, it typically functions as a contract, which is exactly why matching an invoice back to it matters. If the PO says 50 units at $12 each and the invoice shows up for 60 units at $14, that's not a rounding error. That's either a legitimate change nobody documented, or a problem. 

Why small companies don't feel this yet 

At a small company, POs are easy. One person approves the spend, one person orders it, one person pays the bill. There's no real gap between intent and execution because there's barely anyone standing between them. If a PO process exists at all, it's more of a formality than a control; everyone already knows what's being bought because everyone's in the same room, or at least the same Slack channel. 

Growth breaks that. More departments buying things. More vendors. More people with a card or an approval limit. More invoices arriving that nobody remembers authorizing. The PO process that worked fine at 20 employees starts producing exceptions, delays, and arguments at 200, not because anyone got careless, but because the process was never designed to hold up under that much volume. 

Why PO control slips first 

Finance and ops leaders tend to notice cash flow problems, hiring problems, and reporting problems well before they notice PO drift, mostly because PO drift doesn't show up as one big failure. It shows up as a dozen small ones, spread across departments, that don't look connected until someone adds them up. 

Maverick spending 

Someone orders materials or services directly from a vendor without cutting a PO first, because it was faster, and because nobody stopped them. Multiply that by every department and every month, and it adds up to real, untracked spend. Manufacturers feel this one hardest, where raw material and component orders often move faster than the paperwork tracking them, a production line waiting on a part doesn't care whether the PO caught up yet. 

Invoices with no matching PO 

An invoice lands in AP with nothing to check it against. Someone has to chase down who ordered it, whether it was approved, and whether the price and quantity are even right, all after the money's already committed. That chase eats hours every week, and it's rarely tracked as a real cost anywhere. 

No visibility into open commitments 

A PO gets issued but never shows up anywhere finance is actually looking. Nobody can answer "how much have we already committed to spend this quarter" with a straight face, which makes budgeting a guessing game dressed up as a spreadsheet. 

Approvals that exist on paper but not in practice 

The policy says every purchase over a certain amount needs sign-off. In reality, half of it happens over email or Slack, and the paper trail only gets built after the fact, if at all. 

Duplicate and split purchases 

Two departments order the same thing from two different vendors because nobody could see what the other had already committed to. Or one purchase gets split into smaller orders specifically to stay under an approval threshold, sometimes on purpose, more often just because nobody was tracking the running total. 

None of these show up on a dashboard right away. They show up months later, as a controller trying to explain a variance, or an AP team drowning in exception invoices, or a CFO who gets asked "did we actually approve this" and can't answer quickly. 

Here's the part that catches people off guard 

Older AP benchmarking from Ardent Partners and SAP found that best-in-class finance teams tied roughly 73% of invoices to a purchase order, while typical organizations managed closer to 55%. The same benchmark tied that gap to real operational cost: best-in-class teams processed a single invoice for around $2.94 and 3.6 days on average, compared to $15.96 and 16.6 days for everyone else, with an invoice exception rate roughly half as high. The exact numbers move around by report and by year, so treat these as a directional gap rather than gospel, but the pattern holds up in every AP benchmarking study worth reading: the difference between well-run finance teams and everyone else usually comes down to whether purchases get matched against what was actually authorized, and how much manual chasing that mismatch creates downstream. 

Two-way matching vs three-way matching 

This is the actual mechanism that closes the gap, and it's worth understanding both versions, because they catch different problems. 

Two-way matching compares the purchase order against the invoice or bill. It checks whether the quantity and rate being billed line up with what's left on the PO. If a PO has 5 units remaining and a bill shows up for 7, two-way matching flags the difference immediately, before anyone approves payment for units that were never authorized. 

Three-way matching adds a receiving receipt into the comparison. Instead of just checking the PO against the bill, it checks all three: what was ordered, what actually showed up, and what's being billed. A simple example makes the value obvious. Say a PO authorizes 10 units. The receiving dock logs 8 units actually received. The vendor's bill charges for 10. Two-way matching alone wouldn't catch that, the PO and the bill agree with each other. Three-way matching catches it immediately, because the receipt doesn't match either one. That's the exact kind of gap that either costs a company real money or turns into a vendor dispute nobody has time for. 

For any business buying physical goods, and manufacturers in particular, three-way matching is usually the more meaningful control. Services-heavy businesses can often get most of the benefit from two-way matching alone, since there's rarely a physical receiving step to log. 

What getting control back actually looks like 

The fix isn't more policy. Most companies that lose PO control already have a policy, it's just not enforced anywhere the software would actually catch a mismatch. The fix is making the PO, the receipt, and the invoice check each other automatically, so a mismatch gets flagged before payment goes out instead of after. 

That's the piece that's genuinely hard to do by hand once volume grows. A person can eyeball three documents for one purchase. Nobody can eyeball three documents for three hundred purchases a month without something slipping. 

This is exactly where Forwardly's Purchase Orders feature earns its keep. POs still get created in your accounting software, that part doesn't change, but once your accounting software is connected, open POs sync into Forwardly automatically and stay read-only there, so there's no risk of two systems disagreeing about what a PO actually says. From there, you can link one or more POs to a bill, and Forwardly checks the bill's quantity and rate against what's left on the PO, flagging anything that doesn't match before the bill goes out. Add a receiving receipt to the mix and you get full three-way matching, PO, receipt, and bill, checked against each other instead of trusted on faith. Those validation checks are informational rather than a hard stop, so they surface a discrepancy for a human to review rather than blocking a legitimate exception, which matters when the difference is expected and just needs a second set of eyes. 

Paired with AI bill capture reading incoming invoices automatically and an approval workflow that actually lives in the software instead of a Slack thread, it closes the gap between what was ordered and what actually gets paid for. Once a matched invoice is approved, instant payments get it out the door without adding a second delay on the back end of a process you just tightened up. 

Signs it's already time to fix this 

A few honest questions tend to surface the problem faster than any audit: 

• Can your team say, right now, how much you've already committed to spend this quarter, not just what's been paid? 

• When an invoice without a PO shows up, is there a clear process, or does it become a scavenger hunt? 

• Do the same two or three vendor disputes keep coming up because nobody can agree on what was actually ordered? 

• Are approvals happening in the software, or mostly in email threads that nobody can audit later? 

If more than one of those gave you pause, the gap is probably already costing you more than it would take to close it. 

If your PO process still lives in someone's inbox, a shared spreadsheet, or "we'll sort it out when the invoice shows up," that's worth fixing before it costs more than it already has. Learn how Forwardly's AP automation works or take the product tour to see it against your own invoice volume.