3 way matching is the strongest routine control in accounts payable and it is entirely dependent on two documents produced by other people. When both exist the check is cheap, fast and reliable. When either is missing the control does not degrade gracefully, it simply cannot run, and the invoice falls into a manual path that costs many times more.
What it protects against
Paying for goods that never arrived, paying a price nobody agreed, and paying twice for one delivery. Those three risks are the reason the control exists, and no other routine check covers them. It is worth stating plainly, because the case for upstream discipline is usually made in terms of efficiency when it is really about this.
Precondition one: the purchase order
Somebody raised an order before buying. Where that does not happen, there is nothing to compare a price or quantity against, and the invoice must be authorised on its own merits by whoever owns the spend. The share of your invoices in this category is a purchasing measure, and it caps everything matching can do.
Precondition two: the goods receipt
Somebody recorded what arrived, when, against the order line. This is the precondition most often missing, because it belongs to operational staff who are measured on other things. Making it take seconds on a phone at the delivery point is the highest-return change available in most organisations, and it costs nothing.
Questions people ask about 3 way matching
Is two-way matching a reasonable fallback?
For services and anything without a physical delivery it is the correct choice. For goods it removes the protection against paying for undelivered items.
What if the receipt arrives after the invoice?
The invoice waits, which is why receipt timeliness matters as much as receipt existence. Batching receipts weekly adds a week to every affected invoice.
Can tolerance be set to zero?
It can and it produces an exception for every rounding difference. A percentage with an absolute cap, reviewed against real exceptions, is the workable shape.