Case studies are the main evidence available in this category and they are written by vendors. That does not make them useless; it makes them documents to interrogate. Four questions convert a promotional artefact into something you can learn from, and the answers are often more interesting than the headline figure the study was built around.
What was the baseline, and was it measured
A saving is a difference from something. If the before figure was estimated rather than timed, the improvement is an estimate. Round numbers in a baseline are usually a sign it was not measured. Look for how it was established, over what period, and whether it included the exception handling or only the clean path.
What was in scope, and what else changed
Frequently the largest part of a reported improvement came from a process change that happened alongside the software: a purchase-order policy, a consolidated mailbox, a receipt discipline. That is genuinely useful to know, because those changes are free and you may be able to make them first and independently.
How comparable is the organisation
Invoice volume, the share with purchase orders behind them, the industry and how the team was structured before. A hospitality group with thousands of small invoices and a professional firm with a few hundred large ones will get different things from identical software. Comparability determines whether the number means anything for you.
Questions people ask about ap automation case study
Should we ask for a reference customer?
Yes, ideally one of similar size and invoice mix, and ask them what did not go well. That answer is usually the most valuable thing in the whole evaluation.
Can we build our own before figure?
In an afternoon: time ten invoices end to end, count monthly volume, and count how many had an order and a receipt behind them.
Are published industry averages useful?
As context, not as a business case. A finance director will ask where the number came from and whether it applies here, and the honest answer is usually that nobody knows.