How to choose spend classification software

Published 25 October 2026 · 6 min read

Every spend classification vendor says its results are accurate, fast and easy. Demos are run on clean, friendly data. The way to tell tools apart is to ask questions that only have useful answers when the product really works. Here are ten, grouped by what they protect you from. They work for any tool — including ours.

Accuracy: will the output be usable?

Can I see results on my own data before I pay? Not a demo dataset. A sample of your file, categorised, that you can check line by line. If the answer is “after onboarding”, you are paying to find out.
When you quote accuracy, what exactly was measured? “95% classified” often means 95% of rows got a category, which is coverage, not accuracy. Ask how many were checked by a person, and at which level. Our guide to measuring classification accuracy gives you a one-hour method to check for yourself.
What happens to rows the tool is unsure about? A good answer surfaces them for review. A bad one quietly puts them in “Other”.

Structure: will it fit how we work?

Is the taxonomy aligned to a standard? A structure mapped to UNSPSC, NAICS or NACE will still make sense after a reorganisation, and can be compared across years and business units.
Can we change the categories? Rename, merge, split — without a change request.
Does it work in our languages? If you buy in more than one country, ask whether categories come back in the language of the data or are forced into English.

Ownership: can we leave?

What do we get back, and in what format? A plain file with every line and its categories is portable. A dashboard you can only view inside the vendor is not.
What happens to our data? Where is it processed, by which sub-processors, and is it used to train anything? See is it safe to upload spend data.

Cost: what will this really cost?

What is the total for year one, and for year two? Licence, implementation, training, and the cost of a re-run next year. The second year is where surprises live. See the true cost of spend analytics.
How long until we have a categorised file? Days, weeks or a quarter? A long implementation is a cost even if it is free.

How to use the answers

Score each vendor on the ten questions, but weight the first two heavily. A tool that cannot show you accurate results on your own data before you commit is asking for trust it has not earned yet.

How Structera answers them

You get a free sample on your own data before paying; categories aligned to UNSPSC, NAICS and NACE in the language of your data; an Excel file you keep; and a price per file published on our pricing page. We only categorise — if you need sourcing or contract management, you need a different kind of tool.