Invoice capture software: how to choose

Deciding to automate invoice entry is the easy part. What comes next is harder: several products, all of which read a PDF, all of which show a tidy table, and all of which promise saved hours. In a demo they look almost identical. The differences show up in month two — on the invoice nobody expected, and on the question of who is answerable for it.

So choose on answers to a few specific questions rather than on a feature list. Here are the questions, and why each one decides something.

Start with your month, not with a feature list

Before you look at any product, write down your month: how many invoices arrive, how many land in the last two days, how many suppliers send summary invoices with dozens of lines, how many documents come in as photos from employees, and how many times a month you fix the same thing. That list is the only basis on which two different products can be compared. A feature that touches none of your month's lines costs exactly as much as the ones that do.

Does it speak to your accounting system?

This question eliminates the most options, so ask it first. Plenty of tools can read an invoice; the harder part is that the entry lands in your accounting system the way your accountant would have entered it by hand — with the right ledger account, VAT code and dimensions. Don't ask whether they integrate; ask whether it already works with Rivile GAMA, Finvalda, Centas or Odoo, and whether a real client is using it. The gap between planned and working is usually measured in months of your time, not theirs.

What happens to an invoice it gets wrong?

Every product handles a clean PDF. The real difference shows on a bad one: an amount written in by hand, a two-page invoice whose second page was scanned crooked, a supplier who sends the same document three times. The question is simple: does the software quietly record a guess, or does it say it is unsure and put the invoice in front of a person? Ask to see an invoice they failed on. A vendor who has one to hand looks at their own mistakes; a vendor who doesn't simply isn't looking.

Who approves before anything reaches the books

Automation does not mean documents travel into the ledger unseen. Judge three things: whether you can decide who approves what; whether you can see who changed a value and when; and whether an approval can be undone and the invoice sent back for review. Without the third, the first two are just a report — and errors are usually found after approval, not before.

Whose rules are they — yours or the vendor's?

Every company has ten small conventions nobody else knows: this supplier's invoices always belong to one site, cleaning services split across two departments, one contract carries its own cost centre. The question is not whether rules exist, but who sets them. If every change means emailing the vendor and waiting, each small convention becomes a week. If you set them yourself, the rules keep up with the business.

What a trial actually proves

A trial with five clean invoices proves nothing. Take a real month with all of its trouble in it: a credit note, a telecom invoice carrying several VAT treatments, a foreign supplier, a duplicate, and one summary invoice with a hundred lines. Then measure two numbers rather than an accuracy percentage: how many documents had to be touched by hand, and how long the touching took. Those are the only two numbers that will still look the same six months later.

Questions worth asking before you sign

• •Which accounting systems does this already work with for real clients, not on a roadmap?
• •What happens to an invoice it doesn't understand — does it guess, or ask?
• •Do I change the assignment rules myself, or does every change go through you?
• •Who sees which client's documents, and what does the approval flow look like?
• •Will I see who changed an amount or a VAT code, and when?
• •What does a month with twice the usual volume cost?
• •How long does rollout take, and what do I have to prepare?

Price: per document, per user, or per company

Pricing models differ more than they first appear, and they can only be compared using your own numbers. What matters is not the monthly figure in a normal month but how that figure behaves in your worst one: year-end, or the month a new client or department is added. Price all three scenarios — the third one usually reveals which offer is actually cheaper.

If you are choosing for twenty clients, not one company

An accounting firm has a different list. Do you reach every client from one place, or log in twenty times? Is onboarding a new client half a day or a project? Does a team member see only their own clients? Do rules set up for one client stay strictly with that client? If those answers come back vague, the product was built for a single company, and the firm will always be a guest in it.

Where to start

Pick three products, prepare one real month of documents, and run it through all three. The questions in this article get answered in a week rather than a quarter — and after that week the decision is usually obvious.