Back to blog

Never Fear the Hard Stop

Most duplicate payments don’t start in a spreadsheet or a bank file. They start at the keyboard, the moment an invoice is entered. And whether your ERP is configured at that exact moment for a hard stop or a soft stop has more to do with your duplicate-payment rate than almost anything else. My advice, after 25 years of looking at this: never fear the hard stop.

Hard stop vs. soft stop

Every major ERP can be configured to react when someone tries to enter an invoice number that already exists for a vendor. There are two ways to set it up.

A soft stop throws a warning: “This invoice already exists. Do you want to continue?” The processor can click through and keep going.

A hard stop doesn’t ask. It simply won’t let the same invoice number be entered twice for the same vendor. Full stop.

That difference sounds small. In practice, it’s enormous.

Why the soft stop fails

The soft stop assumes the person at the keyboard will pause, investigate, and make the right call. But look at how AP data entry actually works. Processors are measured on volume: invoices entered per day, per week, per month. When a pop-up appears, the fastest way to hit the number is to click “continue” and move on. The warning becomes background noise.

Now put that in an outsourced environment, where the processor has no connection to whose money it is. If an invoice gets entered twice, it’s no skin off their back. The incentive to stop and dig in is essentially zero. The soft stop turns a control into a reflex, and the reflex is to keep going.

“But sometimes it’s a legitimate re-entry”

Here’s the objection I always hear: sometimes people need to re-enter, so they’ll just append an “A” or a “-1” to the invoice number to get past the block. True. But look closely at what that actually is.

Appending a character is a conscious decision. It’s harder to do, and more importantly, it’s visible. Instead of an invisible click that disappears into the void, you now have a deliberate act sitting in your data: one you can find and analyze after the fact.

The hard stop turns a reflex into a decision

That’s the whole point. A soft stop lets a duplicate slip through with a single, forgettable click. A hard stop forces the person to do something, and whatever they do, it leaves a trail.

If they stop, great: the duplicate was prevented. If they append a character to force it through, that’s now a data point. You can run analytics on which processors are appending invoice numbers and how often. Sometimes you’ll find a legitimate pattern. Sometimes you’ll find real duplicates. And sometimes you’ll find a training issue: someone who doesn’t understand the process and is quietly creating exposure every day. Either way, you can see it, and you can fix it.

A soft stop gives you none of that. The click leaves no meaningful signal.

Yes, it adds friction. Build for it.

The reason organizations default to soft stops is fear: fear that a hard stop will block legitimate work and slow the team down. And it’s true that there are real edge cases: a vendor that genuinely reuses invoice numbers, credit memos, rebills.

But the answer isn’t to weaken the control. It’s to pair the hard stop with a defined review path: when a true duplicate is flagged, it goes to someone who checks it, rather than leaving the decision to a processor racing against a volume target. Design the exception process on purpose, and the friction stops being a problem and starts being a control.

Don’t fear the hard stop

A soft stop feels flexible. What it really is, is a duplicate payment waiting for a click. The hard stop feels rigid, but it prevents the error, and when someone works around it, it hands you the evidence to coach, correct, and improve.

After watching this play out at organization after organization, I’ll say it plainly: I believe in the hard stop. Don’t fear it.

Not sure how your ERP is configured, or what’s slipping through? Start a no-cost Proof of Value and we’ll show you what’s really in your data. We only get paid a percentage of what we recover.

Karl Andersson
CEO, AP Impact

Karl has spent 25+ years in AP auditing and analytics, helping finance teams recover lost value and understand their payables. He writes about what he’s actually seen in the field. Read his story →

Keep reading

We Asked AI to Find Our Duplicate Payments. Here's What Actually Happened.
Point AI at your ERP and ask it to find duplicate payments. Here’s what really happens, and how we do it differently.
99.9% Accurate, and Still Losing Millions?
You’ll hear that AP processing is 99.9% accurate. Even if it’s true, do the math, and here’s what really drives your error rate.