Typing an order from the inbox into the ERP system. Copying customer details from the CRM into the invoicing package. An export to Excel that gets reimported every Friday afternoon. Almost every organisation has routes like these, and anyone looking for a hard error rate mostly finds figures that vary wildly by source. The honest answer: the exact number matters less than the mechanism behind it.
That mechanism is straightforward. Every manual step is a moment where something can go wrong, and that risk grows with volume, time pressure and repetition. The question isn’t whether mistakes happen, but where they end up and what they cost you once they’re there.
There’s no such thing as a reliable error rate
How much goes wrong when copying data by hand depends on the complexity of the data, the experience of the person entering it, and the circumstances. Towards the end of a long day, under time pressure, on the hundredth line of a list, the risk of error rises noticeably. More importantly: not every mistake is equal. A typo in a notes field goes unnoticed. A misplaced decimal point in a price or a transposed digit in an account number doesn’t, usually surfacing only after the damage is done. What matters for your organisation, then, isn’t an average from some study, but the combination of how often it goes wrong and where.
The mistakes that keep recurring
The nature of the mistakes is much the same at almost every organisation:
- Transposed digits or letters, such as 1234 becoming 1243
- Format mismatches, for instance dates or amounts that each system records differently
- Skipped fields, especially under time pressure or unclear instructions
- Duplicate entries, leaving the same customer or order in the database twice
- Data linked to the wrong customer, order or department
The nasty part is that these mistakes look perfectly valid at first glance. An amount with a misplaced decimal point sails through any cursory check, and only surfaces in an invoice that doesn’t add up or a report nobody trusts any more.
One input error rarely stays isolated
Systems pick up data from one another. A wrong address in the order system becomes a failed delivery. An incorrect amount flows through to the accounts and from there to the management dashboard, where it steers decisions that look well founded on paper. That’s how a single input error quietly multiplies through the chain; you can read what that means for your decision-making in our article on making decisions on inconsistent data.
Then there’s the clean-up work on top. Tracing where a mistake came from often takes longer than the original entry did, because you have to retrace the entire route. Want to know how big this issue is for you? Don’t count error rates, count the hours your team loses each week to retyping, checking and correcting. That figure tells you more than any average.
Removing the manual step works better than extra checks
Double checks, tighter input formats and the four-eyes principle all help, but they treat the symptom. As long as someone is typing data from screen A into screen B, the risk stays in place; you only make it smaller, and the process more expensive. The structural fix is removing the manual step itself by connecting systems directly. With API integrations, your software exchanges data automatically, in a single format, with validation at the source.
That doesn’t have to be a large project. Often, one well-chosen integration between your two busiest systems already removes most of the retyping and the mistakes that come with it. We’ve been building such data integrations since 2001 for organisations that need to trust their figures. Start with the route where the clean-up work is heaviest; that’s where an integration pays for itself fastest.