Ask an owner what it costs to retype an order into a second system and you’ll usually get a shrug. “A minute, maybe two.” That answer is why the problem sticks around for years.
Here’s the arithmetic, done honestly.
Start with the real minutes
Retyping an order is rarely one action. It’s: open the second system, find or create the customer, key in the line items, check the units, fix the one line that doesn’t match, save, then go back and mark the first system as done. Time it with a stopwatch instead of from memory. In every warehouse and back office we’ve measured, the honest number lands somewhere between four and nine minutes. Not one or two.
Take the low end. Four minutes, sixty orders a day, five days a week. That’s twenty hours a week. One full-time person, doing nothing but reading a screen and typing what it says into another screen.
At $25 an hour all in, that’s roughly $26,000 a year. It’s a real number, it appears in no budget line, and nobody owns it.
Then add the mistakes
People copying numbers get them wrong somewhere between half a percent and one percent of the time. Call it 0.5%. Sixty orders a day is about fifteen thousand a year, so seventy-five orders leave your building with something wrong on them.
The cost of a wrong order isn’t the typing. It’s the phone call, the return label, the credit note, the second shipment, the apology, and now and then the customer who doesn’t come back. Most people, pressed, put the all-in cost of one wrong shipment north of $80. Seventy-five of those is another $6,000 — and that assumes you catch them.
A typing mistake is also the worst kind, because you can’t see it happen. Nothing breaks. The screen looks right. It shows up two weeks later as a stock count that won’t add up.
Now the part nobody prices
The retyping job is almost always done by one person who’s unusually good at it. They know the units are cases in one system and singles in the other. They know customer 4412 is really the same as customer 4412-B. They know which three products always need a manual override.
None of that is written down anywhere. It lives in one head.
When that person is on vacation, output halves and mistakes triple. When they leave, you find out you never had a documented process — you had a person, and you called it a process. That’s the risk that actually keeps owners up at night, and it doesn’t show up in any hourly calculation.
And the part that’s hardest to see
Retyping sets the pace for everything downstream. If orders are keyed in once a day at 4pm, your warehouse can’t pick before 4pm, your stock levels are a day out of date, and your website is promising things you don’t have. The delay isn’t the typing. It’s the batching the typing forces on you.
We watched a company move from a once-a-day manual push to a five-minute automatic sync and find that the real win wasn’t the twenty hours. It was that the website stopped overselling, because reserved stock finally moved in near real time instead of overnight.
What to do about it
You don’t need to replace your systems. You need the two you have to hand each other the specific records a person is currently carrying between them.
Three questions decide whether it’s worth doing:
- Can both systems be reached? If system A has an API or a database you can read, and system B has an API or a database you can write to, the job is doable. Most business software from the last twenty years qualifies, including a surprising amount of the old stuff.
- Is there one shared identifier? A product code, a barcode, an order number, a customer ID. If the two systems agree on one, the connection is straightforward. If they don’t, the first piece of work is agreeing on one — and that’s worth doing on its own.
- What happens when it fails? The answer must not be “quietly.” A good connection writes down every record it moved, retries what it can, and puts whatever it can’t in a queue somebody can actually see.
The honest test
Pick the record you retype most. Count how many times a day someone reads it in one system and types it into another. Multiply by four minutes. Multiply by 250 days.
If the answer is more than a couple of weeks of somebody’s year, this isn’t an annoyance. It’s an unbudgeted salary, an invisible error rate, and a single point of failure with a friendly face.
That’s usually the first thing worth fixing, and it usually pays for itself well before the year is out.