If an order is typed into more than one system, somebody is paying for it: in minutes on every order, and now and then in a mistake that costs far more than the minutes. Most shops know this in general. Very few can say how much, and that makes two decisions hard: whether automating it is worth the money, and, afterwards, whether it actually helped.
The good news is that measuring it is cheap. It takes one ordinary week, a spreadsheet and a little discipline. This article walks through where the manual work hides, how to log it, what to count, how to decide what to automate first, and how to check the result.
Where the manual work hides
Follow one order from checkout to the customer’s door and write down every time a person touches it outside the shop itself. In most shops the list looks something like this:
- Order intake: reading the new-order e-mail, checking stock, checking the address looks real.
- Payment: matching a bank transfer to an order, marking cash on delivery as received, chasing an unpaid order.
- Invoicing: typing the customer, the products and the prices into an invoicing tool, then sending or filing the invoice.
- Shipping: typing the name, phone, address or office and the cash-on-delivery amount into the courier’s website, printing the label.
- Customer questions: "where is my parcel?", answered by looking up the tracking number and copying it into a reply.
- Corrections: a wrong invoice, a parcel sent to an old address, a cash-on-delivery amount that did not match.
- Month end: exporting orders and invoices for the accountant, and reconciling what the courier paid out.
Not all of it is waste. Checking a strange address is judgement. Typing an address that already exists in the order is copying. The measurement is mostly about separating the two.
The one-week log
For one ordinary week, not the Christmas rush and not the quietest week of the year, everyone who handles orders writes down every time they touch an order outside the shop. One row per touch:
date, order_no, task, tool, minutes, retyped_fields, correction
2026-09-14, 1043, invoice, invoicing app, 3, "name, address, VAT id", no
2026-09-14, 1043, label, courier website, 4, "name, phone, address, COD amount", no
2026-09-15, 1047, label, courier website, 6, "name, phone, address", "wrong office, relabelled"
2026-09-15, 1039, tracking question, e-mail, 3, "tracking number", noThe column that matters most is retyped fields: which facts were typed again that already existed somewhere else. That is the copying, and it is the part software can remove.
Three things make the log accurate:
- Make it easy. A shared spreadsheet open on the packing desk, or a two-question form on a phone, gets filled in. A document someone has to find does not.
- Write it down as the work happens, not from memory on Friday. Memory underestimates the small interruptions, the two-minute checks and the customer who calls, and those add up the most.
- Log corrections even when they are embarrassing. They are the most expensive rows in the sheet.
The three numbers
At the end of the week, three numbers:
- Minutes per order: the total minutes, divided by the number of orders that week.
- Retyped fields per order: how many facts were typed again, on average.
- Corrections: how many orders needed something fixed, and how long each fix took, including the customer’s e-mails about it.
Multiply the minutes per order by a normal month’s orders and you have the monthly time the copying costs. Multiply that by what an hour of that person’s time costs the business, and you have a number to compare with the price of automating it.
The corrections deserve their own line. A wrong invoice costs time twice, once to make and once to correct with a credit note. A parcel sent to the wrong office costs a second delivery, a customer who waited, and sometimes a customer who does not come back. Those costs are real even when nobody writes them down.
A shop sends 300 orders a month. The week’s log shows 72 orders, 290 minutes of work outside the shop, and 4 corrections.
minutes per order 290 / 72 ≈ 4 minutes
monthly copying time 4 × 300 ≈ 20 hours
corrections 4 a week ≈ 17 a monthBroken down by task, the log shows 2 minutes per order on courier labels, 1.5 on invoices and the rest on tracking questions. Three of the four corrections were courier offices typed wrongly.
Where the mistakes come from
The log usually shows mistakes clustering around a few fields, the ones that are long, precise and easy to mistype:
- addresses, and especially courier office or locker codes
- phone numbers, which couriers need and customers mistype
- VAT and company numbers on invoices
- the cash-on-delivery amount, which has to match the order to the cent
- product codes and quantities, when a picking list is written by hand
Every one of them already exists in the order. A field that is typed once, by the customer, at checkout, and then read by every document after it, cannot be mistyped a second time. That is the whole argument for connecting the systems, in one sentence.
What to automate first
Rank each task in the log by how often it happens, times how long it takes, plus what its mistakes cost. In the example above, that ranking is clear: courier labels first, because they take the most time and cause most of the corrections; invoicing second; tracking questions third, because automatic "your parcel is on its way" e-mails remove most of them.
- If most of the minutes go into invoicing, connect invoicing first.
- If the corrections cluster around addresses and offices, stop retyping them, usually by creating the courier label from the order.
- If customers keep asking where their parcel is, send them the tracking when it ships.
- If the total is small, the honest answer may be that automation is not worth it yet. Twenty minutes a week does not justify a project.
Automate the top item, then measure again before moving to the next. The second item often looks different once the first is gone.
What to keep manual
Some tasks should stay with a person even when they could be automated:
- checking an address that looks wrong
- a very large or unusual order
- confirming a bank transfer, because only the business can see its bank account
- packing, and checking what goes in the box
- anything where the right answer needs judgement rather than copying
The goal is not a shop with no people in it. It is a shop where people spend their time on the things only people can do.
Measure again afterwards
Keep the spreadsheet. A month after a change, log another ordinary week the same way, with the same columns. Compare the three numbers.
If the minutes per order and the corrections have not moved, the change did not remove the copying, whatever it cost and whatever it promised. If they have, you now have a number for what the change is worth every month, which is also the best argument for, or against, the next one.
For what it looks like when the copying is removed on a real site, From order to invoice to shipping label walks through a live site’s order flow step by step, including what it still does by hand and why.
Want a second pair of eyes on the numbers?
Send me what your week looked like.