Skip to content
Efidis
Efidis

Costs and decisionsUpdated 10 min read

What manual data entry costs you a year (with a formula)

3D illustration of keyboard keys, a stack of paper forms and a lilac pencil.
In this article

Three people spend 6 hours a week each copying orders from WhatsApp into a spreadsheet, and from the spreadsheet into the invoicing system. It doesn't sound like much: one morning per person. Over a year it's 864 hours, close to half a year of one full-time employee's time, spent typing data that was already typed.

The copy tax

Call it the copy tax: what you pay every year to type a second time a piece of data that already existed. It never shows up on an invoice. You pay it in payroll hours, in orders shipped wrong and in reports someone pieces together by hand on Monday morning.

The time part isn't small. In a Smartsheet survey of workers, more than 40% said they spend at least a quarter of their work week on manual, repetitive tasks, and the biggest time sinks were email, data collection and data entry (Smartsheet, 2017). Asana surveyed more than 9,600 workers in six countries: 58% of the workday goes to "work about work," like searching for information and chasing status updates (Asana, 2023).

The error part is the one people forget. Barchard and Pace tested three ways of entering data in the lab. With single entry, errors ran at about 1% of values. Checking by eye didn't help, and both methods left roughly 30 times more errors than entering the data twice and letting the computer compare (Barchard and Pace, 2011). Worse, almost all of those errors were plausible values, like a 7 instead of a 1, so they don't jump out at anyone.

One percent sounds small until you multiply it. If your team retypes 10,000 data points a month (orders, quantities, prices, IDs), that's around 100 wrong entries every month. Some don't matter. Others end up as a wrong shipment or an invoice that has to be voided.

The formula: calculate your own cost

Run the numbers one process at a time, not for the whole company. "Moving WhatsApp orders into the spreadsheet" is one process; "reconciling inventory on Friday" is another. Then add them up.

VariableWhat goes inHow to get it
A. PeopleHow many people retype data in that processCount everyone, including the manager who builds the report
B. Hours per weekHours each person spends retyping data in that processHave them log it for two weeks; memory runs short
C. Cost per hourWhat one hour of that person costs the businessMonthly pay plus benefits and payroll taxes, divided by hours worked per month
D. Weeks per year4852 weeks minus vacation and holidays
Cost of timeA × B × C × DWhat you pay each year in copying hours alone
E. Errors per monthErrors that actually have a consequenceCount returns, credit memos, reshipments and inventory mismatches for one month
F. Cost per errorWhat it takes to fix each oneHours to correct it, a second shipment, lost stock, a discount to the customer
Cost of errorsE × F × 12What you pay each year for copying mistakes
Copy taxCost of time + cost of errorsYour number

How to get the cost per hour

Don't use pay alone. An employee costs the business their pay plus benefits, payroll taxes, paid time off and their workspace. Add up everything you spend on that person in a month and divide it by the hours they work that month. If someone earns $3,000 a month and costs you $4,000 all in, at 170 hours a month their hour is worth about $23.50, not $17.65.

If people at different pay levels do the copying, give each one a row. If the owner or a manager does it, use what their hour is worth selling or negotiating: that is usually the most expensive row.

How to estimate the cost of errors

Counting them is best: go through a month of returns, credit memos, repeat shipments and inventory differences, and flag the ones caused by a mistyped entry. If you don't have that record, estimate: data points retyped per month × 1% (Barchard and Pace, 2011) gives you the wrong entries. Then assume what share has a consequence, and correct that assumption once you have a real count.

Spreadsheets add their own risk. In audits of real-world spreadsheets done since 1995, 94% of the 88 spreadsheets reviewed across seven studies contained errors (Panko, 2005). When data is copied into one spreadsheet that then feeds another, the error travels with it. If your processes run on shared spreadsheets, see how to move off Excel without stopping work.

Worked example: a distributor taking orders on WhatsApp

This is a made-up example with round numbers. A distributor gets about 1,200 orders a month over WhatsApp. Two clerks type them into a spreadsheet and from there into the invoicing system; a third person reconciles inventory against the orders every week. Each of them spends about 6 hours a week retyping data.

ItemExample valueMath
A. People3Two clerks and the inventory reconciler
B. Hours per week6Per person, logged for two weeks
C. Cost per hour$15Pay plus benefits and payroll taxes
D. Weeks per year48
Cost of time$12,9603 × 6 × 15 × 48
Data points retyped per month9,6001,200 orders × 8 fields (customer, product, quantity, price…)
Wrong entries per month969,600 × 1%
E. Errors with a consequence10 a monthAssumption: 1 in 10 reaches the customer or the inventory
F. Cost per error$25Reshipment, credit memo and an hour from whoever fixes it
Cost of errors$3,00010 × 25 × 12
Copy tax$15,960 a year12,960 + 3,000

Close to $16,000 a year for a single process, and that's a conservative number. It doesn't count the customer who stops ordering after two botched deliveries, or the overtime at month-end close.

Also look at which part weighs more. Here, time is 81% of the total. In other businesses it's the other way around: little gets copied, but each error is expensive (a quote with the wrong price, a dosage, a measurement). That difference changes what you should automate first.

Checklist: where your business retypes data

Go through the list with the people who do the work, not just the ones who supervise it. Every item you check becomes a row in the table above.

  • Orders come in by WhatsApp, email or phone, and someone types them in again.
  • The same sale gets recorded in a spreadsheet and then in the invoicing system.
  • The price list lives in a file and gets copied into quotes or orders.
  • Inventory is reconciled by hand, comparing what was sold against what was shipped.
  • Field reps write orders on paper or on their phone, and someone at the office enters them.
  • The weekly or monthly report is built by copying data from two or more files.
  • Commissions are calculated in a spreadsheet with data pulled from somewhere else.
  • Customer payments are looked up on the bank statement and matched one by one.
  • A new customer's details get entered in more than one system.
  • Someone exports a file from one system just to import it into another.
  • There's a file only one person knows how to update.
  • When someone asks "how much did we sell yesterday?", the answer takes more than five minutes.

If you checked three or more, your copy tax is probably already higher than what a system that prevents it would cost. Do the math before you decide: the checklist tells you where to look, the formula tells you what it's worth.

What to automate first

Collecting and processing data are among the activities with the most technically automatable time: 64% and 69%, according to the McKinsey Global Institute (McKinsey, 2017). Being able to automate something doesn't mean it's all worth doing at once. Here's the order:

  1. Rank processes by annual copy taxRun every process from the checklist through the formula and sort them from highest to lowest. The top one is often a surprise: it's rarely the one people complain about most.
  2. Move up the ones with expensive errorsA process with few hours but $200 errors can outweigh one with lots of hours and cheap errors. If a mistake touches a customer or money, move it up the list.
  3. Start where the data is bornAutomate the point where the data first comes in: the order, the sale, the receiving dock. If that data starts out right in one system, the copies downstream go away on their own.
  4. Deal with the one that depends on one personIf a process stops when that person is out sick or quits, it carries a cost the formula doesn't measure. Put it near the top.
  5. Measure before and afterKeep the hours and errors you logged. Three months after automating, redo the math with the same rules. If the number didn't drop, something is still being retyped.

To decide whether a process is worth it, compare the cost of automating it with its annual copy tax. A rule of thumb: if the fix costs less than one year of tax, do it; if it costs between one and three years, look at what else it solves; if it costs more than three years, leave it for later.

When automating isn't worth it

Automation has a cost too: money, your team's time to explain the process and a few weeks of adjustment. Skip it if:

  • The process adds up to less than an hour or two a week across everyone and its errors are cheap. The tax is real, but small.
  • The process is going to change or go away in the next few months. Automating it locks in something you're about to throw out.
  • Nobody really knows how it's done. Write the procedure down and sort it out first; automating a confusing process only makes the confusion faster.
  • A feature your off-the-shelf software already has, and nobody set up, would solve it: an import, a built-in integration, a form. Check that before paying for something new.
  • The fix costs more than three years of copy tax. At that point, manual entry with a good template is still the sensible choice.

If you decide to keep entering data by hand, cut the error rate: use dropdowns instead of free text, validate ranges (a quantity can't be 0 or 10,000) and, for critical data, enter it twice and let the spreadsheet compare. That last technique left about 30 times fewer errors than checking by eye (Barchard and Pace, 2011).

What to do with your number

Once you have your copy tax, you have a budget. If it comes to $16,000 a year, any fix that costs less and gets rid of most of that copying pays for itself. If it comes to $1,500, you don't need a new system; you need a better template.

There are three paths, from simplest to most complete: set up what you already have properly, connect the tools you already use so data moves on its own, or bring sales, inventory and operations into one app where every piece of data is entered once. The third makes sense when the checklist flagged several processes that hand data to each other: if the order, the inventory and the invoice live in the same place, there's nothing to copy between them. If that's your case, this is how we build custom software.

Sources

  1. Barchard, K. A. and Pace, L. A., "Preventing human error: The impact of data entry methods on data accuracy and statistical results," Computers in Human Behavior 27(5), 2011 (opens in a new tab)
  2. Interactive Measurement Group (UNLV), summary of Barchard's data entry and data checking studies (opens in a new tab)
  3. Asana, "Anatomy of Work Global Index 2023," 2023 (opens in a new tab)
  4. Smartsheet, "Workers waste a quarter of the work week on manual, repetitive tasks," 2017 (opens in a new tab)
  5. Raymond R. Panko, "What We Know About Spreadsheet Errors," revised 2005 version (opens in a new tab)
  6. McKinsey Global Institute, "A future that works: Automation, employment, and productivity," 2017 (opens in a new tab)

Frequently asked questions

Enter every piece of data once

Efidis builds your company's app so the data that comes in with an order reaches every place that needs it on its own. On the call we'll run the numbers on your processes with you and pick the one to fix first.