What does manual work really cost, beyond the hours?
Manual work costs more than the hours it consumes. KodDelta counts four layers: the paid time itself, the corrections that follow a mistake, the days a case spends waiting between desks, and the risk of a process that only one person knows how to run. Only the first layer is visible.
No accounting system has a line called "re-typing orders from email". The work is real and it is paid for, but it is spread across salaries where nobody adds it up. This page is about how to add it up — and about the three costs underneath the hours that are usually larger than them.
Why manual work stays invisible
Manual processes are nobody's job title. The sales assistant re-entering orders was hired to support customers; the supervisor compiling Friday's report was hired to run the plant. The cost lands in salaries budgeted for something else, so it never appears as a decision anyone made. It also grows quietly: each new customer or reporting requirement adds a few minutes to an existing routine, no single addition is worth objecting to, and after three years the routine consumes a day a week.
Layer one: the hours
The visible layer, and the easiest to measure. Take headcount, hours per week, the loaded hourly wage and the number of working weeks. Loaded wage means gross salary plus employer contributions divided by actual working hours, not the gross figure on the contract. One warning: do not use estimates people give in a meeting. Ask them to write down minutes at the end of each day for two weeks — the written number differs from the remembered one, in both directions, and it is the only version worth building a case on. The full method is in how to measure software ROI.
Layer two: what a mistake costs after it travels
Manual transcription produces errors. That is not a comment on your staff; it is what happens when a person copies a code from one screen to another several hundred times a week. What matters is not the error rate but how far the error travels before anyone notices. A wrong product code caught at entry costs seconds. Caught during picking, it costs a trip back to the aisle and a re-pack. Caught by the customer, it costs return freight, a credit note, a replacement shipment and a small amount of trust. Quality management has a long-standing rule of thumb that the cost multiplies at each stage — a way of thinking, not a number you can use without measuring your own.
Your own number comes from records you already keep: count credit notes, return authorisations, re-issued invoices and re-sent shipments over three months, then work out what one costs to resolve end to end.
Layer three: waiting
Most processes spend far more time waiting than working. A purchase request needs four minutes of managerial attention and waits three days for it. A quote takes twenty minutes to prepare and goes out on Thursday because the price approval sat in an inbox since Monday.
Waiting matters for two reasons. It is what your customer experiences — nobody outside your company can tell a hard task from a queued one. And it is the cheapest thing to remove, because it needs no new effort from anyone, only a system that puts the request in front of the right person immediately and records the answer. That is most of what a purchase approval workflow does.
Layer four: the process lives in one head
Manual processes accumulate exceptions, and exceptions are remembered rather than written down. Which customer gets the special discount, which supplier needs the order split, which product code the accounting package refuses unless entered a particular way — all of it lives with whoever has done the job longest. The cost appears when that person takes leave or resigns: the task takes three times as long, errors rise, and the company discovers the process was never documented. Encoding the rules in a system is not mainly about speed; it is about the process surviving the person.
A related cost: manual processes leave no usable record. You cannot answer "how many orders were changed after approval last quarter" if the changes happened by phone, so every question about the past becomes a small research project.
An illustration of the arithmetic
The table below applies the layer-one formula to four common workflows, at a loaded wage of $15 an hour across 48 working weeks. These are illustrative figures to show the method — substitute your own headcount, hours and wage.
| Workflow | Manual load | Annual cost of hours alone |
|---|---|---|
| Re-typing dealer orders | 2 people × 6 hrs/week | $8,600 |
| Stock counting and reconciliation | 3 people × 7 hrs/week | $15,100 |
| Compiling production reports | 1 person × 8 hrs/week | $5,800 |
| Chasing purchase approvals | 2 people × 4 hrs/week | $5,800 |
| Four workflows combined | 62 hrs/week across the company | $35,300 |
Corrections, waiting and key-person risk are not in that total. Layer one is the floor, not the figure.
What changes, and what does not
| Step | Manual today | With a system |
|---|---|---|
| Order arrives | Email or phone; someone re-types it | Dealer enters it once, in the format the system needs |
| Stock check | Someone looks, or guesses | Availability shown at entry; quantity reserved on confirmation |
| Approval | Form on a desk until the approver returns | Request reaches the approver on their phone; decision recorded with a timestamp |
| Reporting | Compiled by hand from several files | Generated from the records the work already created |
| Exceptions | Handled by whoever knows the rule | Still handled by a person, but with the rule written down and the history visible |
The last row matters. Automation does not remove judgement, and any proposal claiming a workflow becomes fully hands-off should be read carefully. What it removes is the repetitive majority, so the people who understand the exceptions spend their day on exceptions.
Where to start
- Pick the process with the most handoffs, not the most irritating one. A daily task passing between three people holds more recoverable cost than a monthly task one person hates.
- Measure two weeks before changing anything — hours, corrections and elapsed time. Without the before, the after proves nothing.
- Build one module, not a platform. A focused module reaches production in 4 to 6 weeks, so the second measurement happens while the rest is still a decision.
- Fund the next module from the measured result, which is also how the price bands are structured.
If you want the four layers costed against your own workflow, describe it on the quote form, or read how the same problem is handled end to end in business process automation.
Frequently asked questions
What are the hidden costs of manual work?
Four. Paid hours, visible on the payroll but rarely attributed to the task. Correction cost, which grows the further a mistake travels before anyone notices. Waiting time, the days a case sits between desks. And key-person risk, where a process only one employee understands stops when that person is away.
How do you calculate the annual cost of a manual process?
Headcount times hours per week times the loaded hourly wage times working weeks, then add the annual cost of corrections: corrections per month times the cost to resolve one, times twelve. Loaded wage means gross salary plus employer contributions divided by actual working hours.
Why does a mistake cost more the later it is caught?
Because more work has been built on top of it. A wrong code caught at entry costs a moment. Caught in the warehouse it costs a repick. Caught by the customer it costs return freight, a credit note and a replacement shipment. The escalation is a rule of thumb, not a measurement of your business.
What is waiting time and why does it matter more than work time?
Waiting is the gap between a case being ready for the next step and someone taking it. An approval needing three minutes of thought can sit for three days. Customers experience elapsed time, not work time, so waiting is usually most of what they complain about.
Which manual process should be automated first?
The one with the highest frequency and the most handoffs between people, not the one that feels most annoying. Daily repetitive work with three or more handoffs produces measurable results quickly. Rare complex tasks are better left to people until the frequent work is handled.
Does automation remove all manual work?
No, and a supplier who says otherwise is selling. Exceptions still need judgement and someone still has to check what the system produced. A realistic target is that the repetitive majority is handled automatically and your team spends its time on cases that genuinely need a decision.
Let's talk about what you need.
The 30-minute discovery call is free and carries no commitment.