How to calculate the cost of a process and manual work before you decide on automation

When I ran operations departments, I heard many times that some process "takes a huge amount of time" and needs to be automated. Sometimes that was true. Just as often, once the numbers were in, it turned out that the problem lay elsewhere and the automation would have paid for itself only after several years. That is why, before a decision is made, I calculate the cost of the process. No special tools are needed: a spreadsheet, a few days of measurement and honest assumptions are enough.

Why intuition fails here

People's feelings are about what is tedious, not about what is expensive. A boring task done once a month seems enormous, while a five-minute correction repeated dozens of times a day by the whole team goes unnoticed. On top of that, time estimated from memory is usually understated for routine things and overstated for disliked ones. Without numbers it is also hard to compare several candidates for change and pick the one that will deliver the most.

Step 1: pick one process and count the volume

Start with a single, specifically named process, for example "entering orders from emails into the system" or "matching invoices to deliveries". Establish where it starts and where it ends. If the boundaries are unclear, a short process mapping exercise will help.

Then count the volume: how many cases pass through the process per week or per month. It is best to take the data from the system, the mailbox or a register. Watch out for seasonality. If December has twice as many orders as July, use the average for the whole year, not for the quietest week.

Step 2: measure the time for one case

This is the most important and the most frequently skipped step. Instead of asking roughly how long it takes, measure it. A sample of 20 to 30 cases, handled by different people on different days, is enough. The employee notes the time they start and finish working on a given case. From the sample you calculate the average and look at the spread, because large differences between people are in themselves information about the process.

Measure working time, not waiting time. If a document sits in the approval queue for three days, that is not a labour cost, but it does lengthen the lead time and often generates additional phone calls and reminders. Record both times separately. The first will be used to calculate the cost, the second to assess how the process affects the customer.

Step 3: establish the full cost of an hour of work

Gross salary divided by the nominal number of hours is not enough. The full cost of an hour includes three elements:

• pay together with contributions and other employer-side costs,
• overheads linked to the position: workspace, equipment, licences, a share of management costs,
• the actual number of hours worked, that is after deducting holidays, public holidays and absences.

The exact values depend on the form of employment and the company's structure, so it is best to ask HR or controlling for them. If that is not possible, adopt a single rate for the whole team and describe clearly where it came from. One reasonable rate that everyone knows is better than apparent precision.

Step 4: add errors, corrections and delays

Manual work rarely ends with the first attempt. Count what percentage of cases need a correction and how long one correction takes, including clarifying it with the customer or another department. The cost of errors is the number of cases times the error rate times the correction time times the rate.

Record separately the consequences that are not working time: contractual penalties, lost early-payment discounts, complaints, downtime. These items are sometimes larger than the cost of the hours themselves, but they are harder to estimate. Enter only what can be shown on data from recent months.

Step 5: put it together as the annual cost of the process

The formula is simple:

annual process cost = volume × time per case × hourly rate + cost of errors and corrections

If you find it easier to think in team hours than in cases, you can calculate it like this: number of people × hours per week spent on the process × number of working weeks in the year × rate. For a quick estimate, use the manual work cost calculator on our home page.

Worked example

The figures below are illustrative and rounded; they serve only to show the method. Let us assume that in the order-handling department 8 people each spend 6 hours a week retyping data between systems. I count the working year as 46 weeks, because the rest is taken up by holidays, public holidays and absences. I take the full cost of an hour of work as PLN 70.

• hours per year: 8 × 6 × 46 = 2208 h,
• annual cost: 2208 h × PLN 70 = PLN 154,560.

Now let us assume that the change will cut this time by 30%. That gives 2208 h × 0.3 = 662.4 h, or around 662 hours a year, and PLN 154,560 × 0.3 = PLN 46,368 a year.

I take 30% rather than 80% deliberately. Automation rarely removes all of the work. Exceptions remain, along with checking the results and handling cases the system will not recognise. A cautious assumption protects you from disappointment and gives you an argument that will hold up in front of the board.

Compare it with the cost of the change

A saving only makes sense when set against the cost of the change. To the implementation cost I add preparing and building the solution, licences, maintenance in subsequent years, and the time of the process owner and the team spent on testing, fixes and learning. This last element is the one most often left out, and in practice it can be significant.

I calculate a simple payback period like this:

payback period = implementation cost / annual net saving

The annual net saving is the saving minus the annual maintenance cost. In the example, with an illustrative implementation cost of PLN 40,000 and maintenance of PLN 6,000 a year, the net saving is PLN 46,368 minus PLN 6,000 = PLN 40,368, and the payback period is 40,000 / 40,368, or around 1 year. This is not a full investment analysis, but for a decision about a single process it is usually enough. If, under cautious assumptions, the payback period exceeds two or three years, it is worth first looking for a cheaper approach, for example simplifying the process without any tools.

Benefits you cannot see in the spreadsheet

Some of the effects are hard to convert into money: fewer mistakes, predictable lead times, less dependence on the one person who "knows how to do it". It is worth describing them alongside the calculation, but not attaching invented amounts to them. If the decision only holds up thanks to such items, that is a signal that the hard numbers are weak. I later measure these effects with the indicators I write about in the article on operational KPIs.

The most common mistakes in the calculation

Counting hours that will not be freed up

If each of eight employees gains 45 minutes a week, nobody will be moved to other work and the time will dissolve into the day. The saving is only real when you know what the team will do with the recovered time: handle a larger volume without new headcount, take on tasks there were no hands for, or cut overtime.

Leaving out maintenance

An automated solution needs looking after. Document formats, systems and rules change. A calculation without the maintenance cost looks better than it will turn out in reality.

Automating before simplifying

If the process has unnecessary steps, double approvals and workarounds, automation will cement them in place and raise the implementation cost. Simplify first, then calculate again. I write about how to assess whether a process is ready in the article when to automate a process, and when not yet.

Volume from a single week

Measuring in an atypical period distorts the whole result. Check data from several months before you treat the volume as representative.

Where to start

Pick one process, count the volume, measure the time on a sample, establish the rate and add the errors. This usually takes a week or two and gives you a number that can be discussed. If you want to do it faster and with someone from outside, take a look at the Automation and AI page or check out the Operational Scan, in which we measure the process together with your team and calculate where it is losing time and money.