The hidden costs of manual processes: how to work out whether automation will pay off

The profit and loss account shows salaries, materials and external services. It does not show how much of the order processing department’s salaries goes on retyping data from emails into the system. Nor does it show how much errors in that data cost.

The cost is scattered: some in salaries, some in corrections and complaints, some in overtime at the peak of the season. Each of these items on its own looks like a normal cost of running a business. Only when they are brought together for a single process do they show an amount you can do something about.

This article focuses on the cost-benefit calculation – comparing the cost of a process with the cost of changing it. If you first want to measure the time spent on manual work across the whole company, or the cost of errors and information chaos, we describe this in separate guides.

A calculation for the whole company usually ends with one big number that nobody can defend. A calculation for a single process can be checked against data and compared with a specific offer. So choose one process that meets the conditions below.

A manual process costs money in five places, but not all of them are suitable for a calculation for the management board. Include in the core calculation only the items you can show from data for recent months. Describe the rest alongside as additional benefits, with no amount or with a cautious range.

The table below sets out how to calculate each item and where to look for the data. The order reflects how easy each item is to prove.

Take a hypothetical distribution company. Customers send orders by email, in Excel files, PDFs and in the body of the message. Three people in the order processing department retype them into the ERP – about 1,200 orders a month.

Measurement on a sample of 40 orders showed an average of 8 minutes per order. That gives 1,200 × 8 = 9,600 minutes, or 160 hours a month. We assume a full hourly cost of PLN 70: a gross salary of PLN 7,000 plus employer contributions comes to about PLN 8,430, and about 120 hours a month are left for process work.

Working time therefore costs 160 × PLN 70 = PLN 11,200 a month, or PLN 134,400 a year. On top of that come errors: 2% of orders, i.e. 24 a month, need correcting, and each correction costs PLN 150 on average. That is PLN 3,600 a month, or PLN 43,200 a year.

In total, the process costs PLN 177,600 a year. We have not included delays in shipping orders received in the afternoon or team fatigue during the season. We describe these alongside the calculation as additional benefits.

Automation rarely handles a hundred per cent of cases on its own. In our example, we assume the system will read and enter 70% of orders, with a person only reviewing them. The remaining 30% – unusual files, orders with comments – are still handled manually. That 70% is an assumption that must be confirmed in a pilot on real documents.

That leaves 360 orders × 8 minutes = 48 hours of manual work. Reviewing the 840 automatically entered orders at one minute each takes another 14 hours. That is 62 hours instead of 160 in total, or 98 hours fewer. That is 98 × PLN 70 = PLN 6,860 a month, or PLN 82,320 a year.

We also assume that errors will be halved, which gives PLN 1,800 a month, or PLN 21,600 a year. The target benefit is therefore PLN 103,920 a year, or PLN 8,660 a month. That is less than 60% of the cost of the process – and that is how it should be in an honest calculation.

The most common mistake in cost-benefit calculations is assuming full benefits from day one. In practice, the first few months are taken up with building and testing, then people learn the new way of working, and the system is fine-tuned on unusual cases. The ramp-up schedule below is an assumption for this example – adjust it to your own project.

On these assumptions, the first year is equivalent to 6.3 months of full effect (3 × 0 + 3 × 0.5 + 6 × 0.8). That gives PLN 8,660 × 6.3 = PLN 54,558, or 52.5% of the target annual benefit. The full PLN 103,920 only appears in the second year. If someone promises a full payback within a few months on a schedule like this, ask them to show you the calculation.

The supplier’s offer is only part of the cost. The other part is your own people’s work and expenses that appear only after launch. Before comparing with the benefits, add the items below to the calculation.

Instead of waiting for a quote and wondering whether it is high, turn the question around. Decide over what period the investment should pay for itself, and add up the benefits over that period. In our example, two years give PLN 54,558 + PLN 103,920 = PLN 158,478.

From this amount, subtract the costs from the previous section: your people’s time, plus licences and maintenance over two years. What remains is the ceiling on the implementation price at which the investment pays back in two years. If the offer exceeds it, you have three options: extend the payback period, narrow the scope or walk away.

Prepare the same calculation in three variants: cautious (for example 50% of orders handled automatically), base (70%) and optimistic (85%). If the investment only stands up in the optimistic variant, that is a warning sign.

The management board needs one page, not a twenty-slide presentation. Show the cost of the current process broken down by item, with the source of every figure. Next to it, show the target benefit, the first-year benefit and the three variants.

State plainly what will happen to the hours freed up. Will the company handle more orders without hiring? Will someone move to looking after key customers? Or perhaps you will not replace a person who is retiring? Without this answer, the “hours saved” will remain on paper.

Finally, decide how you will measure the effect: the same indicators as before implementation, i.e. time per order, number of corrections and the team’s hours. Measurements after three and after twelve months will show whether the assumptions held.

Not every manual process is worth automating. The calculation usually comes out badly in a few typical situations. In those cases it is better to simplify the process, or leave it as it is and choose a different one.