How to calculate the ROI of automating a manual process

To calculate the ROI of automating a manual process, add up what the process costs you today (time at loaded hourly cost, errors and delays), estimate how much of that the automation realistically removes, and compare that annual benefit with the total cost of building, running and maintaining it. The payback period is the one-off cost divided by the monthly net benefit.

The arithmetic is simple. The hard part is honest inputs: a measured baseline instead of a guess, and a realistic allowance for the exceptions that will still need a person. This guide gives you the formulas, a fully worked example and the pitfalls we see most often.

The formula

Work with annual figures, then derive payback and multi-year ROI from them.

Annual benefit   = time savings + error savings + delay savings
Time savings     = hours saved per year x loaded hourly cost
Error savings    = errors avoided per year x cost per error
Delay savings    = cost of waiting avoided per year

Annual cost      = run cost + maintenance cost
One-off cost     = build cost (+ internal time spent on the project)

Net annual benefit = annual benefit - annual cost
Payback (months)   = one-off cost / (net annual benefit / 12)
ROI over N years   = (N x annual benefit - one-off cost - N x annual cost)
                     / (one-off cost + N x annual cost)

Use a period of three years for most process automation. Shorter undersells a system that will run for years; longer assumes the process will not change, which is rarely true.

Benefit and cost inputs

Time saved at loaded hourly cost

Hours saved is not the hours the process takes today. It is the hours today minus the hours people will still spend after automation, mainly on exceptions, reviews and monitoring.

Value those hours at the loaded hourly cost, not the salary. A simple way to estimate it:

1. Take annual gross salary.
2. Add employer contributions, benefits and a share of overheads such as office, equipment and software. Your finance team will have a standard percentage.
3. Divide by productive hours per year: working hours minus holidays, sick leave, meetings and training.

Error cost

Each error has a cost: time to find and fix it, plus direct consequences such as credit notes, re-shipments, penalties or a lost customer. Estimate a typical cost per error and multiply by the number of errors per year that automation would prevent. Be conservative: automation removes typing and copy-paste errors well, but not errors in the source data.

Delay cost

Manual processes run in batches, so work waits. Delay can cost money through late payment fees, missed early-payment discounts, slower invoicing to your own customers, stock held longer or goods that cannot be released. Only include delay costs you can trace to a real figure.

Build cost

Design, development, testing, data migration and rollout. Include your own team's time: process owners mapping rules, staff testing, and the parallel run where old and new processes operate side by side. The main cost drivers are the number of systems involved, the quality of their APIs, how many exception types there are, and audit requirements.

Run cost

Hosting, licences, per-task or per-execution fees of any platform you use, and third-party API charges. If you use a tool priced by volume, model how this grows with volume. Our comparison of n8n, Zapier and custom code explains how the different pricing models scale.

Maintenance cost

Systems change. APIs are versioned, suppliers change formats, rules are updated. Budget a yearly maintenance allowance; a percentage of the build cost is a common starting point, adjusted for how often the connected systems change.

A worked example (hypothetical)

Example, all numbers hypothetical. These are placeholder figures chosen to make the arithmetic easy to follow. They are not our prices or a real client's numbers. Replace each one with your own.

A finance team of 4 spends about 5 hours each per week matching supplier invoices against orders: 20 hours per week in total, over 46 working weeks.

Benefits

• Hours today: 20 h/week x 46 weeks = 920 h/year.
• Hours after automation: exceptions still need about 4 h/week = 184 h/year.
• Hours saved: 920 - 184 = 736 h/year.
• Loaded hourly cost: salary of €48,000, plus 25% for employer costs and overheads = €60,000, divided by 1,500 productive hours = €40/h.
• Time savings: 736 h x €40 = €29,440 per year.
• Error savings: 10 errors a month, each costing 1.5 hours to fix (€60) plus €50 in direct costs = €110 per error. 120 errors a year, of which automation prevents 70%: 84 x €110 = €9,240 per year.
• Delay savings: €2,000 a year in late fees and missed early-payment discounts, assume half is avoided = €1,000 per year.
• Annual benefit: €29,440 + €9,240 + €1,000 = €39,680.

Costs

• One-off build cost: placeholder of €30,000, including internal time.
• Run cost: €1,200 per year for hosting and services.
• Maintenance: 15% of build cost = €4,500 per year.
• Annual cost: €1,200 + €4,500 = €5,700.

Results

• Net annual benefit: €39,680 - €5,700 = €33,980, or about €2,832 per month.
• Payback period: €30,000 / €2,832 = about 10.6 months.
• 3-year ROI: benefit of €119,040, total cost of €30,000 + €17,100 = €47,100, net €71,940. ROI = €71,940 / €47,100 = about 153%.

Now stress-test it. If exceptions take 8 hours a week instead of 4, time savings drop to €22,080 and payback stretches to about 13.5 months. Running your numbers at a pessimistic case tells you whether the project still makes sense if things go less well than planned.

How to measure the baseline

A guessed baseline is the most common reason ROI cases fail. Measure before you build:

1. Time it: ask the people doing the work to log time on the process for two to four weeks, including interruptions and follow-ups, not just the core task.
2. Count volumes: records processed per day or week, and how volume varies through the month.
3. Count exceptions: how many records need extra work, and how long each type takes.
4. Count errors: from correction logs, credit notes, complaints or rework tickets.
5. Trace delays: how long items wait between steps, and what that waiting costs.
6. Shadow the work: watching the process often reveals steps nobody mentioned, such as checking a second system or emailing a colleague.

Our article on the automation discovery process describes how we run this step.

Soft benefits worth listing

Some benefits are real but hard to price. List them separately rather than forcing a number:

• Scalability: handling more volume without hiring in proportion.
• Speed of detection: problems surface in minutes instead of the next morning.
• Audit readiness: every decision logged automatically.
• Staff retention: people spend their time on judgement, not copy-paste.
• Better data: structured records you can report on.

For example, the reconciliation system we built for a medical and home supplies distributor removed about 20 hours a week of manual work, but the team also gained real-time discrepancy detection and an automatically populated compliance log, which cut audit preparation time. The case study has the details, and our invoice reconciliation build vs buy guide covers the approach.

Pitfalls to avoid

• Overestimating savings: assuming 100% of the time disappears. Something always remains: exceptions, reviews, monitoring.
• Ignoring exceptions: a small share of unusual cases can take a large share of the effort. Measure them separately.
• Using salary instead of loaded cost, or using loaded cost but forgetting that saved hours only create value if they are reused or avoid hiring.
• Leaving out internal time: process owners and testers spend real hours on the project.
• Forgetting maintenance: a system with no maintenance budget slowly breaks.
• Volume-based fees: per-task pricing that looks small in a pilot can grow quickly at full volume.
• Automating a broken process: fix or simplify the process first, then automate it.

Frequently asked questions

What is a good payback period for process automation?

There is no universal threshold; it depends on your cost of capital, how stable the process is and how long the system will run. Agree a target with your finance team before you build the case, and check it against the pessimistic scenario, not just the expected one.

Should we count saved hours if nobody is let go?

Yes, if the hours are reused for valuable work or avoid future hiring as volume grows. Be explicit about which, because an hour saved that is not reused does not show up in the accounts.

How do we estimate costs before we have a quote?

List the systems involved, the number of exception types and the audit requirements, since these drive cost most. Then ask for an estimate based on that list, and add your own internal time on top.

How do we check the ROI after go-live?

Repeat the baseline measurement three to six months after launch: time spent, exceptions, errors and delays. Compare against your original case and adjust the rules or exception handling where the gaps are largest.

If you have a process you suspect is costing more than it should, our business process automation service starts by measuring it. Book a call and we will help you build the ROI case, or request a quote if you already have the numbers.