Most automation projects are approved on gut feel and evaluated on vibes. That is why so many of them get built and then quietly abandoned — nobody defined what success looked like before the work started.
ROI calculation for process automation is not complicated, but it requires honest inputs. This guide gives you the formula, walks through three real examples, and covers the mistakes that consistently make automation look better on paper than it performs in practice.
The ROI formula
Process automation ROI has two sides: what it costs to build and run the automation, and what it saves or earns.
Total Build Cost = Development cost + First-year maintenance
ROI (Year 1) = (Annual Savings − Total Build Cost) ÷ Total Build Cost × 100
Payback Period = Total Build Cost ÷ (Annual Savings ÷ 12) months
The key inputs you need before you can run this calculation honestly:
- How much time does the manual process currently take per week?
- What is the fully-loaded hourly cost of the person doing it?
- How often does the manual process produce errors, and what does each error cost to fix?
- What could that person do with the recovered time?
Example 1: Invoice processing
A 20-person professional services firm processes 150 invoices per week. Each invoice takes 4 minutes to receive, check, code, and enter into the accounting system. An accounts administrator earning £55,000/year (approximately £27/hour fully loaded) handles this.
Annual time cost: 10 hrs × £27 × 52 = £14,040/year
Error rate: 3% of invoices require correction (4.5/week)
Error correction time: 15 min each = 1.1 hrs/week
Error time cost: 1.1 × £27 × 52 = £1,544/year
Total annual cost of manual process: £15,584
Automation build cost: £8,000 (one-time)
Annual maintenance: £1,200
Total Year 1 cost: £9,200
Year 1 savings: £15,584 − £9,200 = £6,384
Year 2+ savings: £15,584 − £1,200 = £14,384/year
Payback period: 7.1 months
Example 2: Customer support triage
An e-commerce brand receives 300 support tickets per week. 45% are routine queries (order status, returns policy, tracking numbers) that a support agent handles in 5 minutes each. The support team costs £22/hour.
Time: 135 × 5 min = 675 min = 11.25 hrs/week
Annual cost: 11.25 × £22 × 52 = £12,870/year
AI agent handles 80% of routine tickets automatically.
Time saved: 11.25 × 80% = 9 hrs/week
Annual savings: 9 × £22 × 52 = £10,296/year
Build cost: £12,000 | Annual maintenance: £2,400
Year 1 total cost: £14,400
Year 1 net: £10,296 − £14,400 = −£4,104 (investment year)
Year 2+ net: £10,296 − £2,400 = £7,896/year
Payback period: 16.7 months
This example shows why payback period matters. The support automation is a good investment — but not in year one. If the business expects immediate payback, they will be disappointed. If they plan for an 18-month horizon, the economics are compelling.
Example 3: Weekly reporting
A operations manager at a logistics company spends 6 hours every Friday pulling data from three systems, formatting it into a board report, and distributing it. Their cost is £45/hour.
Annual cost: 6 × £45 × 52 = £14,040/year
Automated report: pulls data, formats, and emails automatically.
Residual review time: 30 min/week
Time saved: 5.5 hrs/week
Annual savings: 5.5 × £45 × 52 = £12,870/year
Build cost: £6,000 | Annual maintenance: £800
Year 1 net: £12,870 − £6,800 = £6,070
Payback period: 6.3 months
Common mistakes that inflate the numbers
- Using salary, not fully-loaded cost. Fully-loaded cost includes employer taxes, benefits, office space, and management overhead — typically 1.3–1.5× base salary. Using salary alone understates the real cost of manual labour.
- Assuming 100% automation rate. Most automation systems handle 70–85% of cases. The remainder still need human review. Factor this in or your savings estimate is wrong.
- Ignoring maintenance cost. Automation systems require ongoing maintenance as your processes, data formats, and integrations change. Budget 15–20% of build cost per year.
- Not measuring the current state first. If you do not know how long the manual process actually takes today, you cannot calculate what the automation saves. Measure first, then decide.
- Counting recovered time as savings without a plan for it. If automation saves 10 hours per week but those hours are absorbed into unfocused work, the saving is theoretical. The ROI is only real if the recovered time is redirected to measurable value.
What to measure before you automate
Most ROI calculations fail not because the arithmetic is wrong but because the baseline is guessed. If you do not know what the manual process actually costs today, every number downstream is fiction. Spend one week measuring before you model anything.
Four figures give you a defensible baseline:
- Volume. How many times does this process run per week? Count it, do not estimate it — people routinely misjudge frequency by a factor of two in both directions.
- Handling time. How long does one run take, start to finish, including the context-switching and the chasing? Time five real instances rather than asking someone how long it takes.
- Touch count. How many people handle it? A process touching four departments carries coordination cost that never appears in a single person's time estimate.
- Failure rate. How often does it go wrong, and what does putting it right cost? Rework is usually the largest hidden line in the whole calculation.
A useful sanity check: if the measured cost of a process is under roughly £3,000 a year, automating it rarely pays for itself no matter how annoying it is. Irritation and cost are not the same thing, and automation budgets should follow the second.
Payback period vs ROI: which number to use
These two metrics answer different questions, and using the wrong one is how good projects get rejected and bad ones get approved.
Payback period answers “how long until I get my money back?” It is the build cost divided by the annual net saving. It is the right metric when cash is tight, when the process might change soon, or when you need to convince a sceptical finance team — because a nine-month payback is intuitively persuasive in a way that a percentage is not.
ROI answers “how much value does this create overall?” It is the net gain over a period divided by the cost, expressed as a percentage. It is the right metric when comparing several competing projects, because it normalises for size: a £5,000 automation returning £15,000 and a £50,000 automation returning £150,000 both show 200%, which makes the comparison fair.
| Situation | Use |
|---|---|
| Limited cash, need approval | Payback period |
| Choosing between projects | 3-year ROI |
| Process may change within a year | Payback period only |
| Board or investor reporting | Both, with assumptions stated |
Whichever you use, state the assumptions in writing alongside the number. An ROI figure without its inputs is unfalsifiable, and unfalsifiable numbers get ignored the moment anyone senior pushes back. You can model both figures with our automation ROI calculator.
The benefits that never make the spreadsheet
Direct labour saving is the easiest thing to count and rarely the largest thing you gain. Three others matter and are worth naming explicitly in a business case even when you cannot price them precisely:
Cycle time. A process that took two days now takes four minutes. If that process sits between a customer and their purchase, or between a claim and its payment, the compression has revenue and retention value far beyond the staff hours saved.
Consistency. Automated processes do not get tired at 4pm on a Friday. For anything with a compliance or audit dimension, the removal of variance is often the actual justification, and error-cost savings understate it.
Retention. Repetitive data entry is one of the most reliably cited reasons capable people leave operational roles. Recruiting and onboarding a replacement typically costs a significant share of annual salary — a cost that lands in a completely different budget line from the one paying for the automation.
You do not need to fabricate numbers for these. Listing them as named, unpriced benefits alongside the hard figures is more credible than inflating the labour saving to compensate.
Getting started
Pick one process. Measure the current state honestly — time per week, error rate, cost per hour. Run the formula with conservative assumptions (70% automation rate, 20% annual maintenance). If the payback period is under 18 months, it is worth pursuing. If it is over 36 months, look for a higher-impact target first.
The best first automation is usually the one your team finds most tedious — not the most technically impressive one.