A repetitive task often survives because each individual occurrence looks cheap. Five minutes to copy a record. Ten minutes to build a status update. A few minutes to chase an approval. The operating cost becomes visible only when the business measures the full pattern: how often the task happens, who touches it, what goes wrong, what waits on it and what else those people could have been doing.
Small manual tasks become infrastructure when the business depends on them every day.
This article sets out a practical way to build that picture. The aim is not to produce the largest possible number. It is to produce a number that a finance director, an operations lead and the people doing the work would all recognise as fair.
Why the cost stays hidden
Repetitive work rarely appears as a line in a budget. It is absorbed into salaries, spread across several roles and performed in short bursts between other work. Nobody owns the total. The person doing the task sees a few minutes; their manager sees a team that is busy; the finance team sees headcount. The cost is real but structurally invisible.
It also tends to grow quietly. A task that was tolerable at fifty occurrences a month becomes a constraint at five hundred, yet nothing in the operation signals the change until a queue forms or a mistake reaches a customer.
Calculate direct handling first
Start with frequency multiplied by average handling time. If a task takes eight minutes and occurs 300 times a month, the operation consumes 40 hours before any error, waiting or management overhead is counted.
Use realistic observations rather than the fastest possible completion time. Include the time required to open systems, find context, switch between tools and confirm completion. Where possible, take frequency from system evidence such as ticket counts, record creation logs, email volumes or export histories rather than from memory, and sample handling time across several people and several days.
Add coordination cost
Many repetitive tasks trigger follow-up work: asking whether something was done, finding the latest version, confirming an owner or waiting for another team. Coordination time is often spread across several people and therefore disappears from a simple task estimate.
Track messages, status meetings, handoffs and queue checking that exist because the process does not expose reliable state. A useful test is to ask: if the status of this work were visible to everyone who needs it, which of these conversations would stop?
Measure correction and exception work
Manual activity can create errors through re-keying, missed steps or inconsistent interpretation. Record how often work must be corrected and how much time the correction consumes. Include downstream rework when one bad input creates several later fixes.
Exceptions deserve their own line. The routine version of a task may take minutes, while the one case in twenty that does not fit the pattern takes an hour and a senior person's attention. Averages hide this, so record the routine path and the exception path separately.
Include delay where it matters
Not every minute of waiting has a financial value, but some delays constrain revenue, service, cash collection or capacity. A manual step that adds two days to invoicing can matter more than its ten minutes of handling time.
Separate labour savings from business-cycle improvement. This avoids pretending that every saved minute becomes cash while still recognising when faster flow has commercial value.
Account for interruption
Short tasks can be expensive when they repeatedly interrupt higher-value work. Context switching makes the effect difficult to price precisely, so avoid invented multipliers.
The research supports caution in both directions. In a laboratory study of an email-handling task with 48 participants, Gloria Mark and colleagues found that interrupted people actually finished in less time with no measurable difference in quality, but reported more stress, frustration, time pressure and effort. The lesson for a cost model is that interruption does not always show up as lost minutes; it can show up as strain that is harder to price and easier to ignore. Record how often specialists or managers are pulled into low-value administrative work and whether that work could be routed differently, and describe the effect rather than inflating the hours.
Build a defendable baseline
A practical monthly cost model can include direct handling hours, coordination hours, correction hours and any clearly attributable external cost. Use loaded labour rates only when appropriate to the decision and show assumptions separately.
Public-sector appraisal practice offers a useful discipline here. The UK Treasury's Green Book asks appraisers to compare every option with a business-as-usual benchmark: what is expected to happen if current arrangements simply continue. It also values resources at their opportunity cost, the value of the next best use they could be put to. For repetitive work, that means the baseline is not zero. It is the current cost trajectory, including growth in volume, and the time released only has value if there is a realistic better use for it.
An illustrative worked example
The following is a hypothetical example, not client evidence. Suppose a finance team re-keys supplier invoice details from email into an accounting system.
| Component | Basis | Monthly hours |
|---|---|---|
| Direct handling | 400 invoices × 6 minutes | 40 |
| Coordination | Approval chasing, roughly 3 hours a week | 13 |
| Correction | 20 corrections × 30 minutes | 10 |
| Exceptions | 15 unusual invoices × 45 minutes | 11 |
| Total | 74 |
Direct handling alone suggested 40 hours. The fuller picture is closer to double that, before any value is placed on late-payment risk or month-end delay. Every figure here is an assumption that the real team would need to replace with observation.
Compare with the lifecycle cost of change
Then compare the baseline with the lifecycle cost of improvement: design, implementation, licences, monitoring, maintenance, exception handling and change management. The Green Book also warns about optimism bias, the documented tendency for costs and timescales to come in higher, and benefits lower, than first estimated. Apply the same scepticism to automation proposals: assume the build will take longer, that exceptions will persist and that someone will need to watch the solution after go-live.
Sunk effort should not distort the decision either. Time already spent on a half-finished tool is not a reason to keep investing in it.
Do not automate merely because the number is large
A large manual cost may indicate a process that should be removed, simplified or redesigned. Automation should follow the root cause. A repeated task with unstable rules or high judgment may be a poor automation target even when the labour total is significant.
The options are wider than automate or tolerate. The task may be eliminated because nobody uses its output; consolidated because two teams do the same work; moved upstream so the data is captured correctly once; integrated so systems exchange the information directly; or deliberately retained as human work because judgment is the point. See understanding how the work actually moves before choosing.
Warning signs that the cost is being misjudged
- The estimate uses the fastest observed completion time.
- Only one role's time is counted, although several people touch the work.
- Exceptions are excluded because they are "rare".
- Every saved hour is converted to cash, with no plan for the released capacity.
- The improvement case includes build cost but not monitoring and maintenance.
A simple decision worksheet
- Occurrences per month, and how that is trending.
- Average handling time, taken from observation.
- People involved per occurrence.
- Coordination and follow-up time.
- Correction and exception rate.
- Business delay created.
- Required judgment.
- Available system interfaces.
- Expected implementation and operating cost.
What better looks like
The organisation can explain the cost of repetitive work with transparent assumptions. That evidence supports better choices about simplification, integration, automation or deliberate retention of human work. The numbers are revisited after change, so the business learns whether its estimates were sound. For the investment side of the same decision, see how to calculate the business case for automation.
Sources and further reading
- https://ics.uci.edu/~gmark/chi08-mark.pdf (last verified 10 October 2026)
- https://assets.publishing.service.gov.uk/media/698dbcd17da91680ad7f4308/The_Green_Book_2026.pdf (last verified 10 October 2026)
Related Mellorca reading
- How to find automation opportunities from operational evidence
- Why growth breaks systems that worked when the business was smaller
- How to find the manual work nobody is measuring