How to decide whether a manual process is worth automating

Not every repetitive task is worth automating. Some automations pay for themselves within months. Others cost more to build and look after than the time they save. Here’s a simple way to tell them apart.

Start with three numbers

For any manual task, estimate three things.

1. How often it happens, per day, week or month.
2. How long it takes each time, including the small interruptions around it.
3. What a mistake costs. A wrong invoice, a missed order or an unhappy customer all have a price, even if it’s rough.

The first two give you the time the task takes. The third is often the bigger number, and the one most people forget.

A worked example

Say someone spends 15 minutes copying each new order from an email into a spreadsheet and then into the accounting system. There are 40 orders a week.

That’s 10 hours a week, or roughly 500 hours a year. If a few orders a month are copied wrongly and each mistake takes an hour to sort out, add a few dozen more hours, plus whatever it costs in customer goodwill.

Now compare that with the cost of automating it: building the automation, testing it on real orders and looking after it over time. If the automation costs less than about a year of the manual work, it’s usually a good investment. If it would take five years to pay back, it probably isn’t, unless mistakes are very costly.

Good candidates

• Tasks that happen often and follow the same steps each time.
• Copying information between systems that both offer a way to connect.
• Reports built from the same sources in the same way every week.
• Reminders and handovers that depend on someone remembering.

Poor candidates

• Tasks that happen a few times a year.
• Work that needs judgement every time, such as handling a complaint.
• Processes that are still changing, or that people disagree about.
• Tasks where the inputs vary so much that every case is different.

A poor candidate isn’t always a no. Sometimes part of the task can be automated while a person handles the rest. Pre-filling a form for someone to check is often worth more than trying to remove the person completely.

Hidden costs to include

Automation isn’t free once it’s built. The systems it connects to will change. Someone needs to know when it fails and what to do about it. Include this in the comparison, and design the automation so failures are visible rather than silent.

The question behind the question

Sometimes the best answer isn’t to automate a task but to remove it. Before automating a step, ask why it exists. If the answer is “because the old system couldn’t do something”, there may be a simpler fix.

If you’re unsure, it helps to understand the process properly first. The numbers are much easier to estimate once you can see how the work really happens. And if you’d like a second opinion on a specific process, tell us about it.