How do you work out what a manual process actually costs?

Start with a number, not an opinion

Every manual process has a cost. Almost none of them have a number against it, which is why they survive budget after budget: nobody can put the process next to the thing competing with it and say which is worse.

The arithmetic isn’t complicated. Hours, rate, frequency. What makes it hard is that nobody has ever timed the process, and the person doing it will underestimate it, because the parts they have automated in their own head do not feel like work any more.

The three numbers

How long it takes, measured rather than remembered. Ask someone how long onboarding a new starter takes and you’ll get the time the form takes to fill in. You won’t get the follow-up email, the chase three days later, the correction when IT provisions the wrong licence, or the ten minutes spent working out who approves it this time. Sit with the person and time one, end to end, including the waiting.

What an hour costs. Not the salary. The fully loaded cost: superannuation, leave loading, payroll tax, the seat and the software. For most Australian mid-market organisations that lands somewhere between 1.3 and 1.5 times base salary. Finance already has this number for headcount planning, and using theirs rather than inventing one means the result survives the first conversation with the CFO.

How often it happens. This is where the number usually turns out bigger than anyone expected. A process that takes forty minutes is nothing. A process that takes forty minutes thirty times a month is twenty hours, every month, forever.

Multiply, then check the second number

Hours multiplied by rate multiplied by frequency gives you the direct cost. That figure is defensible and you should lead with it.

Then there’s the cost you can’t invoice: the work that doesn’t happen because the week went on the manual process. That one is harder to evidence, so we don’t put a dollar figure on it. We name it, and we let the direct number carry the case.

A real one

Achieve Australia runs thirty onboarding and offboarding events a month across 1,700 staff. Each one touched five systems, with no single record of what had happened to whom.

Timed end to end, the manual pipeline came to 120 hours a month. Against the loaded cost of the people doing it, that was $126K a year, and none of it appeared as a line item anywhere. It was distributed across HR, IT and service managers in slices small enough that nobody had ever added them up.

The build that replaced it is a single request that starts the whole pipeline, with provisioning driven off that request and an audit record at the other end. The 120 hours came back. You can read the full story.

What to do with the number

Three things change once you have it.

The first is that the process becomes comparable. It can be ranked against other things asking for the same budget, and it usually wins, because most of what it’s competing with doesn’t have a number at all.

The second is that it sets the size of the fix. A process costing a few thousand dollars a year doesn’t justify a six-week build, and knowing that stops you buying one. We’ll tell you when that’s the case rather than scope the work anyway.

The third is that it gives you something to measure against afterwards. The same three numbers, taken again six months later, are the only honest way to know whether the thing you paid for worked.

Where to start

Pick the process people complain about most. Not the biggest one, and not the one that looks worst on a diagram — the one that generates the most complaints, because volume of complaint tracks frequency, and frequency is the multiplier that makes the number large.

Time one instance properly. Get the loaded rate from finance. Count last month’s instances from whatever system logged them. You’ll have a defensible figure in an afternoon, and it will be bigger than you expected.