How to work out what your manual processes are really costing you
Every business has a few processes that everybody complains about and nobody has ever costed. They feel expensive. That feeling is usually right, and it is almost never precise enough to act on.
Putting a number on it is not difficult. It takes about an hour of arithmetic and two weeks of paying attention. What follows is the method, with the parts people usually get wrong.
Start by writing down what actually happens
Pick one process. Something like getting a purchase order approved, or producing the monthly sales report, or onboarding a new employee.
Now write down every step, including the ones nobody counts. The waiting counts. The chasing counts. The bit where somebody exports a file and emails it to somebody else so they can import it counts, and is usually where the real cost is hiding.
Do this by watching, not by asking. If you ask a team how a process works you will get the version from the manual. If you sit with somebody while they do it, you will find the three workarounds they invented in 2023 that everybody has quietly adopted.
The arithmetic
You need four numbers.
- How long one run of the process takes, measured rather than estimated. People underestimate familiar tasks by a wide margin, usually because they forget the waiting.
- How often it happens. Per week is the easiest unit to think in.
- The loaded hourly cost of whoever does it. Not their salary divided by 2,080. Salary plus benefits plus employer contributions plus a share of overhead. For most small businesses that lands somewhere between 1.25 and 1.4 times base salary.
- The rework rate. What fraction of runs have to be redone or corrected, and how long the correction takes.
Then multiply. Time per run, times runs per week, times 52, times loaded hourly cost. Add the rework. The number you get is the annual cost of that process continuing to work exactly as it does today.
A worked example. Invoice entry takes eleven minutes a time, happens around ninety times a week, is done by somebody on a loaded cost of 480 pesos an hour, and about one in twelve needs correcting at roughly fifteen minutes each. That is a little over 143 hours a year on entry, plus about 97 hours on corrections. Call it 240 hours, or around 115,000 pesos a year, for one task nobody has ever put on a budget line.
Do that for five processes and you usually have somebody’s salary.
The costs the arithmetic misses
The hours are the easy part. Three other costs matter and none of them show up in a spreadsheet.
Delay is the first. An approval that sits in an inbox for three days is not costing you three days of labor. It is costing you whatever the delay causes downstream, which might be a supplier discount you missed or a customer who went elsewhere. Measure the elapsed time from request to completion as well as the hands on time. The gap between those two numbers is often enormous and is invisible in a labor calculation.
Errors are the second, and they get more expensive the later they are found. A number typed wrong at entry costs a minute to fix that afternoon. The same number, found by a client six weeks later, costs a phone call, a credit note, and some of your reputation.
The third is the one people leave out entirely. Key person risk. If one person is the only one who knows how the month end close actually works, the process does not cost you their hours. It costs you their hours plus the standing risk that they resign, get sick, or go on leave in the wrong week.
Which processes are worth changing
Not all of them, and not the most annoying one.
Work that repeats often, follows consistent rules, and moves structured information between systems is where change pays for itself quickest. Copying figures from one screen into another. Routing a request to whoever is next. Producing the same report every Monday from the same source.
Work that is rare, or needs judgment every time, or depends on information that arrives in a different shape each time, is far harder to change and usually not worth it. A process that runs four times a year will not repay much effort no matter how tedious it is.
There is a decent rule of thumb here. If you cannot write the rules down clearly enough that a new hire could follow them without asking questions, the process is not ready to be automated. It might be ready to be simplified first, which is a different and often cheaper piece of work.
Simplify before you automate
This is the step almost everybody skips, and it is the one with the best return.
A lot of processes contain steps that exist because of something that stopped being true years ago. A form that gets printed because a manager used to want it on paper. A double approval introduced after one bad invoice in 2019. A spreadsheet maintained in parallel with the system because somebody did not trust the system when it launched.
Automating those steps preserves them forever and makes them harder to remove. Ask of every step whether the reason it exists is still true. You will usually delete two or three, and the process gets faster before anybody has spent anything on software.
What to expect afterwards
Two things worth setting expectations on.
The hours you save rarely turn into a smaller payroll. They turn into the same people doing work they were hired for and did not have time to do. That is a genuine gain and it is much harder to put in a business case, so decide in advance which one you are actually buying.
And automated processes are not maintenance free. A supplier changes their invoice layout, a system gets updated, somebody adds a field. Things break quietly, and a broken automation is worse than a manual process because nobody is watching it. Whoever owns the process needs to keep owning it after it stops being manual.
None of this argues against making the change. It argues for going in with a real number, so that six months later you can tell whether it worked.
