To know if an automation actually saves you money, measure before and after on the same yardstick and subtract: what the task cost by hand (time times cost of the hour, plus the errors it caused) minus what it costs now (the little time it still takes plus what it costs to keep the automation alive). If the result comes out positive and holds month after month, you're saving. If not, either you're not saving, or you haven't measured it properly.
The condition almost everyone skips is measuring the starting cost before automating. If you didn't note how much time and how many errors the task cost by hand, you've got nothing to compare against later, and the saving stays a pleasant feeling that shows up in no account.
Why isn't "it saves me time" enough?
Because saved time doesn't always turn into money. If an automation makes a two-hour task take ten minutes, but the person who did it spends those two hours looking at their phone, the business hasn't saved anything: it's gained slack, which is fine, but it isn't the same as cash. Time turns into money when it's reinvested in something that produces or when it avoids hiring someone.
That's why "it saves me time" is the start of the sentence, not the end. The question that finishes it is: has that time turned into more sales, into fewer people needed, into fewer errors that cost money? If the answer is no, the saving is potential. It can be valuable anyway, but it's worth calling it by its name and not putting it in the profit column as if it were hard cash.
What exactly do you have to measure?
Four things. The first two count in favour of automating; the last two subtract and almost no one counts them:
- Direct time. How long the task took by hand, times how many times it's done. It's the obvious one and the first you see.
- Errors avoided. The mistakes doing it by hand caused —a duplicated invoice, a mistyped order— and what each one cost in money, returns or angry customers.
- Maintenance cost. What it costs to keep the automation alive: platform fees, the time to fix it when it fails, the time to update it when the process changes. An automation isn't furniture; it has to be fed.
- Build cost. What it cost to create it, spread across the months it'll last. An automation that cost a lot but lasts years weighs little per month; a cheap one you have to rebuild every quarter weighs more than it looks.
The real saving is the subtraction: (direct time + errors avoided) minus (maintenance + a share of the build). Plenty of automation that "saves time" stops saving once you add the last two lines.
How do you do it, step by step?
Six steps, and the first has to be taken before automating anything:
- Measure the current cost before touching anything. Time the task by hand: times, minutes, who does it, what their hour costs. Without this starting number, there's no comparison later.
- Note the errors too and what they cost. How many mistakes the task by hand causes and what each is worth in money and aggravation.
- Add build and maintenance. To the cost of creating it, add fees, fixes when it fails and updates when the process changes.
- Measure the cost after with the same yardstick. After a while, measure the same things in the same unit. Change the yardstick and the comparison is worthless.
- Subtract and translate into payback months. Monthly saving = before minus after minus maintenance. Build cost divided by monthly saving = months to pay for itself.
- Review it after a few months. A real saving can stop being real if the process changes or the automation starts failing. Look every few months, not once.
You don't need a sophisticated dashboard for this. A sheet with two columns —before and after— and the four lines above is enough. The hard part isn't the sum; it's having started it in time.
An example with numbers
A task of moving orders from an email into a spreadsheet, with made-up but realistic figures:
| Item | Before (by hand) | After (automated) |
|---|---|---|
| Time per month | 20 h (1 h/day) | 1 h (reviewing) |
| Cost of the time (at €15/h) | €300 | €15 |
| Errors per month | 3 wrong orders (~€150) | Almost none |
| Maintenance | — | €30 (fee + fixes) |
| Monthly cost | €450 | €45 |
Saving: €405 a month. If building it cost €1,600, it pays for itself in four months and after that it's gain. Swap the numbers for yours and you'll have your answer. If, on filling in the table, the saving comes to €30 a month, maybe that task wasn't the one to automate.
When does an automation NOT save, even when it looks like it does?
- When the process changes often. If it has to be rebuilt every quarter, the build cost isn't spread over years but over months, and it eats the saving.
- When it fails and needs watching. A fragile automation where someone spends half an hour a day checking it went well can cost more time than it saved.
- When the freed time isn't used. If no one reinvests the saved hours and no hiring is avoided, the saving is slack, not cash. Legitimate, but don't book it as profit.
What mistakes do people usually make?
Almost all come from measuring badly or not measuring.
- Not measuring the before. The mother mistake. Without a starting cost, any saving is a feeling, not a number.
- Forgetting maintenance. The build cost gets counted and what it costs to keep alive gets ignored, which is sometimes what decides whether it pays off.
- Counting saved time as cash always. It only is if it's reinvested or avoids hiring. Otherwise it's slack.
- Measuring it once. The first month's saving can evaporate by the third if the process changes. You review it, you don't take it for granted.
Frequently asked questions
How do I calculate the saving of an automation?
By subtracting two costs measured on the same yardstick: what the task cost by hand (time times cost of the hour, plus the errors it caused) minus what it costs now (the little time it still takes plus the maintenance of the automation). If the result is positive and steady, you're saving. The essential condition is having measured the starting cost before automating.
How long does it take to recover what automating costs?
You work it out by dividing what it cost to build by what it saves per month. If building it cost the equivalent of 2,000 euros and it saves 400 a month, it pays for itself in five months; from there on, it's gain. As a rule, an automation that doesn't pay back in under a year is worth looking at closely: either the saving is smaller than it seemed or maintenance is eating the benefit.
Do I count saved time as real money?
With caveats. Saved time is real money if it's reinvested in something that produces or if it avoids hiring. If it simply makes a two-hour task take ten minutes but that person doesn't spend the other two hours on anything useful, the saving is potential, not hard cash. It's worth telling apart the saving that turns into cash from the one that only eases the load.
When does an automation cost more than it saves?
When the process changes so often that it has to be rebuilt every so often, when it fails a lot and someone spends more time watching it than they were saving, or when a barely-done task was automated. In those cases maintenance eats the saving, and sometimes it works out cheaper to go back to doing it by hand.
Without the before number, no saving counts
To know if an automation actually saves money, measure before and after on the same yardstick and subtract, always adding the cost of maintaining it. The mother mistake is not measuring the starting cost: without it, the saving is a pleasant feeling that shows up in no account.
Tell apart the time that turns into cash from the one that only gives slack, count the maintenance, and review the number every few months, because a real saving can evaporate when the process changes. A sheet with two columns is enough; the hard part is starting it in time.
Not sure whether what you automated pays off?
If you have automations running but couldn't say how much they save, we can set up the before-and-after sum and put clear numbers on it, maintenance included. Sometimes it turns out it pays off easily; sometimes, that one was cheaper by hand. Tell me which automation makes you wonder.
See the data analysis service Let's talk about your case