How You Can Tell After a Few Months Whether an Automation Has Paid Off in Your Business

After a few months, your automation either feels good or feels like a nuisance. Neither is evidence. Whether it has paid off shows in three figures you recorded before the start and now measure again: time per case, error rate, and your customers' waiting time. Against those you set all costs, including the hours for supervision and correction.

Why your impression after three months is not an evaluation

Your impression measures what you noticed most recently, not what happened over months. A glitch last week colors your judgment more than a hundred cases that ran through in silence.

That works in both directions. When the process runs smoothly, you forget how much time it used to cost you. When it stalls once, that is mainly what you remember.

The mechanism behind it is simple. A case that runs by itself creates no event in your day. A glitch creates one: a query, a phone call, a correction. Your memory collects events, not runs.

Take an appointment request that arrives by email, lands in your calendar automatically and triggers a confirmation. If that runs cleanly three hundred times, you fail to notice it three hundred times. If one appointment lands in the calendar twice, you talk about it for a week.

An impression means you are judging from memory. A figure means you are comparing two measurements that came about the same way. For your decision on whether to expand, adjust or switch off the automation, only the second one carries weight.

Your impression is not worthless, though. It shows you where to look. If the process feels like a nuisance to you, that is a pointer to a spot you can measure. The impression asks the question, the figure answers it.

In businesses I rarely see a lack of effort during the rollout. What I see is that the baseline is missing. Baseline means: the figure that describes your process before the automation took it over. Without it, all you have after a few months is the question of how it feels.

Which figures can you measure yourself without a controlling department?

Three figures are enough for you: time per case, error rate, and your customers' waiting time. You can collect all three with a stopwatch, a tally sheet and the timestamps in your inbox, with no extra software and no department of your own.

Time per case means: the minutes a person in your business actually works on a single run. That is working time on the case, not the time it sits around. You measure it from the first action to the last.

Define beforehand what counts as the first action and what counts as the last. For a request, that can run from opening the email to sending the confirmation. Once you have written that boundary down, you later measure the same stretch and not a different one.

Error rate means: how many out of a hundred cases you have to correct afterwards. For that you count two things, the cases in total and the corrections. A tally sheet next to the workstation is all you need.

Here too you need a definition, namely what counts as an error for you. A wrong date in a confirmation is one. A query from your customer because a detail is missing can count as well. What matters is that you apply the same rule before and after.

Waiting time means: the span between the moment your customer asks for something and the moment they have the result. Both points in time are usually already in your inbox or your order system.

This third figure is the only one your customers feel themselves. Time per case and error rate describe your business from the inside. Waiting time describes it from the outside. That is why it belongs in your evaluation, even though it never shows up in any cost calculation.

On top of that you need a fourth number, which is not a measurement but carries every calculation: the number of cases per month. A saving per case tells you nothing as long as you do not know how often the case occurs in your business. You find it in your inbox, your calendar or your order list.

You do not need more figures to begin with. Three numbers you collect cleanly carry your decision better than twelve you stop maintaining after two weeks.

When you measure decides what your figure is worth

You measure twice in the same way: once before the start, once after the new process has settled in. Whoever measures only afterwards has a figure, but no comparison.

You take the before measurement over a normal period, say two ordinary working weeks. Do not pick an exceptional week, neither the quietest nor the busiest. You write down who measured, how it was measured and on which cases.

That record looks like a side note and is half the measurement. In a few months you will no longer know whether the queries were counted or whether only one person was timed. If it is on a sheet of paper, you can build the second measurement exactly like the first.

You do not take the after measurement in the first few weeks. That is when things are still being adjusted, your team double-checks a lot, and the figure shows the transition, not the operation. For the same reason our pricing page sets a deadline of three months after which a saving has to be proven (source: product information impruvement.com, Ersparnis-Garantie).

For the second measurement, watch for comparable conditions. If your business fluctuates over the year, you compare an ordinary week with an ordinary week. Setting a busy week against a quiet one shifts your figure without anything in the process having changed.

After that you measure again at a fixed interval, for example once a quarter. A process changes because your business changes: new services, new staff, different requests. Your figure from six months ago describes the business of six months ago.

Put the date for the next measurement in your calendar as soon as you finish the current one. A measurement without a date takes place when things are quiet, and in your business things are usually not quiet.

How do you measure the time saved if nobody wrote anything down beforehand?

You reconstruct the baseline and label it as a reconstruction. That is weaker than a real before measurement, but far stronger than an estimate from memory.

The first route: have someone who used to do the case by hand do it by hand once more. Ten runs with a stopwatch give you a usable value for time per case. Use real cases from your business for this, not made-up practice examples.

The second route: use the traces your business has left behind anyway. Old emails carry timestamps for arrival and reply. From those you read off your customers' former waiting time, without anyone ever having noted it down.

The third route applies to the error rate: search old cases for corrections, meaning credit notes, second versions of a document or emails with an amendment. You will not find every one, but you will find a lower bound.

You may be thinking now: a re-enacted measurement is flattering, because the person knows they are being timed. The objection is valid. Whoever is watched works faster than in everyday life with the phone ringing and interruptions. Your reconstructed time is therefore more likely too low than too high.

For your calculation that does no harm. A baseline that is too low makes your saving smaller, not larger. If the automation supports itself even with this cautious figure, it supports itself with the real one all the more.

What matters is that you stay honest with yourself. Next to every reconstructed figure, write down how it came about. A figure with a known weakness helps you more than a figure whose origin you no longer know after three months.

For the next process you automate, you will not need these detours anymore. Then you measure before the start, and the reconstruction was a one-time task.

How do you calculate the costs honestly against the benefit?

You set all monthly costs against the monthly benefit, including the ones you never receive an invoice for. The largest of those is the working time your business spends on supervising and correcting the automation.

The calculation for the benefit looks like this: minutes per case before, minus minutes per case after, times cases per month. You convert the result into hours and multiply it by what an hour of that person's work costs you.

The after time includes everything a person still does on the case. If someone in your business checks every result before it goes out, that check is working time on the case. It belongs inside the measurement, not next to it.

On the cost side go four items: the running fees for operation and software, the hours for supervision, the hours for corrections after glitches, and the effort for adjustments when your process changes. The one-time setup costs you list separately.

Then you subtract. Benefit per month minus running costs per month gives you your net benefit. The one-time costs divided by this net benefit give you the number of months after which the setup has paid for itself.

So that you can see the steps in front of you, here is an example with assumed values. The numbers come from no real business, they only show the steps. Before: 12 minutes per case, after: 3 minutes, 200 cases a month. That is 9 minutes times 200, so 1,800 minutes or 30 hours.

If an hour of work costs you €40, your benefit is €1,200 a month. Against that stand, assumed, €300 in fees, 3 hours of supervision and 2 hours of correction, together €500. Your net benefit is €700. A setup for €3,500 would have paid for itself after 5 months in this example.

Now put in your own values. A single change tips the picture: at 50 instead of 200 cases, the benefit in the example drops to €300 and sits below the running costs. Volume often decides more than the saving per case.

I do not know a universal benchmark for the number of months, and I will not give you one. It depends on your cases, your costs and your process. Your calculation with your figures is the only one that applies to you.

One more point belongs to honesty. Saved hours only become money once they are used differently in your business. So write down what the freed-up time is now spent on: on quotes, on callbacks, on work that used to be left lying.

Error rate and waiting time you keep next to the calculation. Both can only be converted into euros with assumptions, and every assumption weakens your figure. Note them as their own line: before, after, difference. That is enough for your decision.

How do you recognize that an automation is running but bringing you little?

You recognize it by the fact that it works technically while your three figures have barely moved. Running is a technical state. Benefit is a difference between two measurements.

The first sign is duplicated work. Your team keeps its own list alongside the automation or fully checks every result. Then the after time sits close to the before time, and your calculation shows you that.

Duplicated work rarely comes from convenience. It comes because your team does not yet trust the result, and there is usually a reason for that. Ask about the incident that triggered the second list. Once it is fixed, the full check can become a spot check.

The second sign is volume. The case occurs less often in your business than you assumed when planning. A large saving per case times few cases makes a small benefit.

The third sign is unchanged waiting time. Your customers wait as long as before, even though the automated step has become faster. Then the delay sat somewhere else in your process, for instance at an approval that still waits for a person.

This sign is especially valuable to you. It shows you which step really governs your process. The next point of attack then lies there, and not in speeding up further a step that is already fast.

The fourth sign is correction time. The hours you spend after glitches are as high as the hours you save. Your net benefit then sits at zero, and you know what to work on next.

None of these signs is a verdict on your decision. Each one shows you a spot where you can readjust. Without measuring, you do not see that spot.

What do you do with the result?

You make one of three decisions: expand, adjust or switch off. Each of them is right if it follows from your calculation.

If the net benefit is clearly positive, you carry the approach over to the next process. You take your measurement with you, because you now know how to record a baseline. The second process starts with a real before measurement instead of a reconstruction.

If it is thin, you look for the biggest item on the cost side. Usually it is the manual check or the correction time. That is where you readjust, and you measure again next quarter. Change one thing, not three, or afterwards you will not know what made the difference.

If it is negative and stays negative after the adjustment, you switch it off. That is not a defeat but a decision with evidence behind it. You then have a figure where others have a hunch, and you have learned which kind of process does not suit your business.

Record the decision with a date and a reason. If you stand in front of the same question a year from now, you look it up instead of guessing again. Measured beats felt.

Figure before impression applies not only to your processes. It also applies to whether you yourself work and decide the way your goal demands, because your business grows exactly as far as you do. That is exactly where the Challenge begins: four days, the entry into Become → Do → Have. You start it at impruvement.com/en/challenge.