How to measure a process before automating it

Most failed automation projects did not fail technically. They started from an impression — “this takes us ages” — which, once measured, turned out to be something other than it seemed.

The measuring takes two weeks, is done on paper or in a spreadsheet, and needs no tools.

First: name the process exactly

Not “the paperwork”. Not “the admin side”. Something with a clear beginning and end:

“From the moment a request for a quote arrives by email, until the reply leaves for the customer.”

If it cannot be stated in one sentence with a beginning and an end, it is not one process — it is several, tangled together, and automating them as one never works.

The four things to record

1. How often it happens

Per week or per month. The real number, counted, not estimated. This is where the biggest surprises appear: processes that “eat all our time” turn out to happen four times a month, while others dismissed as minor happen fifty times.

2. How long one instance takes

Timed, not guessed. Include the interruptions: time spent waiting for an answer is part of the process duration, even if it is not actual work.

3. How many hands it passes through

Every handover from one person to another is a place where time is lost and errors appear. A process with four handovers is much harder to automate than one with a single handover — but also much more worth simplifying.

4. How often it departs from the usual path

The most important and the most frequently ignored. If nine out of ten cases follow the same route, automation makes sense. If every case is different, what you actually have is judgement work disguised as repetitive work — and that does not automate.

What the numbers show

After two weeks you have a picture. Three typical situations:

• High volume, stable route, few handovers. The ideal case. This is where automation produces a visible effect.
• High volume, but every case different. Do not automate yet. Simplify the rules first, then measure again.
• Low volume, however annoying. The build effort is not justified. It is often better solved with a template or a checklist.

What almost always gets discovered

Three things appear in nearly every measurement:

1. Part of the process is not necessary. A step done because “it has always been done”. It disappears with no technology at all.
2. The time goes into waiting, not into work. If the process takes three days but the actual work is twenty minutes, the problem is coordination, not speed.
3. The real process is not the described one. People have workarounds nobody mentions. Automating the official version would automate something that does not happen.

What to do with the result

If the numbers say it is worth it, you also have the baseline to check six months later whether it helped. If they say it is not, you have saved yourself an entire project — and that is a good result too.

If you want, we will send you the sheet we use for the two weeks of measuring. There is nothing special in it, but it saves you the question “what exactly do we record”.

← Back to the blog