Where the time goes

Frequency tells you what usually happens. Duration tells you what it costs. The second is normally where the value of an analysis sits.

Switch the model from frequency to performance

With a model selected, switch the view from frequency to performance. Activities and arrows are now colored and labeled by how long they take rather than how often they occur.

The important part is what the arrows now mean. An arrow’s duration is waiting time — the gap between one activity finishing and the next starting. In most processes, most elapsed time is waiting, not working. A step that takes four minutes of effort but sits in a queue for three days is a four-minute activity with a three-day arrow, and the arrow is the problem.

Use the median, not the mean

Set the aggregation to median duration first. A handful of pathological cases — one order stuck for eight months waiting on a customer — will drag a mean far above anything typical, and you will end up investigating a step that is fine for 99% of cases.

Then look at the mean as well, and treat a large gap between the two as its own finding: it means the step is usually fast and occasionally catastrophic, which is a different problem from being uniformly slow, and it has a different fix.

Find the batching

Run Dotted Chart on the log. Every event becomes a dot, positioned by time. Patterns that are invisible in a process model are obvious here:

• Vertical stripes — many cases hitting the same activity at the same moment. That is a batch job or a weekly meeting, not a process step. Everything arriving before it waits for it.
• Empty bands — nights, weekends, holidays. A case that finishes at 17:05 on a Friday does not resume until Monday, and no amount of process redesign changes that.
• A widening or shifting cloud — the process changed at some point during the period covered.

Batching is worth hunting specifically, because it is one of the few findings where the fix is scheduling rather than redesign, and scheduling is cheap.

Turn a finding into a sentence

An analysis is finished when you can state it plainly:

Orders over €10,000 wait a median of 3.2 days between Approve and Ship, against 4 hours for everything else. The gap is a manual second approval that runs once daily at 16:00.

That is specific, quantified, and points at something changeable. “The process is inefficient” is not.

Next

Bring your own data and do this on a process you care about.