Work through them with the person who actually does the job, not with the person who owns the process on paper. Forty minutes is usually enough. Write the answers down verbatim — the wording people use is half the finding.
Before you can fix a process you have to agree on what it is. Most disagreements about a process are really disagreements about where it starts.
Teams usually name the first step they own, not the first step that happens. The real trigger is often an email, a phone call or a spreadsheet row that nobody counts as part of the process.
Listen for: an answer that begins “well, by the time it reaches us…”
Count people and teams, not job titles. Four handovers in a five-step process is a coordination problem, not a work problem.
Listen for: roles nobody mentioned in the first telling — quality, finance, an assistant who chases things
Anything the next person has to ask for is a gap in the handover, and it costs the same amount of time on every single case.
Listen for: “we usually call them to check what they meant”
Cycle time is mostly waiting. The work itself is rarely the problem — the gaps between the work are.
Name the wait, not the step. “Waiting for a signature” and “waiting for someone to notice the email” look identical on a timeline and need completely different fixes.
Listen for: “that one can sit there for a week”
Almost every process has a single constraint. Improving anything else while it stays in place moves nothing.
Listen for: one name that comes up in three different answers
The spread matters more than the average. A process that runs in two days or in three weeks depending on luck cannot be promised to a customer.
Listen for: a confident average followed by “…but it can be much worse”
Manual work is easiest to spot and easiest to cost. This is usually where the first quick win lives.
Re-keying between a system, a spreadsheet and an email is the most common friction in mid-sized companies — and the one people have stopped noticing because they have always done it.
Listen for: the word “just”: “I just copy it across”
Rework tells you where the real quality gate should be. It is almost never at the step that failed; it is upstream, where the information was captured badly.
Listen for: “half of them come back incomplete”
“By hand” and “by interface” are very different processes drawn the same way. Write the transport method on every arrow.
Listen for: an export, an upload folder, or a monthly file someone sends
Exceptions, control and visibility. This is the part that never appears in the official process document.
If a third of cases are exceptions, the exception path is the process. It usually runs on one experienced person and is written down nowhere.
Listen for: “that depends” as an answer to a yes/no question
Visibility is a process property, not a reporting problem. If the answer is “I would have to ask around”, nobody can manage the flow — only react to complaints about it.
Listen for: a status that lives in someone's inbox or head
Some controls are required and must stay. Others were added after one bad incident years ago and now cost more than the incident did. Ask what would actually happen if each one were removed.
Listen for: “we've always done it since that time when…”
One row per piece of friction you found. Score impact (what it costs when it happens) and frequency (how often) from 1 to 5, and effort to fix from 1 to 5. Priority = impact × frequency − effort. Fix the top three; ignore the rest until they are done.
| Friction found | Where in the process | Impact 1–5 | Frequency 1–5 | Effort 1–5 | Priority |
|---|---|---|---|---|---|
People describe the process they are supposed to follow. These get them describing the one they actually follow.
Three processes, and the friction the twelve questions tend to expose in each.
Describe the process in plain words. Modilo asks these questions, maps the answer as BPMN and names the friction it finds.