A conversation between a leader and somebody who does not report to them directly. Run properly it is the best correction available for filtered information. Run badly it is the fastest way to destroy a manager’s authority.
The difference is almost entirely in the rules, which have to be stated to everybody before the first one happens.
Set it up before you start
Tell the managers first. Explain what it is, why you are doing it, that everybody at that level rotates through, and that you will not be taking decisions in these meetings. A manager who learns their people are meeting the boss without warning will assume they are being investigated, and they will be partly right to.
Rotate on a fixed list. Everybody, in order, so that being invited carries no signal. Selective skip-levels are interpreted as either favour or investigation, and both readings distort the conversation you were trying to have.
State the two rules aloud in the meeting. This is not a performance review, and nothing you say goes back to your manager attributed to you. Then hold both without exception, because the first breach ends the programme for everybody.
The format
Thirty minutes. Their space rather than your office, if that is possible. No notes taken during it, which changes the register more than anything else you can do. Roughly every eight to ten weeks per person, which for most leaders means two or three a week.
The first three or four minutes will be careful whatever you do. This is not a problem to solve; it is the cost of entry, and it shortens with each subsequent conversation as the person accumulates evidence that the rules hold.
The questions
Six, in roughly this order. You will not get to all of them and should not try.
What are you actually working on this week? Opens with something concrete and frequently reveals a gap between what you believe people are doing and what they are doing.
What takes longer than it should? The most productive question in the set. Everybody has an answer, it costs nothing to give, and the answers point directly at process problems that never reach an executive.
What do you need that you do not have? Tools, information, authority, a decision from somebody.
If you ran this place, what would you change first? Invites a view of the whole rather than of their corner, and people give considerably more thoughtful answers than leaders expect.
What are people talking about that I am probably not hearing? Direct, and it works because it names the filter rather than pretending it does not exist.
What am I not asking about? Closes it, and is where the useful thing sometimes arrives once the person has relaxed.
What to avoid saying
Anything about their manager. Do not ask how it is going with them, do not ask what they think of a decision the manager made. If the person raises a genuine problem, hear it, do not comment on it, and deal with it afterward through the manager.
Any decision. Overriding or approving something in a skip-level bypasses the manager and teaches the organization that the route to a decision runs around them. Where somebody raises something that needs deciding, say that you will take it up properly.
Any promise. The temptation to fix something on the spot is considerable and the cost of an unmet promise here is higher than elsewhere, because the whole exercise runs on the person believing what you say.
Afterward
The part that determines whether the second round is worth holding.
Write down what you heard, that evening, for yourself. Not attributed. Patterns across eight conversations are the actual output, and no single conversation contains them.
Take anything actionable to the manager, unattributed and framed as your own observation. I have been hearing that the permit handoff is slow. This is the discipline that keeps the programme from becoming a back channel.
Report back to the group, not the individual. Every quarter or so, say what came out of these and what changed. Three items is enough. This is what converts the exercise from something people tolerate into something they prepare for.
How it fails
Nothing ever changes. Two rounds of this and people stop bringing anything. The programme now costs time and produces the appearance of consultation, which is worse than not doing it.
Something gets traced back. One instance of a comment reaching a manager attributed to the person who made it ends candour permanently, across the organization, within about a week.
It becomes a channel. When people start bringing you things instead of their manager, the manager has been hollowed out. The correction is to keep redirecting: that is a conversation for Dana, and I will let her know you are coming.
It lapses. Started with enthusiasm and abandoned after six weeks, this teaches something specific about which of your initiatives are real. Better to run four a quarter indefinitely than twenty in one month.
Scale
Below roughly twenty-five people this is unnecessary, because you already talk to everybody and adding a formal version introduces distance where none existed.
Above about two hundred you cannot rotate through everybody, and the honest version is a sample: a fixed number each quarter, drawn across functions and locations, with no claim that it is comprehensive.
In between is where this does the most work, which is also the range where a leader has just stopped being able to see the organization directly and has usually not noticed yet.
Edited by Patrick J. Wolf, PhD