Key takeaways
- →A skip-level meeting exists to give you information your direct reports cannot give you, not to check whether they are doing their job.
- →Say the confidentiality rule out loud in the first ninety seconds and describe exactly how you will pass things on without names attached.
- →Ask about systems and specifics - last week, this project, this process - rather than inviting people to rate their manager.
- →Thirty minutes one-to-one, or forty-five minutes for a group of six to ten; anything longer drifts into a status update.
- →Brief the manager in between before you start, not afterwards, or the whole exercise reads as going around them.
- →Close every cycle by telling people what changed, what did not, and why - a skip-level with no visible output is a listening tour nobody attends twice.
A skip-level meeting is a conversation between a leader and someone two or more levels below them, without the manager in between present. Its only real purpose is to give you information that cannot reach you through the reporting line - not because anyone is hiding it, but because every layer a message passes through smooths it, shortens it and rounds it toward the acceptable.
That purpose is easy to state and easy to lose. Run badly, a skip-level becomes a performance review by proxy, a complaint channel that bypasses a manager, or a thirty-minute status update the person has already given twice this week. Run well, it tells you which process everyone works around, which decision nobody understood, and which person is three weeks from resigning.
This guide covers the confidentiality contract you have to state out loud, the questions that produce specifics rather than reassurance, run sheets for both the one-to-one and the group format, and how to close the loop so people turn up to the next one.
What a skip-level is for, and what it is not
You are looking for three categories of information. First, friction that is invisible from above: the approval step that adds four days, the tool two teams both pay for, the handover that only works because one person does it manually every Friday. Second, comprehension: whether the strategy you announced in March means anything concrete to the people executing it. Third, early signals about people - who is stretched, who is bored, who has stopped putting their hand up.
It is not a place to evaluate the manager in between. If you go in with that question, you will get one of two answers, both useless: a defensive endorsement, or a complaint that arrives without the context you would need to act on it. Information about a manager will surface anyway, sideways, through the description of how work actually gets done - and that version is far more useful than a rating.
It is also not a substitute for the regular one-to-one. If someone's only meaningful conversation about their work is a twice-yearly skip-level, the problem is upstream and no amount of skilful listening in this session will fix it. Sustained attention from a direct manager is what drives engagement, and Gallup's work on the returns from engagement is a reasonable argument for fixing the manager relationship first rather than routing around it.
The confidentiality contract, stated out loud
People calibrate how honest to be in the first ninety seconds, and they calibrate on what you say about where their words are going. Vague reassurance - 'this is a safe space' - is worse than useless, because everyone has heard it before and it carries no information about your actual behaviour.
Be concrete instead. Say something like: I take notes for myself, I do not share them, and when a theme comes up more than once across the team I will raise it with your manager as a theme with no names attached. If something you tell me needs to be acted on with your name on it, I will ask you first and I will not do it if you say no. Then honour that exactly, including the awkward case where you disagree with their answer.
There is one carve-out to state at the same time, and stating it builds rather than damages trust: anything that involves someone's safety or a serious policy breach has to go somewhere, and you will tell them where before it goes. People are far more comfortable with a clearly bounded promise than with an unlimited one they suspect is not real. Psychological safety in a team is built from exactly these small, observed consistencies, not from declarations about culture.
Prepare the cycle, not the conversation
Most of the work happens before anyone sits down. Get these six things right and the conversation itself is mostly listening.
- ✓Manager in between briefed on the format and questions before any invite goes out
- ✓Whole layer scheduled on a rota, so no individual invitation reads as a signal
- ✓Three questions sent 48 hours ahead, framed as a starting point
- ✓Confidentiality rule written down so you say it the same way every time
- ✓Your last cycle's commitments reviewed, so you can open with what changed
- ○Thirty minutes held, camera on, no laptop between you and the notes
Questions that surface truth, and the ones that do not
The reliable pattern is to ask about a specific, recent, bounded thing. Broad questions - how are things going, are you happy here - produce broad answers, because answering them honestly requires the other person to decide unilaterally what you meant and how much risk to take. A narrow question does that work for them.
Ask what took longer last week than it should have. Ask which part of the current project they would cut if they had to lose a week. Ask what they had to ask someone else for that they should have been able to do themselves. Ask what they think we decided about the roadmap in March, and listen for the gap between that and what you thought you announced. Each of these gets you a concrete, checkable answer that opens into the systemic issue behind it.
Then ask the two questions worth asking every time. What is one thing you would change if you could change it tomorrow. And: what have you noticed that you assume I already know about. The second is the highest-yield question in the format, because most of what you are missing is not being withheld - it is simply assumed to have already reached you.
The 30-minute one-to-one run sheet
0-3 min: the contract and the reason
State the confidentiality rule in your standard wording. Say why they are on the list - because you are working through the whole team, not because of anything specific. If this is a second cycle, spend sixty seconds on what changed since last time, including the thing you did not manage to fix and why.
3-12 min: one specific, recent thing
Open with the narrowest of your three questions and stay on it. Resist the urge to move on when the first answer arrives - the useful material is usually two follow-ups deep. Ask how long it took, who else was involved, whether it happens every time. You are trying to get from an anecdote to a mechanism.
12-22 min: their agenda
Hand over explicitly: what did you want to talk about that I have not asked. Then stop talking. A silence of eight or ten seconds here is normal and productive; filling it with a second question is the most common way leaders talk themselves out of the best part of the meeting.
22-27 min: the two standing questions
What would you change tomorrow, and what do you assume I already know. Write the answers down in front of them - visibly taking a note signals that something will happen to what they said, which matters more than the note itself.
27-30 min: agree what leaves the room
Name the two or three things you will act on and say explicitly how each one will be described and to whom. Ask directly whether they are comfortable with that. If they are not, drop it - and be seen to drop it, because whether you honour a no is the thing the next person will hear about.
Choosing the format for the size of the layer
The format follows the number of people and the kind of information you need. Most leaders default to one-to-ones for everything and then quietly stop doing them when the calendar bites; a mixed cycle is more sustainable and covers more ground.
A practical pattern for a layer of thirty: an anonymous round to set the agenda, two group sessions of ten to work the themes, and six one-to-ones with people you have not spoken to in six months. That is roughly four hours of your time per quarter and it reaches everybody.
Running the group version without it going flat
The group skip-level fails in a predictable way: you ask an open question, two confident people answer, everyone else recalibrates to the tone those two set, and you leave with a smaller and more polished version of what you already believed. The fix is to collect before you discuss.
Open with an anonymous prompt everyone answers on their phone at the same time - one open text question such as what should we stop doing, and one rating on something concrete. Ninety seconds of writing, then put the responses on the shared screen and read them out. Session Flo is built for this shape: anonymous collection, a visible response count while people write, and a reveal you control. The room now has a set of raised topics that belong to nobody, which is a far easier place to start a conversation than an open floor.
Then work the top two or three themes rather than trying to cover everything. Ask who has an example, not who agrees. And expect the observation effect to be real: people behave differently when a senior leader is watching, so a group session will always give you a slightly tidier picture than the anonymous round did. Treat the written responses as the honest baseline and the discussion as the elaboration.
Four habits that separate a good skip-level from a polite one
Do not solve it in the room
Fixing a problem live feels responsive and quietly teaches people that the way to get something done is to reach you directly. Take it away, route it through the manager, come back with the outcome.
Count how much you talk
Aim to be speaking for under a quarter of the session. If you have explained context twice, you have turned a listening session into a briefing.
Write the theme, not the quote
Notes phrased as quotes are one forward from being a leak. Convert to a theme the same day, while you still remember enough to do it accurately.
Ask the last question anyway
When someone says everything is fine, ask what they would change if they had a free week. It reliably produces the thing they decided was not worth raising.
The cycle that makes it sustainable
A skip-level programme lives or dies on the last step. People will forgive you for not fixing something. They will not turn up again if they never find out whether you heard them.
Week 0 - brief the managers
Share the questions and the confidentiality rule with the managers in between. No surprises, no exceptions.
Week 1 - anonymous round
Two questions to the whole layer, open for three days. Read everything yourself before scheduling anything.
Weeks 2-5 - the conversations
Group sessions first, then one-to-ones with the people the themes point toward and anyone overdue.
Week 6 - themes to managers
Each manager gets the themes for their area, unnamed, plus what you have committed to on their behalf.
Week 7 - close the loop publicly
Tell the whole layer what you heard, what is changing, what is not and why. This is what earns the next cycle.
Closing the loop without breaking a confidence
Write the close-out as three lists. Here is what came up repeatedly. Here is what is changing, with who owns it and by when. Here is what is not changing, and the actual reason - which is sometimes that you disagree, and saying so plainly costs you far less credibility than silence does.
Keep every item at theme level. 'Several people described the release approval as adding about a week' is safe and actionable. 'Someone in the platform team said the approval process is a joke' identifies a person to anyone who was in the room. If a theme came from exactly one person and is important enough to raise, go back to that person first and agree the wording together.
Then do the smallest thing on the list quickly. One visible fix inside a fortnight does more for the credibility of the next cycle than three large initiatives announced with a timeline of six months. People judge whether the format is real by watching what happens to the easy items, because those are the ones where nothing but intent stands in the way.
Frequently asked questions
How often should you run skip-level meetings?
One-to-one skip-levels work well on a rolling six-month cycle: if you have four managers reporting to you covering thirty people, that is roughly one thirty-minute conversation a week and everyone gets seen twice a year. Group skip-levels can run quarterly. What matters more than frequency is predictability - people prepare differently for a scheduled rhythm than for an invitation that appears out of nowhere, and an unscheduled skip-level almost always reads as a signal that something is wrong.
What should you never do in a skip-level meeting?
Never ask someone to evaluate their manager directly, never make a promise you cannot keep in the room, and never repeat something you heard in a way that identifies the source. The first turns the conversation into a loyalty test, the second destroys the next three cycles when you fail to deliver, and the third ends the format permanently. One breach travels faster through a team than any amount of stated good intent, and you will not get an honest answer from that group again.
How do you stop a skip-level from undermining the manager?
Brief the manager before the cycle starts, tell them precisely what you will ask, and commit to what you will share back. Frame it as something you do with every team on a rota, not as a response to a concern. Then follow through: if a theme comes up that belongs to them, take it to them within a week as a theme rather than a quote, and let them own the fix. A skip-level undermines a manager when it is a surprise or when the leader starts solving problems that were never theirs to solve.
What if nobody says anything useful?
Silence in a skip-level is data, and it usually means one of three things: the person does not trust the confidentiality rule, they have never been asked before and have nothing prepared, or the questions were too broad. Fix the third one first - ask about a specific week or a specific project rather than how things are going generally. If it happens across a whole team, run an anonymous round before the conversations so people can raise themes without owning them, and use those themes as your opening questions.
Should a skip-level meeting have an agenda?
Send three questions in advance and nothing else. A full agenda makes it feel like a review; no agenda at all means the first ten minutes go on small talk and orientation. Three questions give people something to think about on the way in, which matters most for anyone who processes better with preparation time. Then be explicit that the questions are a starting point and they can spend the whole thirty minutes on something you did not ask about.
Keep reading
- running anonymous feedback without losing trust — How to handle the anonymous round that precedes a group skip-level.
- effective one-to-ones — The regular manager conversation a skip-level should never replace.
- getting quiet people to contribute — Techniques for the group format when two people dominate.
- pulse checks — A thirty-second way to open a group skip-level with real data.
- Session Flo — Anonymous collection and live themes for group listening sessions.