Most of what a manager of managers hears about their organization has already passed through someone else’s judgment about what was worth saying. A skip-level meeting is the deliberate exception: a conversation with someone two levels down, without their manager in the room. It runs alongside the reporting line underneath it rather than instead of it.
Done well, it is the fastest way to find out what is actually happening before it turns into a resignation or a surprise in a review cycle. Done badly, it is a status meeting with extra awkwardness. Here is how to run one, what to ask, what not to ask, and what to do with what you hear before the next one.
Who this guide is for: the person calling the meeting. A manager of managers, or a founder with a second layer underneath them. Everything below assumes you schedule the skip-level, open it, and have to decide what happens afterwards. If you are the report going into one rather than running it, this will still show you what a good skip-level looks like from the other side of the table.
What a skip-level meeting is — and what it isn’t
A skip-level is a conversation between a manager and someone who reports to one of their direct reports: two levels down the org chart, sometimes more. The name describes the structure, not the intent. You are not skipping the direct manager as a practice, you are stepping around them for one meeting, on purpose, with them knowing it is happening.
The easiest mistake is treating a skip-level like a smaller 1:1. It is not. A 1:1 is where a manager tracks a report’s work and growth over time, and their own manager already does that with them every week or two. A skip-level answers a narrower question: what does this person see from where they sit that has not made it up two levels of summarizing yet? Ask about their current sprint and you waste the one channel that exists for something else. You also spend their time on a second status meeting they never asked for.
Why run them: the two things you only learn two levels down
Start with the stakes. Gallup found that managers account for at least 70% of the variance in employee engagement scores across business units. That layer makes most of the difference between a steady team and a struggling one, and it is the layer you observe least directly.
What reaches you about that layer has usually passed a filter first: a manager who reasonably wants their own team to look competent. That filter is not dishonesty. It is ordinary judgment about what deserves your attention: the manager solves what they can solve, and passes up a version that is accurate but tidier. Good managers do it as readily as weak ones, which is what makes it hard to spot. People are reluctant to pass bad news upward, an effect Rosen and Tesser named the MUM effect in 1970 that has held up in the research since. The gap it leaves is wide: when Milliken, Morrison and Hewlin interviewed 40 employees across industries, 85% described at least one occasion when they felt unable to raise something to their boss that they themselves thought was important. Only half said they generally felt comfortable speaking up at work. A skip-level is the structural way around that.
Two kinds of information suffer most from that filter, and they are the two a skip-level recovers best. Take them one at a time. The first is what the work looks like where it actually happens: the manual step nobody has automated, the handoff that quietly breaks every third time, an on-call rotation that runs unfairly for weeks. From the team manager’s perspective everything looks good and on-call is running, but it is not running fair.
The second is how a report’s own manager is doing. When the same study asked which issues people could not raise upward, concern about a colleague’s or a supervisor’s competence topped the list at 37.5%, ahead of pay, policy and career worries. The subject most likely to stay unsaid is the one a skip-level exists to reach. How you ask matters more here than anywhere else in the meeting.
Before the meeting: what to prepare
Light prep beats heavy prep. You are not building a case; a report who watches you flip through pages of notes about their work stops treating this as a conversation. Skim whatever you have from the last conversation, mostly so you do not ask about a project that already shipped. In Tagup that is the person’s profile timeline; in a notebook it is turning back a page.
Bring three or four genuine threads rather than fifteen questions, and hold them in reserve. The report should drive most of this conversation, not you. Open by asking what is on their mind and spend the bulk of the time there; your threads fill the gaps when they run out. If you want a structure rather than a blank page, here is a copyable agenda and a longer question bank →. Adjust it per person.
Questions that actually get answers
The best skip-level questions are specific enough to answer honestly without putting the person on record about anything.
About the team
- “What’s something the team deals with regularly that never reaches me?”
- “If you could fix one process tomorrow with no approval needed, what would it be?”
- “Who’s doing work right now that isn’t getting noticed?”
About their manager — lead with something positive-framed first. You are not asking them to grade their boss, and a question that sounds like you are gets a guarded, useless answer.
- “What does [manager] do that makes your work easier?”
- “Is there anything worth [manager] knowing that hasn’t reached them yet?”
About the work
- “What’s taking longer than it reasonably should, and why?”
- “What would you stop doing tomorrow if you were sure nobody would notice?”
About the person
- “What are you proud of recently that didn’t get much recognition?”
- “Where do you want to be in a year, and is anything here in the way of that?”
Questions that get you nothing, and why
Each of these is common, and each fails for a reason worth knowing.
“How’s it going?” is too open to answer honestly on the spot. It gets “fine,” and “fine” tells you nothing you did not already assume.
“Any concerns about your manager?” is the single most common skip-level question, and it is close to useless. It asks someone to criticize the person who controls their next review, on the record, to that person’s boss. Anything short of glowing feels like informing. The answer comes back guarded, and you can rarely act well on the secondhand information underneath it.
“Do you feel the team is well managed?” is the same trap in different words, a leading question with an implied correct answer.
Any yes/no question about sensitive territory. “Are you happy here?” gets “yes.” “What would make you consider leaving” covers the same ground, and nobody can answer it in one syllable.
How to open — and what to say about confidentiality
Say plainly why you are there before you ask anything. This is not an evaluation of them or a review of their manager. You want to hear how things look from where they sit, because that view does not always travel up two levels intact.
Then be honest about confidentiality, and specific rather than absolute. The reasons people gave for staying silent in the Milliken study were almost all relational: fear of being labeled a troublemaker (30%), fear of damaging a valued relationship (27.5%), fear of retaliation (22.5%). Your promise about confidentiality has to answer those fears. Break it once and you prove every one of them right.
“Nothing you say leaves this room” is a promise you usually cannot keep: if someone tells you something that needs fixing, you are going to act on it. Make the narrower promise instead. You will not repeat what they said back to their manager with their name attached, and if something needs to change, you will change it in a way that does not point back to this conversation. Say that out loud. It is smaller, and you can keep it.
If recording comes up, treat it like any conversation with someone you do not manage directly: ask first, and do not record until the person says yes. More on how consent for recording actually works →.
What you do with what you hear — the part everyone skips
Most skip-levels do not fail in the meeting. They fail afterwards. You ask well, nod along, feel like a thoughtful leader for forty-five minutes, and then let all of it evaporate before the next 1:1 with your own directs. Do that a few times and the reports notice. The next conversation gets guarded answers, because nothing came of the last one.
The research is direct about the cost. A quarter of the people in the Milliken study named futility as their reason for silence: speaking up would not make a difference, or the person hearing it would not act. Nobody starts out believing that. They learn it from skip-levels that changed nothing.
What you hear splits into two kinds of things, and both should be handled differently.
The first is a specific commitment: “I’ll ask about that,” “let’s check back in a month.” A promise with no note behind it rests on your memory across a quarter of other meetings, so write it down before the next thing on your calendar starts, with a date attached. In a notebook that is a dated line under the person’s name, read back before the next conversation. In Tagup, log the skip-level under that person’s profile → like any other conversation with them, and turn what you promised into a tracked follow-up → with a real due date rather than a mental note that quietly expires. Either way, you open the next conversation with what you said you would do.
The second is harder: themes visible only in aggregate, across several skip-levels over time. One person who calls a process frustrating is a data point. Three people who raise the same friction unprompted, in three separate conversations, is a pattern. Memory alone will not catch it, because those arrive as unrelated meetings spread across a quarter. Keyword search does not catch it either, since all three people will have described the same problem in different words.
That gap is what the Conversation Hub closes. Ask it across people rather than about one person: “what has come up more than once about the release process this quarter?” It reads every tagup, note and transcript by meaning, so “the deploy is a mess” and “we lose a day every release” come back as one thread instead of two unrelated notes. It surfaces the candidates; you still decide which ones are a pattern worth acting on.
Whatever you heard, one move is always wrong: handing the raw list to the person’s manager. “Two of your reports said the same thing about you” points straight back at the room even with the names stripped out, and it breaks the narrower promise you made at the start. Worse, it turns the report into the source of a criticism they thought they were giving you in confidence. What to do instead takes more care, and it is the next section.
Following up without going around the manager
Silence is not the main risk here. Undermining the manager whose team you just talked to is. Swoop in and personally fix something a report mentioned, and you have told that manager their authority is conditional, and told the report that going around their manager works. Neither habit is worth encouraging.
Sort what you heard into two piles. Systemic problems, like a broken process or a tooling gap, you fix yourself, and you fix them without naming who raised them. Anything about a specific manager’s behavior goes through that manager, privately, rather than around them. If a report flags an on-call rotation as unfair, do not reassign it yourself. Take it to the team’s manager, ask what they have already noticed, and let them own the fix. The team then sees their own manager fix it, not you. That keeps the manager’s authority intact, and the change sticks.
The other half is going back to the report. You do not owe them the detail, only the fact that you acted, and “I looked into what you mentioned” is usually enough. If you never close that loop, they are left to assume nothing happened, which is exactly the futility that keeps people quiet in the first place. The next skip-level then turns back into a status meeting.
Cadence: how often, with whom, how long
Thirty minutes is usually enough. An hour turns into a status meeting by minute twenty.
How often depends on how many people sit below your direct reports. Count them once: six managers with six people each puts thirty-six names on your list, and nobody runs thirty-six skip-levels a month. A quarterly rotation, a few per month, keeps anyone from going more than a quarter without one. Smaller organizations can go monthly. Picking the wrong interval is not the real risk. Picking one you cannot keep is. A cadence that quietly lapses reads to your reports as a loss of interest, not as a busy calendar.
Five mistakes worth avoiding
Turning it into a status meeting. Ask about sprint progress and you have picked the wrong channel for the wrong purpose.
Fishing for dirt on the direct manager. It puts the report in a bad spot and gives you information you are usually not equipped to act on well.
Promising confidentiality you cannot actually keep. Say the narrower, true thing instead: no names attached, but you will act on what needs fixing.
Hearing something real and doing nothing with it. This is the one that kills the practice over time, and it is the one the research names most clearly.
Running it once and letting it lapse. One skip-level followed by silence looks like a gesture rather than a practice, and your reports will read it that way. Choose the slowest cadence you can genuinely keep, and then keep it.
Skip-level meetings: quick answers
What is a skip-level meeting?
A conversation between a manager and someone two levels down the org chart, typically a direct report’s direct report, without that person’s manager in the room. It surfaces what does not reach a manager through the usual reporting line; it does not replace the 1:1 relationship that already exists one level up.
How often should you hold skip-levels?
Often enough that nobody goes more than a quarter without one, and no more often than you can sustain. For a manager of managers with thirty or more people two levels down, a rotating quarterly cadence is usually more realistic than monthly.
What should you not ask in a skip-level?
Avoid leading questions about the person’s own manager, like “any concerns about your manager?” — they ask someone to criticize the person who controls their next review, and usually get a guarded non-answer. Overly open questions like “how’s it going?” have the same problem from a different angle: too easy to answer with nothing.
Should the direct manager know what was said?
Not by name and not verbatim. If something specific needs fixing, fix it without pointing back to who raised it; if it is about the manager’s own behavior, raise that with the manager privately rather than acting around them.
Should you record a skip-level meeting?
Usually, no. Most of the value is in a candid, off-the-record feeling conversation, and a recording changes what people are willing to say. If you do record one, the honest approach and the required approach are the same: ask first, and do not record until the person has agreed.
What’s the difference between a skip-level and a 1:1?
A 1:1 is the regular conversation between a manager and their own report, about that person’s work and growth. A skip-level happens less often, with someone a level further down, and it exists to catch what the reporting line leaves out. Treat it like a smaller 1:1 and you waste it.
None of this requires much infrastructure; a notebook and a decent memory get you most of the way there. Memory across a quarter of other meetings is the part that breaks, and that is the part Tagup holds for you. If you manage managers, see how that fits your day-to-day →, or everything the product does →.