🇸🇬 HireDeveloper.sg

I Ran 40 Skip Level Meetings in Singapore Before I Found the 1 Question That Makes Engineers Tell You the Truth

Two colleagues talking across a desk in a quiet office, illustrating a skip level conversation between an engineering leader and an engineer in Singapore
Bryan

Bryan

Delivery & Offshore Teams Expert · November 17, 2026 · 12 min read

TL;DR

  • •The problem: my first 20 skip levels produced almost nothing. I opened with how are things going, got a polite version of fine, and left with warm impressions and no facts.
  • •What changed it: one question about last week instead of about feelings. What took longer than it should have. Concrete, safe to answer, and it reaches the friction that general questions never touch.
  • •The 7 steps: state the purpose, pick a cadence you can hold for a year, brief the manager first every time, open on the work, take notes you would show the engineer, close the loop in 2 moves, then track 3 signals instead of attendance.

For about eighteen months I ran these meetings badly and did not know it. I would book thirty minutes with an engineer two levels below me, open with how are things going, receive a courteous and entirely unusable answer, and leave with the feeling of having done something useful. Twenty of those happened before I understood that the feeling was the whole output.

What follows is the version that works, built from the next twenty and the four years since. The change that mattered was small and almost embarrassing: it was the first question.

Step 1: Decide What the Meeting Is For, and Say It Out Loud

A calendar invitation from someone two levels up, with no stated purpose, is not a neutral event. In the absence of a reason, people supply one, and the reasons they supply are reorganisation, performance concern, or something happening to their manager. I have had an engineer arrive at one of these visibly braced, having spent two days assuming the worst.

So pick one purpose and keep it narrow. Mine is: understanding how work actually flows, as opposed to how the process documents say it flows. That purpose is honest, it is genuinely something only someone at my level can act on, and it does not trespass on the manager in between.

Then state it twice. Once in the invitation, in a sentence. Once in the first minute, along with the two things it is not: it is not about their performance, and nothing they say will reach their manager attached to their name. Saying the second part out loud is awkward and it is the price of the next twenty five minutes being worth anything.

Step 2: Pick a Cadence You Can Hold for a Year

Do the arithmetic before you announce anything, because this is where most programmes die. Thirty engineers on a quarterly rotation is one hundred and twenty conversations a year, between two and three every week with no gaps. That survives contact with a calm quarter and collapses the moment a release slips.

Team sizeQuarterlyTwice a yearWhat is realistic
10 engineers40 a year20 a yearQuarterly is comfortable
25 engineers100 a year50 a yearTwice a year, or quarterly for a rotating subset
50 engineers200 a year100 a yearTwice a year, and delegate a tier down

Choose the lower frequency. An abandoned rotation is worse than no rotation, because the first version teaches people that this was an initiative rather than a commitment, and the second attempt then has to overcome evidence rather than indifference. I have watched a well intentioned monthly programme stop in week nine and make the subject harder to raise for two years.

One practical note on scope: include the people who never ask for your time. The engineers who volunteer for these conversations are rarely the ones holding the information you are missing.

Step 3: Brief the Manager First, Every Single Time

Not the first time. Every time. This is the step people skip once they feel established, and it is the one that determines whether your managers support the programme or quietly undermine it.

Before the invitation goes out, tell the manager three things: who you are speaking to, why, and what you will do with what you hear. Thirty seconds in a message. What a manager fears is not the conversation itself but finding out about it afterwards and not knowing what was said about them, and that fear is entirely reasonable.

Afterwards, close it from the other side: share the themes that touch their team, with nothing personal and no attribution, and where you want to change something in their area, decide it with them rather than around them. Handled this way, managers tend to become advocates, because the themes hand them evidence for arguments they were already making upward. Ours now ask us to bring conversations forward when they want a problem corroborated. Make it a standing rule rather than a judgement call, for the same reason a written error budget policy beats deciding case by case: the rule holds on the week you are busy and tempted to skip it.

The First Question Decides the MeetingSame engineer, same 30 minutes, two different openingsAsking about feelings“So, how are things going?”Requires a judgement about the whole job,delivered to someone 2 levels up,with no idea what you will do with it.“Good, yeah. Busy but good.”20 meetings, warm impressions, 0 facts.Asking about last week“What took longer than it should have?”Specific, recent, and about the systemrather than about any person.Safe to answer honestly on day one.“Waiting 3 days for staging access.”Named by 4 engineers. Fixed in a week.The second question is not cleverer. It is answerable without having to trust you yet.

Step 4: Open With a Question About the Work, Not About Feelings

Here is the whole shift. Stop asking how things are going. Ask instead: what took longer last week than it should have?

How are things going asks an engineer to deliver a summary judgement on their entire working life to a person two levels above them, without knowing what will be done with the answer. Fine is the only sensible response. It is not evasion, it is a correct read of the risk.

What took longer than it should have is a different kind of question. It is about last week, so it relies on memory rather than opinion. It is about the system rather than any person, so answering it does not require criticising a colleague. And it assumes something normal and true, which is that some things take too long everywhere, so it is safe to answer honestly in the first five minutes of a first meeting.

The first time I used it I got: three days waiting for staging environment access. Mundane, completely outside what I had been looking for, and said by four of the next six engineers I spoke to. That pattern had been invisible through every status report and every metrics dashboard, because it was nobody job to report it and it did not show up as a missed deadline. It took a week to fix.

Have two or three follow-ups ready and no script beyond them. What would you change if it were yours to change. What do you think we are trying to achieve this quarter, which is the question that most often reveals the gap between the strategy you believe you communicated and the one that arrived. And who outside your team do you depend on most, which is how dependency pain surfaces before it becomes an incident. Those also feed directly into blameless postmortems, where the same friction turns up later wearing a more expensive costume.

When the answer is that the team is simply short handed

Sometimes the friction is not process at all, it is two missing engineers. Tell us what the team actually needs to cover and we will shortlist pre-vetted developers against that scope.

Get started

Step 5: Take Notes You Would Be Willing to Show the Engineer

Write notes during the conversation, say that you are doing it, and say what the notes are for. Then hold yourself to a hard standard: nothing goes in that you would be uncomfortable showing the person sitting opposite you.

That rule does real work. It means you record staging access took three days, mentioned by several people rather than seems frustrated with the platform team. The first is a fact about the system you can act on. The second is a characterisation of a person, formed in thirty minutes, which will sit in a document for two years and shape decisions long after you have forgotten the conversation that produced it.

Offer to share what you wrote. Almost nobody takes it up, and the offer still changes the conversation, because the thing an engineer is calculating in the first ten minutes is what happens to their words after they leave the room. Answer that question before they have to ask it.

Keep two records, not one. A private per person note, which is only commitments you made, so you can check later whether you kept them. And a running themes document across all conversations, with no names, which is the part that is actually valuable and which you can show your managers without exposing anyone.

Step 6: Close the Loop in 2 Moves

Most skip level programmes fail here rather than in the room. The conversation goes well, the engineer raises something real, and then nothing visible happens, and they draw the only available conclusion.

Move one, before the meeting ends: say what you will do and what you will not do. Both halves. If staging access is fixable this month, say so and name the date. If the thing they raised is a deliberate decision that will not change, say that plainly and explain the reasoning. An engineer who is told no with a reason stays engaged; an engineer who is told nothing learns that raising things is a waste of their afternoon.

Move two, within a fortnight: report back. A short message naming the thing they raised and where it went. Including, especially, when the answer is that you looked at it and decided against. The message is the entire mechanism by which the next conversation is honest, and a programme that skips it produces steadily more polite and less useful meetings until someone concludes skip levels do not work.

Where a theme turns out to be a person rather than a process, handle it in your own one to one with that manager, not in the skip level, and never with enough detail to identify a source. If the theme is genuinely shared across several engineers, you can raise it as a pattern without exposing anyone. If it came from one person, you usually cannot, and the right move is to help them raise it directly. That boundary is also what keeps these conversations distinct from exit interviews, which arrive far too late to be your early warning system.

The Rotation, End to EndBrief the manager, ask about the work, then report back even when the answer is no1. State the purposein the invite and minute 12. Sustainable cadence1 to 2 per engineer a year3. Brief the managerevery time, before the invite4. Open on the workwhat took too long5. Showable notesthemes and commitments, not people6. Close the loop, 2 moveswill do and will not do, then report back7. Track 3 signalsrepeat themes, items closed, early warningSkip step 6 and the meetings continue while becoming steadily more polite and less true.The signal that it is working: problems start arriving before they are urgent.A dependency named 6 weeks early is the whole return on this, and it never shows up in attendance figures.

Step 7: Track 3 Signals Rather Than Attendance

Counting meetings held tells you the rotation ran, which is not the same as it working. Three signals are worth watching:

  1. Repeat themes across different engineers. The same friction named independently by three or four people is the most reliable output of the whole exercise, and it is the thing no dashboard produces. A theme raised once is a data point; raised four times, it is a decision waiting to be made.
  2. Raised items actually closed. Count what was raised and what was resolved or explicitly declined with a reason. A ratio near zero predicts the programme death about two quarters ahead of the meetings becoming empty.
  3. Lead time on bad news. Over a year, are problems reaching you earlier? A dependency flagged six weeks before it bites, rather than during the incident review, is the entire return on this investment.

The third is the one to judge the programme by, and it is uncomfortable because it cannot be put on a slide. You will know it by absence: the quarter where something did not become an emergency, and you can trace the reason back to a thirty minute conversation in which somebody mentioned, almost in passing, that a team they depend on had gone quiet.

The Mistake That Cost Me a Manager Relationship

Early on, an engineer told me something serious about how their manager was handling a project. I had promised the conversation was confidential, and I meant it when I said it. Then I acted on what I had heard, quickly, because it needed acting on.

The manager worked out where it had come from within two days. It was not difficult: only three people had the information, and I had changed a decision in the week after a skip level. I had broken something with both of them simultaneously. The engineer learned that my promise of confidentiality was worth nothing. The manager learned that I would make decisions about their team based on things they had not been told, which is the exact fear that makes managers resist these programmes.

What I should have said, and now always say, is narrower and less comforting: I will not attribute what you tell me, and if what I hear is serious enough that I have to act, I will tell you before I do. That promise is keepable. It also, counterintuitively, gets me more rather than less, because people can tell the difference between a commitment someone has thought about and a reassurance offered to move the conversation along. Do not promise absolute confidentiality in a skip level. You cannot deliver it, and the moment that becomes obvious is expensive.

Frequently Asked Questions

What is the difference between a skip level meeting and a one to one?
A one to one is between an engineer and their direct manager, and it is where individual work, growth and performance belong. A skip level is with their manager manager, and its subject is the system rather than the person: how work arrives, where it gets stuck, what the team believes the priorities are, and whether those beliefs match reality. Conflating them breaks both. If a skip level starts covering performance or progression, you have taken over the manager job and they will reasonably stop investing in their own one to ones. If it only produces pleasant general conversation, it is a social call with a calendar entry. The test before any skip level is whether what you want to learn is something only a system owner can act on. If the manager should handle it, the right move is to ask the manager.
How often should skip level meetings happen?
Often enough to be ordinary, rarely enough to sustain through a difficult quarter, which for most engineering leaders in Singapore means once or twice a year per engineer. Do the arithmetic before committing publicly: thirty engineers at quarterly is one hundred and twenty conversations a year, two or three every week, which most leaders cannot hold once delivery pressure arrives. Thirty at twice a year is sixty, which is manageable. Choose the lower frequency and keep it, because the failure mode is not an insufficient cadence but an abandoned one. A rotation that runs two months and stops is read as an initiative rather than a commitment, and it makes the next attempt harder because people now have evidence about what these programmes are worth.
What should I do if an engineer complains about their manager?
Separate a pattern from an incident, and do not promise confidentiality you cannot keep. For a single frustrating episode, ask whether they have raised it with the manager directly and what happened; usually they have not, and the useful contribution is helping them have that conversation rather than having it for them. For a pattern across several engineers, you have a management issue for your own one to one with that manager, handled without identifying sources, which is only possible if the theme is genuinely shared. Be honest when it arises: you cannot promise silence if what you are hearing is serious, but you will tell them before you act and will not attribute it. Promising absolute confidentiality and then acting on it is the fastest way to destroy the programme.
Do skip level meetings undermine the managers in between?
They do when they are a surprise, and they do not when the manager is briefed first and sees what came out. What a manager fears is not the conversation but learning about it afterwards from someone else, without knowing what was said about them. That fear is reasonable and the remedy is procedural. Tell the manager before the invitation goes out: who, why, and what you will do with what you hear. Afterwards, share the themes touching their team, excluding anything personal, and where a commitment falls in their area, make it with them rather than around them. Done this way managers tend to become advocates, because the themes give them evidence for arguments they were already trying to make upward.

Change one question this week

Ask your next three what took longer than it should have, and write down what repeats. If the repeated answer is that nobody has the bandwidth, tell us the scope and we will shortlist against it.

Get started