Last updated: August 5, 2026
By Kris Drouet, Engineering Executive, in partnership with KORE1
Coaching a new engineering manager works when you name the specific behavior, hand back the decision they are avoiding, and set a check-in date. A postmortem explains why someone failed. A coaching conversation happens early enough to change the outcome. The gap between those two meetings is usually about five weeks, and most directors miss it.
She said the thing I have now heard something like forty times, in almost the same words, and it always shows up around week six.
“I don’t know what I did today.”
She had been an outstanding staff engineer. Payments platform, high volume. The person you wanted holding the pager during a rate move. Six weeks earlier I had signed off on promoting her. Now she was sitting across from me describing a day that contained eleven meetings, four Slack threads she had refereed, one architecture doc she had read twice and commented on zero times, and no output she could point at. She was not complaining. Complaining would have been fine. She was reporting it flatly, the way you report a build failure you have not diagnosed yet, and flat is the tone I have learned to worry about.
I have written before about promoting your best IC to engineering manager and about how long the transition actually takes before someone finds footing or washes out. This piece is the other half. It is the meeting itself. What you say, in what order, on the day you first notice the wobble.
Because you will notice it. And then most of you will wait. The waiting is the most expensive habit I see in directors who are otherwise excellent at their jobs, and I see it constantly, including on searches where I end up working next to KORE1’s engineering staffing agency recruiters after the window has already closed.
A Coaching Conversation and a Postmortem Are Not the Same Meeting
A coaching conversation is a forward-looking meeting held while the outcome is still changeable, where a leader names one specific behavior, transfers ownership of a decision back to the person, and sets a date to check on it. A postmortem is retrospective. It explains a result that has already landed.
Years ago I worked for someone who reviewed every pull request personally and had opinions about variable naming. Brilliant engineer. Three strong people left inside eighteen months because they were executing somebody else’s vision all day. When I started leading teams I made a deliberate call to go the other way. Set the direction, hand over the context, then get out of the road and stay out of it even when staying out of it is uncomfortable. If someone decides differently than I would have, I want their reasoning before I question the result. Nine times out of ten the reasoning holds up. I learn something.
The tenth time becomes a coaching conversation, not a postmortem.
That line is not a mood. It is a scheduling instruction. The tenth time is a specific, identifiable moment, and it is early. It is the first decision that went sideways, not the fourth. Hold the meeting there and you are adjusting a trajectory, cheaply. Hold it after the team has quietly routed around your new manager, after two engineers have started bringing their questions straight to you, after the skip-level has turned into a support group, and you are not coaching anybody anymore. You are documenting. The checkpoint-by-checkpoint version of that window sits on KORE1’s engineering leadership transition coaching page.
Six Weeks In, the Signals Are Quiet
Nobody escalates a wobble. Which is precisely why it costs so much, because every reporting line and every dashboard you have built is tuned to surface failure, and a wobble is not failure yet.
Wash-outs get escalated. Somebody files a complaint, a project misses by a quarter, an engineer resigns and names the manager in the exit interview. By then you have a personnel problem with an audit trail attached to it, and every option in front of you is expensive. The wobble, meanwhile, the part you can still actually fix, produces almost no signal. A slightly quieter standup. A manager still writing code at 11 p.m. because that is the only stretch of the day that felt like competence. I have been that manager.
Here is what I actually watch for, and what each one usually means.
| What you observe | Wobble (coach it) | Wash-out (different meeting) |
|---|---|---|
| They are still coding | Nights and weekends, on non-critical path work, because it feels like progress | They took the critical path story for themselves and the team is blocked behind them |
| Their 1:1s | Status readouts. Twenty minutes, all Jira, no person in the conversation | Cancelled two weeks running, and nobody on the team asked why |
| Decisions | They bring you calls that are clearly theirs, framed as “wanted to loop you in” | Their engineers have stopped bringing them calls at all |
| Conflict on the team | Avoided, deferred, or solved by the manager doing the work themselves | Two engineers have stopped speaking and it has been three sprints |
| How they talk about the team | “I need to get them to…” Ownership language, wrong direction | “They won’t…” Blame language, settled tone |
| Where your name shows up | On threads they could have closed themselves | On threads their engineers started, without them on it |
That last row is the one I would tattoo on a director’s forearm if they let me. When an engineer skips their manager to get to you and does not even think of it as skipping, the org has already re-routed. You can still fix it. It is just no longer a coaching conversation. It is a rebuild, and rebuilds take a quarter you did not budget.
The stress data backs up how invisible this is from the outside. In Gallup’s State of the Global Workplace, managers report higher daily stress than individual contributors, 45 percent versus 39 percent, and 57 percent of managers say it is a good time to be looking for a job, against 49 percent of ICs. Your wobbling new manager is stressed and quietly browsing. Both at once. Neither shows up in a status report.
Run the Conversation in This Order
Order matters more than wording. Considerably more. I have watched directors say every right thing in the wrong sequence and produce a manager who walked out of the room quietly certain they were being managed out, which is the opposite of the meeting they thought they just held.
Book it as a normal 1:1 slot. Not a new invite with a vague title. A calendar event called “quick sync” from your VP is its own kind of message, and the twenty minutes they spend spiraling beforehand is twenty minutes of coaching you have already spent.
Open with the specific thing you saw. One thing. Named, dated, observable. “Last Thursday the schema change to the ledger topic went out and I found out from the incident channel.” Not “I have been hearing some concerns.” Vague openers make people defend everything, and a person defending everything cannot hear anything. So be specific.
Then stop talking. This is the part that gets skipped. Ask what happened and let the silence sit, even when it gets uncomfortable, even when you are pretty sure you already know the answer, because the reasoning they give you is the entire diagnostic and you cannot get it if you fill the gap yourself. Nine times out of ten I learn something here. Sometimes the call was defensible and I had bad information.
Now separate the two things that are actually tangled together. Almost every wobble is one of these, and they need opposite treatment.
- They did not know the decision was theirs. That is a clarity problem, and it is yours, not theirs. Fix it in the room, out loud, with a specific boundary. “Rollbacks under a five percent error rate are your call, and you do not need to find me first.”
- They knew it was theirs and did not want it. Different animal entirely. That is a confidence gap, or an avoidance pattern, and telling them the boundary again will not touch it.
Get that wrong and you will spend a month solving the wrong problem. I have done it twice, both times confidently. I once spent six weeks re-clarifying decision rights for a manager whose actual issue was that he was quietly terrified of one specific senior engineer on his team, a nine-year veteran who sighed audibly in design reviews and had a way of saying “sure” that ended conversations. No org chart was going to fix that. We named it out loud in about four minutes. Whole thing loosened. He stayed.
Hand the decision back before the meeting ends. Not advice about the decision. The decision. “The staffing call on the reconciliation project is yours. Tell me what you land on Friday, and I am not going to weigh in before then.” You have just made the next week a rep instead of a wait.
Set the date out loud. “Two weeks from Thursday, same slot, we talk about how that went.” Then write it down where they can see it. If you skip this step the conversation degrades into encouragement, and encouragement is not coaching. It is just being nice at somebody.
The Four Sentences That Carry Most of the Weight
I keep a short list. Not a script exactly. More like four moves I know work, because I have watched them work across mortgage tech, payments, and two very different company sizes.
“What made that the right call at the time?” It presumes competence and asks for reasoning in the same breath. People answer it honestly. They do not answer “why did you do that” honestly, because that one has a verdict baked in and everybody hears it.
“That one is yours, and I will back whatever you decide.” Say the second half. Out loud. Without it the first half is a trap, and they know it is a trap, and you have just taught them that delegated authority at your company comes with an asterisk they are expected to read.
“You are six weeks in. I did not know what I was doing at six weeks either.” True, deflating, and worth more to the relationship than any amount of structured feedback. Mine was worse than hers. I reorganized a team in month two for reasons I could not defend by month four. Nobody stopped me.
The fourth one closes every single one of these meetings for me, without exception, and I have never once had it come back empty. “What is the thing you have been avoiding telling me?” Then wait. The answer takes about eleven seconds to surface and it is almost always the actual meeting.
What Not to Say, Even Though You Mean Well
Opening with “how’s it going” is not a coaching conversation. It is a hallway greeting wearing a meeting’s clothes. They will say fine. You will both feel better walking out than you did walking in, nothing whatsoever will change, and in five weeks you will be running a much worse version of this conversation with an HR partner taking notes in the corner.
Do not lead with the team’s feedback. If three engineers told you something, and you open with “the team feels,” you have just told your new manager that their reports have a back channel to you and they are on the outside of it. Whatever they say next will be political. Use the observable thing you personally saw. You have one, or you would not be reading this far into a piece about coaching conversations on a Tuesday.
Do not offer to take something off their plate. It feels generous. It reads as a demotion, and worse than that, it installs the exact reflex you are trying to break, which is the belief that when things get hard the correct move is for a more senior person to absorb the work. That reflex is one of the things a senior IC has to unlearn to lead, and you can accidentally reinstall it in one well-meaning sentence.
Do not do it over Slack. I know. Distributed teams, calendars are full, and you can type the whole thing out more precisely than you will say it. Type it anyway. Then read it to them on a call. Written feedback with no face on it gets read fourteen times at midnight in the least generous possible voice.
The Part Everyone Skips
The conversation is maybe twenty percent of this. The rest is whether anything you both said survives contact with next Tuesday.
Here is the honest version of why most coaching fails at this stage. It is not that the advice was bad. It is that the director had one good meeting, felt the relief of having finally said the thing, and then went back to their own calendar. Three weeks later nothing had changed and the interpretation on the other side was that the feedback must not have been that important. It was important.
So the follow-through, and I mean the minimum viable version, not a development plan with a template.
Two weeks: one question about the specific decision you handed back. Did they make it? Did they make it on time, without checking? Whether it was the call you would have made is not the metric. The metric is whether it got made at their level. Nothing else counts.
Six weeks: ask two of their engineers who they go to when they get stuck, phrased as casually as you can manage inside your own 1:1s. You are not fishing for complaints. You are checking the routing. If both of them name their manager, the wobble is closing.
Twelve weeks: now you can look at output, and not one week sooner. Cycle time, sprint predictability, whatever metric your org already trusts. Judging a nine-week-old manager on team throughput is like judging a migration on week-one latency. You will get a number. The number will be about the migration.
Gallup’s research on manager development is unusually direct about the payoff here. Managers who get training focused on coaching and people development see up to 18 percent higher engagement on their teams and 20 to 28 percent lower turnover. Fewer than half of managers worldwide get any management training at all. So the bar you are clearing is genuinely low, which should be encouraging and is mostly just embarrassing for the industry.
Show me the data on a manager at twelve weeks and I can usually tell you whether the coaching landed. Show it to me at four and I can tell you nothing except that they are tired.
When It Really Is a Postmortem
Sometimes you are late. It happens. Pretending otherwise helps nobody, least of all the person who has spent four months suspecting that the org already made a decision about them and just has not said it out loud yet.
You are past coaching when the team has stopped bringing this person problems, when a peer manager has quietly started covering for them, or when you have run the same conversation three times and the third one felt exactly like the first. That repetition is the tell. Coaching that is working feels different each time, because the problem moves. Coaching that is not working feels like a rerun, right down to the nodding.
At that point the kind thing, the genuinely kind thing, is to stop coaching and start being honest about the fit. Sometimes that means back to a staff or principal track with the title and the comp intact, which is a real outcome and not a punishment, and about a third of the ones I have seen go this way end up happier and materially more valuable to the company inside a year. Sometimes it means the role needs somebody who has done it before. That is fine too. Neither of those is a failure of the person. Both of them are a failure of the transition plan, which was probably yours.
If the answer is an experienced hire, hire for the second thing on your list, not the first. Everybody screens for technical depth. Almost nobody screens for whether the candidate can sit in silence after asking a hard question, which happens to be the single behavior that separates leaders who develop people from leaders who just hand out assignments. KORE1’s engineering recruiters run direct hire searches for engineering leadership roles across 30-plus U.S. metros, and their 12-month retention rate on placements sits at 92 percent, which in this particular role is the only number I would care about. A manager who leaves at month ten costs you the transition twice.
Before You Book the Meeting
These come up almost every time I walk a director through this.
Am I jumping the gun at week three?
Week three is fine if you have a specific observable behavior. Week three is a mistake if all you have is a feeling. The trigger is not the calendar, it is whether you can name one dated thing you personally saw.
Directors tend to err four weeks late, not early. The fear is that raising it early will shake a new manager’s confidence. The opposite happens. Every time, in my experience. Early feedback on a small thing reads as attention rather than as evidence. The same feedback at month four reads as a case being built, because by then it usually is.
What if the real problem is that I promoted the wrong person?
Have the conversation anyway, and run it exactly the same. It doubles as the cheapest fit diagnostic you own.
Six weeks of watching somebody struggle tells you less than forty minutes of hearing them reason out loud. And promoting the wrong person is rarely the real story anyway. Usually the person was fine and the transition plan did not exist, which is a thing I have written about at some length because it keeps happening to people who should know better.
My new manager is more technical than me. Does that change how I coach them?
Not really, and the reason is that you are not coaching them on the technology. You are coaching them on decision ownership, on conflict, and on where their attention goes.
What it does change is your credibility on one narrow axis, so do not fake it there. If they are deeper in the stack than you are, say so out loud and move on. The awkwardness of admitting it lasts about four seconds. The awkwardness of pretending lasts the rest of your working relationship. Staying close enough to the build to not be fooled is different from being the best engineer in the room, and I have argued that distinction at length elsewhere.
Should HR be in the room?
Not for this one. The moment a third chair appears it stops being coaching and becomes process, and everybody in the building will have decoded that within a day.
Loop them in when you move toward a role change or a formal plan. Not before. Not one day before. A coaching conversation depends on the other person believing they can say the true thing out loud, and that belief does not survive a third chair.
How many of these does one new manager need?
Three or four across the first two quarters, spaced far enough apart that something can actually change between them.
More than that and you are not coaching, you are supervising, and the team will feel it before you do. One every two weeks turns a new manager into your direct report in the worst sense, and the engineers below them will read the frequency correctly and route around both of you. Space them. Let the reps happen.
The Meeting You Do Not Regret
I have never once regretted having this conversation too early. Not one time in 25 years.
I have regretted waiting on maybe six occasions I can still name, and in every single case the cost was not the manager. It was the two engineers who left before I got there, and the eighteen months it took to rebuild what that team had been. The Bureau of Labor Statistics puts growth for computer and information systems managers at 15 percent through 2034, roughly 55,600 openings a year. Read that as a supply forecast. The pool of people doing this job for the first time is about to get a lot deeper, and most of them will get the title on a Monday with no plan attached to it. Your ability to catch a wobble at week six matters more, not less.
So book the twenty minutes and name the one thing you actually saw, then hand the decision back before anybody stands up.
Then show up two weeks later. That is the part. That is the entire part, and it is the thing that separates directors who develop engineering leaders from the ones who just keep hiring them.
Sitting on one of these right now? Or on the receiving end, six weeks in, wondering why nobody has said anything? Connect with me on LinkedIn. I answer these. And if the honest answer turns out to be an experienced engineering leader rather than a coaching plan, KORE1’s recruiters average 15-plus years in engineering staffing and you can talk to their team about the search.
