70%
of the variance in engagement between teams is attributable to the manager. Not the stack, not the comp band, not the roadmap.
Gallup, State of the American Manager ↗
For engineering managers
Built by a former VP of Engineering
You would never ship untested. So why are you testing leadership in production?
You were promoted for shipping, and nobody taught you the rest. Now the cost of one badly handled conversation is a staff engineer taking a recruiter call. Voohy is the staging environment: rehearse the conversation against a character who pushes back, get scored, then go and have it for real.
No credit card · 10 minutes · Async, any timezone
- Session
- 10 minutes, then a debrief
- Scored on
- Empathy, clarity, listening, resolution
- Library
- 26+ scenarios, engineering specific
- Teams
- Readiness scores from 5 seats
Session transcript
11 min 27 sec
DK
The Senior Who Is Checked Out
Dana is your strongest backend engineer. Reviews have gone terse, she skipped the last two design discussions, and you have a feeling.
Dana
Everything is fine. The tickets are moving. Was there something specific?
You
The tickets are moving, yes. What I noticed is that you have stopped arguing with me in design reviews, and you used to be the person who caught the thing everyone else missed. That change is what I want to understand.
↳Used a specific observed behavior rather than a vague read on her mood. Much harder to deflect.
Dana
Honestly? I stopped because it did not seem to matter. We shipped the version I flagged problems with anyway, and then spent a quarter fixing exactly what I said would break.
You
That is a fair thing to be angry about, and you were right. I want to talk about what would have to change for you to bother speaking up next time.
Debrief
Scored automatically
79
Solid session
Got to the real reason inside four turns
Empathy
81
Clarity
84
Active listening
86
Resolution focus
64
✓ You cited a specific behavior change instead of asking whether she was okay, which is why she gave you the real answer.
✎ You conceded the point but never committed to anything. Retention risk conversations need a concrete change with a date on it.
Illustrative session
The whole idea, in one line
Do not test leadership in production.
You have a staging environment for a config change. You have none at all for the conversation that decides whether your best engineer is still here in March.
01 / The gap
Promoted on technical signal. Measured on something else entirely.
The thing that got you here was legible: you shipped, you debugged the hard thing, you were right in design review. None of it predicts whether you can hold a performance conversation without damaging a person you need.
25.5%
of tech workers rate their manager as highly effective. 36.5% rate theirs ineffective. Manager quality was the strongest driver of burnout in the whole dataset.
Lenny’s Newsletter, 2026 survey ↗44%
of managers worldwide report ever receiving formal training for the role. Engineering is not the exception here. It may be the worst offender.
Gallup, State of the Global Workplace ↗0.5 – 2×
annual salary, the cost of replacing the engineer who leaves. You feel it in velocity months before it shows up in a budget line.
Gallup on turnover cost ↗01
Debugging people does not work
Your instinct is to find the root cause and fix it. People are not systems, they do not want to be fixed, and the diagnostic tone that works in an incident review lands as an interrogation in a one to one.
02
The technically right, socially expensive engineer
They are correct. They are also the reason two people have asked to move teams. Handling this badly costs you either the engineer or the team, and there is no rollback.
03
Judgment is now the job
AI made every deliverable arrive polished, so weak judgment stays invisible until something ships wrong. More weight lands on the manager as the judgment layer, and the promotion pipeline never adjusted.
02 / The loop
Pick the moment. Have the conversation. Get told the truth.
Twelve minutes end to end, async, no cohort to join. The same loop you already trust for code: change something, run it against a hostile input, read the output, change it again.
-
01
Pick the moment you are dreading
Scenarios written for engineering teams and opened on the sentence a real person would say. The rewrite argument, the disengaged staff engineer, the promotion you have to deny.
-
02
Have the conversation
Reply in your own words. The character holds a position, has a mood, and knows things they will not volunteer unless you make it safe. Push too hard and they shut down.
-
03
Debrief with a score
Overall score plus empathy, clarity, active listening and resolution focus, tied to specific turns in your transcript. Then run it again and beat it.
Why this format
Model the behavior, practice it, get feedback, repeat. A meta-analysis of 117 studies found the knowledge gained this way decayed over time, but the effects on skills and on actual job behavior held steady or kept improving. Transfer was strongest when people practiced scenarios drawn from their own work. The management book you read in January is the part that fades.
Taylor, Russ-Eft & Chan, A Meta-Analytic Review of Behavior Modeling Training, Journal of Applied Psychology, 2005 · Record
Put another way
Nobody gets better at incident response by reading about incident response. You run the game day. This is the game day for the conversation.
The Senior Who Is Checked Out
“It is nothing personal. I just got an offer I could not really say no to.”
Retention · Hard
The Brilliant Jerk
“My code review comments are accurate. If people find that harsh, that is not my problem.”
Performance · Hard
Promotion Denied
“So, did the promotion go through? I have been waiting to hear back.”
Feedback · Hard
Pushing Back on the VP
“I need your team to take this on too. I know you are stretched, but it is a priority.”
Negotiation · Hard
The Rewrite Argument
“The whole service needs rewriting. I am not going to keep patching this thing.”
Conflict · Medium
Missed Deadlines, Again
“I know my numbers are down. But I have been dealing with a lot outside of work.”
Feedback · Medium
Plus 20 more across performance, hiring, layoffs, peer conflict and managing up.
Try one free04 / For the managers you manage
If you run a group, you are already paying for this.
Just not on an invoice. You are paying in regretted attrition, in the team that keeps losing people, and in the coaching time you spend cleaning up conversations that went badly.
01
Reps before your seniors are the test subjects
Your managers rehearse against a simulator instead of practicing on the staff engineer you cannot afford to lose.
02
Readiness per manager, not attendance
A weighted score across skill, practice volume, scenario range, recency and consistency, capped by session count so it cannot be gamed with three lucky runs.
03
A skill heatmap for the whole group
When resolution focus is weak down the entire column, that is a group problem you can fix once rather than five separate coaching conversations.
04
Async across timezones
No workshop to schedule, no travel, no cohort start date. Managers run it when they have ten minutes.
05
Small seat counts, no procurement
Team plans start at five seats on one time passes. You can expense it without opening a vendor review.
What admins can and cannot see
✓ Readiness scores, session counts, competency trends, completion against assigned tracks, AI summaries.
× Conversation transcripts. Those stay private to the person who practiced. Nobody rehearses honestly while being recorded for review.
05 / Questions
What engineering leaders ask first.
Something not answered here? Write to the founder, who ran engineering teams before building this and has been on the wrong side of most of these conversations.
[email protected] →
I came up through the IC track. Is this pitched at my level?
+
It is built by a former VP of Engineering, so the scenarios are the ones that actually happen on engineering teams: the staff engineer who has quietly disengaged, the technically right person who is socially expensive, the rewrite argument, the VP who wants one more thing from a team already at capacity. Not generic corporate role play.
How is this different from role playing with ChatGPT?
+
The characters hold a position and have context they will not volunteer unless you make it safe to. They push back rather than agreeing with you. Every session is scored on the same four competencies, so progress is comparable across weeks instead of being a fresh chat each time, and on team plans the scores roll up into a manager readiness view.
I do not have time for another training program.
+
Ten minutes for the conversation, two for the debrief. Run it the night before a one to one. There is no cohort to join, no scheduled session and no travel, which is the entire reason it works for engineering managers across timezones.
Can I roll this out to the managers who report to me?
+
Yes. Team plans start at five seats. You invite by email or CSV, assign a scenario or a certification track to everyone at once with a due date, and see a readiness score and skill heatmap per manager. Conversation transcripts stay private to the person practicing.
What does it actually score?
+
Empathy, clarity, active listening and resolution focus, plus an overall session score out of 100 and written feedback tied to specific moments in your transcript. Scores accumulate into a skill profile and feed the certification tracks.
Does this replace coaching or a leadership program?
+
No. It is the practice layer between the framework and the real conversation. Programs teach the model, coaching works on you as a leader, and Voohy is where you get the reps so the model is available when someone is upset in front of you.
Your team deserves a manager
who has practiced.
Run one scenario. Ten minutes. See your score. That is usually enough to know whether this belongs in front of the managers who report to you.
No credit card · One scenario · Instant debrief