Your best engineer is now running a team of six. She ships. She reviews well. She knows the system better than anyone. And every one-on-one she runs takes eleven minutes, because she treats the conversation like a standup with fewer people.
Nobody taught her anything else. You promoted her and handed her a calendar.
Here is my position, and I know it will irritate people: coaching skills should be mandatory for every engineering manager in your org. Not optional. Not a nice-to-have on a development plan nobody reads. Mandatory, with a budget line and a deadline.
I already hear the pushback. Mandatory training is theatre. Engineers hate being made to sit in rooms. Half your managers will show up with a laptop open and their camera off. All of those objections have merit. I still think you should do it, and I want to explain why the resistance is the reason rather than the obstacle.
Your managers were never trained, and the numbers are ugly
Robert Walters surveyed 2,000 UK white-collar professionals and found 66% of managers had no formal training for the role. They call them accidental managers. Another 22% were quietly promoted with no acknowledgement, pay rise, or title change. Over eight in ten walked into leadership with no preparation.
The part of the survey people skip over: 35% of those bosses had repeatedly asked their employer for training. Asked. More than once. Nearly half of them then reported feeling overwhelmed and underequipped.
So the story your org tells itself, the one where managers resist development, falls apart on contact with the data. A third of them are begging for it.
Meanwhile Gallup's State of the Global Workplace 2026 reports global engagement fell to 20% in 2025, its lowest since 2020. Manager engagement dropped to 22%, down nine points since 2022. Low engagement cost the world economy around $10 trillion in lost productivity, roughly 9% of GDP.
Gallup also found something worth staring at: inside organizations with best-practice engagement work, 79% of managers were engaged. Nearly four times the global figure. The difference is not the people. It is what the org does with them.
My own research points the same way. When I surveyed people about their working lives, 99.5% said they had experienced one or more types of bad boss. Not 60%. Not 80%. Effectively everyone. We have built an industry where the bad boss is the default experience, and then we act surprised when retention numbers sag.
Coaching is a skill, not a personality trait
The most common objection I get from engineering leaders sounds reasonable. "Some people are naturally good with people. You cannot teach it."
The research disagrees. Jones, Woods and Guillaume ran a meta-analysis of workplace coaching across 17 studies in the Journal of Occupational and Organizational Psychology. Coaching showed positive effects on organizational outcomes overall at δ = 0.36, with skill-based outcomes at 0.28, affective outcomes at 0.51, and individual-level results at 1.24.
Two details from the moderator analysis matter for anyone building a program. Internal coaches produced stronger effects than external ones. And effect size did not vary by format, so face-to-face, blended, and remote all held up. The expensive external consultant with the branded workbook is not what makes the difference. Your own people, trained and given time, outperform them.
Gallup's own work puts 70% of the variance in team engagement down to the manager. Seventy percent. If a single component in your architecture accounted for 70% of your latency variance, you would not leave its configuration to individual preference. You would standardize it, measure it, and review it in the design doc.
Your management layer is architecture. Treat it like architecture.
Why optional training fails, every time
Here is the structural problem with voluntary programs, and it is the whole argument in one sentence.
Optional development selects for the people who need it least.
The managers who sign up for the coaching course are the ones already reading about coaching. They already run good one-on-ones. They already ask questions instead of issuing verdicts. You put on a course and it lands on the top quartile, who get slightly better, while the bottom quartile carries on grinding people down. Then someone runs the numbers, sees no movement in the engagement survey, and concludes training does not work.
Training worked fine. Your distribution was broken.
The manager whose team has the highest attrition in the department will not volunteer. He believes he is doing well. Nothing in his week tells him otherwise, because the people who would tell him have left, and exit interviews get sanitized before they reach him.
Mandatory removes the selection problem. Everyone goes, so nobody is singled out, and the manager who most needs the material is in the room without anyone having to make it a performance conversation first. Compulsory is kinder than targeted.
The resistance is a signal, not a verdict
When you announce it, you will get pushback. Read the pushback carefully, because the shape of it tells you things.
"I do not have time." Translation: nobody in the chain above him has said his people are part of the job. He is being measured on delivery alone and he is responding to the incentive he was given. Fix the incentive.
"My team is fine." Sometimes true. Ask him to show you the evidence. If he has no idea what his team would say about him, you have your finding.
"This is fluffy HR stuff." Show him the meta-analysis. Engineers respond to effect sizes. I have watched a skeptical staff engineer flip in one meeting because someone put a number on the thing he assumed was unmeasurable.
"We tried this before and nothing changed." Usually the most honest objection in the room, and usually correct. Ask him what happened. He will describe a one-day workshop with role-play, no follow-up, and no change to what leadership rewarded. He is not arguing against coaching. He is arguing against your last program.
How to make it stick
Mandatory does not mean a one-day workshop with a certificate. Here is what works, based on what I see in the field.
Run it in cohorts over weeks, not hours. Six sessions of ninety minutes beats one full day. Skills need reps and a gap between them for the reps to land.
Use real cases. No fictional scenarios. Every participant brings a live situation from their own team, anonymized if needed. Practice on the thing you have to do on Thursday.
Managers of managers go first. If the director skips the program, everyone below reads the signal correctly and treats it as low status. Start at the top, publicly, and let them be visibly bad at it in front of their reports. The modeling does more than the content.
Practice in pairs, in the room. Coaching is a motor skill. Reading about questions does not teach you to ask them while someone is crying about a stalled promotion.
Build it into the promotion gate. Nobody moves from senior engineer to manager without the program. Nobody moves manager to director without evidence of using it. Put it in the ladder document where people read it.
Measure the right things. Not attendance and not smile sheets. Ask team members whether their one-on-ones are useful, whether they got feedback in the last fortnight, and whether they know what good looks like in their role. Track it quarterly.
The uncomfortable part
Some of your managers should not be managers. A structured program will surface those people faster than three more years of hoping.
This is where most orgs lose their nerve. They run the training, they see clearly who is unwilling rather than unable, and then they do nothing, because moving someone out of management feels like a demotion and demotions are awkward. So the program becomes another thing done to people with no consequence attached, and the cynics who predicted it get to be right.
If you are not prepared to act on what the program reveals, do not run it. You will do more damage than doing nothing, because you will have raised the expectation and then confirmed the suspicion.
The alternative is worse. You leave 70% of the engagement variance in the hands of people you never trained, most of whom asked for help and did not get it, and you keep paying for it in attrition you attribute to the market.
I work on this with organizations through Step It Up HR, where the workshops run on a framework of Awareness, Acceptance and Action. The order matters. Awareness first, because most managers genuinely do not know how they land. Acceptance second, because knowing and admitting are different jobs. Action last, once the first two have done their work.
Your engineering org has a code review process, an on-call rota, and a deployment pipeline, all of them mandatory, all of them written down, none of them left to individual preference. Your management practice has none of it.
Which of those two systems does more damage when it fails?