Your team ships code every day. Small commits. Automated tests. A rollback button. Feedback in minutes.
Then one of your engineers gets promoted, and the company hands them a two-day workshop... nine years later.
I saw a line from Lee Woollsey recently. "Your leadership training is too slow, too rigid, too...you." He is right on the first two. The third one is where it gets interesting, and where most of the fixes go wrong.
The nine-year gap
Jack Zenger and Joe Folkman studied 17,000 leaders going through development programs. The average age of a participant was 42. Fewer than 5 percent were under 27. Only 10 percent were under 30. Their conclusion, in the title of the Zenger Folkman study, was "too little, too late."
Most people take their first team somewhere around 30.
So the typical engineering manager runs a team for close to a decade before anyone teaches them the job. They learn on live traffic. No staging environment. No rollback. The blast radius is other people's careers.
We would fire an architect who shipped like this.
The money is already there
US organizations spent $102.8 billion on training in 2025, up nearly 5 percent, according to the 2025 Training Industry Report. An average of $874 per learner, across 40 hours a year.
Look at the split by company size. Small companies spend $1,091 per learner. Large corporations spend $468. The bigger the org, the thinner the spend per head... and the larger the number of people reporting to each untrained manager.
Nobody is starving this problem of funding. The delivery model wastes the funding it gets.
Too slow, too rigid... and the evidence backs it
I expected to argue this from opinion. Then I found the data.
Lacerenza and colleagues published a meta-analysis of leadership training in the Journal of Applied Psychology in 2017, pulling together 335 independent samples. Their headline finding surprised the field. Leadership training works better than people assumed. Reactions δ = .63, learning δ = .73, transfer to the job δ = .82, organizational results δ = .72. The average program lifts a leader from the 50th percentile to the 78th.
The moderators are where the real story sits:
- Spaced sessions beat one big block. Transfer of .88 spaced against .71 massed.
- Practice beats lecture. Role plays, action learning and on-the-job assignments produced the strongest transfer of any single delivery method at .87.
- Needs analysis first. Transfer of .91 with it, .72 without.
- Feedback improves on-the-job behavior, whether from one person or from a 360.
- On-site beats off-site for organizational and team outcomes.
Read the list again. Spacing. Practice. Requirements gathering. Feedback loops. Deploy where the work happens.
You have seen this list before. It is continuous delivery. Software teams have run on these principles for fifteen years and call them obvious.
Where the usual fix breaks
The standard prescription for slow, rigid training goes like this: make it accessible, asynchronous and adaptive. Put it on video. Let people self-serve. Buy an LMS.
Two out of three.
The same meta-analysis reports face-to-face delivery producing transfer of .86. Self-administered training produced .64. Around a quarter of the effect walks out of the door with the instructor.
Async is not the win. Async is what you reach for when scheduling is hard. It makes training cheap and available, and it strips out the two ingredients doing most of the work: someone watching you do the thing, and someone telling you what they saw.
Self-serve video is documentation. Documentation earns its keep. Nobody claims reading the docs turns you into a senior engineer.
What untrained managers cost you
Our own research turned up something blunt. 99.5% of survey respondents said they had worked for one or more types of bad boss. Not a minority experience. Close to everyone.
Step It Up HR leads with the Gallup number alongside it: 21% of employees worldwide are engaged at work.
Put those next to your attrition dashboard. Your senior engineers do not leave over the tech stack. They leave over the person running their week. Every regretted resignation costs you months of ramp, a chunk of tribal knowledge and a recruiter fee. Compare it against $468 per learner and the business case writes itself.
Build the pipeline you already know how to build
If you run engineering and you own manager development, stop shopping for courses. Start building a pipeline. The mental model is sitting in your head already.
Small batches. One 90-minute session a month for a year, instead of a two-day offsite. Spacing carried a .17 lift in transfer in the meta-analysis. Structurally free. You are reordering time you already spend.
Staging environments. Nobody rehearses a layoff conversation on a live human. Pair two managers, hand them a real scenario from last quarter, put a third person in the room to observe. One hour. This is the practice effect, the strongest delivery method in the data.
Code review for behavior. A new manager writes up how they plan to handle an underperformer. A senior manager reads it before the conversation and debriefs after. Feedback moves on-the-job behavior. Feedback nobody gives moves nothing.
Requirements before build. Needs analysis moved transfer from .72 to .91. In engineering terms: stop shipping features nobody asked for. Ask your managers what breaks in their week, then build for those failures.
Run it on-prem. On-site training beat off-site. Your managers learn best in the building where the problems live.
Ship to production early. Train people at 30, when they take their first team. Not at 42, when the habits have set and three cohorts of engineers have already resigned.
The "too you" part
Lee's third jab lands hardest. Most leadership training reflects whoever signed the purchase order. It teaches the style of the person at the top, or the consultant's framework, or the book the CEO read on a plane.
Engineers smell this instantly. Give a room of senior developers a model with four quadrants and no evidence behind it and you lose them inside ninety seconds. Then everyone walks out concluding engineers hate soft skills training.
Engineers do not hate soft skills. They hate unfalsifiable claims delivered with confidence. Bring them effect sizes. Tell them practice moves transfer further than lecture, and put the number on the slide. Watch the room turn around.
I'm not sure about this part: I have seen this shift happen in rooms I have run, and it matches what the transfer research implies about needs analysis and buy-in, but I have no study measuring engineer-specific resistance to leadership training. Treat it as my experience rather than evidence.
Start Monday
Pick your three newest managers. Ask each one which part of the job they are guessing at. Take the most common answer. Build one 90-minute session on it, in your own office, with 40 minutes of practice and someone giving feedback.
Repeat next month with the second most common answer.
You have shipped an MVP of a leadership pipeline for the cost of one afternoon. Iterate it the way you iterate on everything else you own.
Your deployment pipeline gets a retro every time it fails. What does your management pipeline get?