How To Manage People

The actual work of managing people

Most managers treat supervision like a checklist. They hold weekly 1-on-1s, set OKRs, and assume everyone knows what good looks like. It doesn't work that way. The gap between expectation and execution is where management actually happens, and it is rarely a communication problem. It is usually a sequencing problem. I spent years running engineering teams, and the first thing I learned is that delegation without scaffolding is just abandonment with better branding. You tell someone what to build. They come back with something functional. You spend three weeks reworking it because you never explained the constraints that matter. By the time you fix it, you are behind and they feel micromanaged. Nobody wins. So you start by writing the invisible requirements before the work begins. Not a spec document, just a short message that says what decisions this person gets to make independently and which ones need your sign-off. Scope changes above a certain threshold. API contracts that affect other teams. Anything that touches the billing or audit path. Put it in writing. Keep it visible. This alone cuts rework by half on my teams, typically turning a two-week cleanup cycle into something closer to three days.

How To Manage People in Practice

People-management frameworks always sound cleaner on paper than they do on a Tuesday afternoon. The part that matters most is feedback cadence. Most managers give feedback once a quarter during reviews. That is too late. Real-time feedback works because it stays tied to the event that triggered it. You mention something the day it happens or within twenty-four hours. You do not collect offenses like points in a video game and then unload them during an annual review. There is a trap here that almost nobody warns about. Negative feedback delivered immediately can feel punitive if the relationship is not already strong. Positive feedback delivered immediately feels supportive. So I spend most of my relational budget early, giving small acknowledgments for normal competent work. Not inflated praise. Just straightforward recognition that the work landed. It builds a reserve you can draw on when correction is actually needed. I learned this the hard way after I gave blunt written feedback to a senior engineer who handled it fine for six months and then quietly disengaged. I had not established enough baseline trust. Took another nine months to fix it. Another thing that surprises people who read management books: context transfer beats instruction transfer. Telling someone the steps to solve a problem is faster in the moment. Teaching them how to think about the problem domain makes them faster later. When I onboarded someone onto a payment system with tight latency requirements, I did not hand them a troubleshooting flowchart. I walked through why the architecture was built that way, where the failure modes live, and what the trade-offs were. Two weeks later they were diagnosing issues without asking me. A checklist version would have kept them dependent on me forever.

When the system breaks down

Every management approach has a failure mode. The one that costs teams the most is over-optimizing for predictability. You structure work so tightly that people stop solving problems and start following processes. I saw this happen on a team where we had excellent pull request templates, mandatory design docs for anything above a certain complexity, and a scheduling system that left zero slack. Output was consistent. Innovation stopped. We shipped fewer high-value features because the friction to start one was higher than the friction to stay on maintenance work. The workaround is deliberate slack. Reserve about twenty percent of capacity for unplanned problem-solving. Do not fill it. Do not schedule it. Just leave it open and enforce the boundary yourself, which means saying no to meetings that would eat into it. It feels irresponsible at first. It is not. It is what lets good engineers stay productive instead of becoming bottlenecks for minor issues that could have been resolved quickly if they had the space. Remote management adds another layer. You lose the ambient awareness that comes from being in the same room. You cannot overhear a quick question that reveals a misunderstanding before it becomes a week of wasted work. The replacement is structured async updates. Not status reports written for visibility, but short written summaries of what was attempted, what was found, and what is blocking next steps. I required these on a Wednesday morning across every distributed team I ran. It took fifteen minutes to write and saved me hours of context-switching. People also got better at diagnosing their own issues because writing the summary forces you to articulate the problem clearly.

Specific edge cases

High-performer management is harder than average-performer management. Average performers respond well to clear expectations and steady feedback. Top performers often push against constraints and will openly challenge your process. The instinct is to let them run free. That works until they run free in the wrong direction and take others with them. I had a senior developer who wrote excellent code but refused to document anything and dismissed code reviews as a waste of time. Performance was outstanding individually. Team velocity dropped because everyone else was doing the documentation he would not. I tried the standard approach first: private conversation, clear expectations, a written agreement on review participation. It failed for four months. The workaround was to change the incentive structure. I made documentation a visible part of promotion criteria, not as a fluffy "soft skill" requirement but as a concrete deliverable tied to the same comp band as code quality. He complied within a quarter. Not because he changed his mind, but because the system finally rewarded the behavior he was ignoring. Another edge case is managing peers. You do not have authority over their workload or their performance reviews. What you do have is influence. The trick is to make your requests about shared outcomes instead of personal preference. "I need this by Thursday" gets pushed back. "If this misses Thursday, the launch date moves and marketing has to reschedule everything" gets action. Same request. Different framing. It is not manipulation. It is accuracy about how work actually connects.

What does not work

Honest assessment matters here. Micromanagement fails because it creates a single point of failure around the manager. Delegation without feedback fails because people drift without correction. Vague feedback fails because it is impossible to act on. All of these are common. All of these are preventable. The method that sounds good and rarely works is incentive stacking. Throw enough bonuses, titles, and public recognition at a team and expect performance to improve. It works short-term. It burns people out long-term. I watched a team hit quarterly targets through a combination of comp bumps and mandatory weekend work, then lose three key people the following spring. The remaining two operated at about sixty percent capacity for six months after. Compensation fixes retention symptoms. Culture and clarity fix the cause. Another thing to watch: meeting-heavy management. If your team spends more than twenty-five percent of their week in meetings, you are managing by ceremony instead of by outcome. I tracked this metric explicitly on every team. When it crept above that threshold, I cut meetings first, not features. The work did not suffer. It improved, because people had time to actually do it.

A practical framework that actually scales

Weekly 1-on-1s. Thirty minutes. Agenda set by the direct report, not the manager. Manager takes notes, follows up on blockers, and does not use the time for status updates that could have been an email. This one rule alone improves the quality of those conversations more than any coaching technique I have seen. Monthly team retrospectives. Not sprint retrospectives that focus only on process, but broader sessions where the team discusses what is working, what is not, and what they want to change. Rotate the facilitation. Let different people run it each month. It distributes ownership instead of concentrating it in one person. Quarterly career conversations separate from performance reviews. These are not about bonus numbers. They are about whether the person's work aligns with where they want to go. If they do not align, the conversation is about adjusting the work, not about adjusting the person. Most managers skip this entirely. It is the single highest-leverage practice I have found for retention.

Write things down. Short, clear, referenceable. A one-page management playbook for each direct report that covers working style, communication preferences, decision-making boundaries, and current goals. Update it every quarter. When you have five or six reports, this prevents you from treating everyone the same when they are not the same. I once put off writing these for a new hire and spent six weeks guessing what she needed. She could have told me on day two. Management is mostly logistics. It is removing obstacles, clarifying expectations, and creating conditions where competent people can do good work without constant interference. The rest is optional.