The Conversation Before the Conversation
The conversations leaders regret aren't usually the hard ones. They're the easy ones we postpone.
When I was a manager at Adobe, I had a senior engineer on my team whom I was lucky to have hired. He brought years of experience and a clear sense of where the platform’s architecture was breaking down. He kept trying to push some genuinely needed improvements, and he kept getting drowned out by other engineers who were louder, more emphatic, and very good at framing their priorities as emergencies.
I agreed with him. I just didn’t want to cause a confrontation with everyone else. He wasn’t argumentative, so disappointing him was the easier choice. He told me he was frustrated in one-on-one after one-on-one. I kept asking him to be patient. I kept telling myself I’d get to it.
By the time I decided to prioritize his work, he’d already accepted an offer somewhere else to lead an architecture group. It was a big loss for the team. He hadn’t given up on me. He’d just stopped believing the conversation he needed me to have was actually going to happen.
I lost him not from a wrong decision, but from not deciding soon enough. Delaying action was the cause.
Most regrets aren’t about the conversations we had
Leadership advice often focuses on how to have tough conversations, rather than when. Yet timing is where most leaders go wrong.
There’s a specific kind of regret most experienced leaders carry, even if they don’t talk about it. It isn’t the regret of having said the wrong thing. It’s the regret of not having said something earlier, when the situation was still small enough that a small conversation could have changed it. The senior engineer who should have been talked to before they took the offer elsewhere. The peer who should have been challenged on a pattern before it became a culture problem. The team that needed direct feedback before they decided you were absent on the issue that mattered most to them.
The phrase that gives it away every time is some version of “I knew earlier.” Usually followed by some version of “I didn’t have enough information yet.”
“I didn’t have enough information yet” is what we tell ourselves when we mean “I didn’t want to do this yet.”
Two costs that don’t show up on the same timescale
The surface question is when to bring something up. The actual tension lies between two costs that don’t show up on the same timescale.
The cost of being wrong is immediate and visible. You said something. The person was having a bad week. You damaged a relationship over what turned out to be nothing. I’ve done it. Most leaders have.
The cost of being right but late is delayed and diffused. The pattern hardens. The person adapts to the absence of feedback. The team makes up its own story about why nothing is being said. By the time the conversation is unavoidable, you’re no longer having the conversation you wanted to have. You’re managing the wreckage of not having had it.
Leaders tend to avoid the immediate, visible cost of being wrong, but quietly accept the bigger, invisible cost of waiting too long. Incomplete data isn’t a problem to solve; it is just part of leading.
Letting patterns persist teaches you to discount your instincts. Over time, this undermines your own judgment as a leader.
The conversation a manager didn’t have
A few years after the Adobe situation, I was at Spotify, and one of the managers reporting to me had a developer who wasn’t meeting expectations. The developer would get feedback, improve for a while, and then drift back. The manager kept seeing little signs of improvement and kept telling himself the next month would be different. He also wanted to be liked by everyone on the team, and the developer was friendly and well-meaning, which made the avoidance feel less like avoidance and more like patience.
After a few months of coaching him on how to handle it, I eventually scheduled a one-on-one with the developer myself. We had a direct conversation about where the gaps were, what specifically needed to change, and what would happen if it didn’t. The developer thanked me for the honesty, made real changes, and performed significantly better for years afterward.
If their manager had had that conversation earlier, the developer could have avoided months of struggle where they genuinely didn’t understand what they were doing wrong. The cost of waiting wasn’t borne by the manager. It was borne by the person who didn’t get the information they needed to do their job.
Five signals that the conversation is already overdue
When I find myself hesitating, I use a simple checklist to decide if it’s time for a conversation. There are five signals to watch for. If two or more are present, the conversation is already overdue. If only one shows up, I pay closer attention. Sometimes a single signal is enough to justify a check-in, especially if my gut says something is off. At the very least, I set a reminder to revisit and see if another signal appears. Acting early can be as simple as opening a small conversation or just staying intentionally observant.
The first is pattern, not incident. The same thing twice, with the same person or in the same situation. One occurrence is a data point. Two is the start of a trend you’re already discounting.
The second is distance. You’re hearing about the issue from somewhere other than the person involved, and the intensity of what you’re hearing is going up rather than down. Secondhand information getting louder is a leading indicator that the people closest to it have lost confidence that direct routes work.
The third is workarounds. You or someone else has started routing around the person or the topic. The team has quietly redistributed something. A peer has stopped looping them in. You’re choosing different agendas for your one-on-ones to avoid the area. The organization is adjusting before anyone has said anything out loud.
The fourth is cognitive load. You’re spending real mental energy on the avoidance itself, not on the underlying issue. You’re scripting the conversation in the shower. You’re rehearsing it on walks. The thing the situation costs you has stopped being the situation and started being the management of it.
The fifth is inflation. The version of the conversation you’re rehearsing keeps getting bigger. Three weeks ago, it was a casual check-in. Now it’s a serious sit-down. Soon you’ll be thinking about whether HR should be in the room. The seriousness of what you’re imagining is rising even when the situation hasn’t dramatically changed. That’s a tell that your own waiting is now the major factor in how the conversation will go.
This framework doesn’t tell you what to say. It reminds you that if you’re hesitating, your decision is already made. You’re just delaying action.
How to have the earlier version
Once you’ve decided the conversation needs to happen, the goal is to keep it small and recoverable.
Lead with observation, not interpretation. “I noticed X” lands very differently from “I’m worried you’re Y.” The observation can be discussed. The interpretation feels like a verdict.
Make the time deliberate, but small. Not a calendar invite labeled “feedback.” Not an ambush in the hallway. A brief, intentional opening that signals this is a conversation, not a performance review.
Be specific about what you noticed and nonspecific about what it means. The whole point of the earlier conversation is that you don’t yet know what it means. That’s why it’s still cheap.
Be willing to be wrong out loud. Something like “I might be reading this wrong, and if I am, tell me, because that’s also useful information.”
If you feel nervous before starting, a quick mindset shift can help: remind yourself that your goal is clarity, not confrontation. Take a moment to ask yourself: What do I actually see or notice? What outcome would I like? How can I make this feel like a two-way conversation rather than a verdict? Naming even one intention, such as “I want to help” or “I’m trying to understand”, grounds your approach. Sometimes, just taking a breath and framing your purpose before speaking builds confidence to begin.
If any of this sounds like nonviolent communication, it should. If you haven’t read that book, it’s worth your time.
Four ways the earlier conversation goes wrong
The first failure mode is saving it for the regular one-on-one next week. Timing kills it. By the time the slot comes up, you’ve reattached to the avoidance, and the conversation becomes a calendar item instead of a decision.
The second is loading the conversation with everything you’ve been holding back. The earlier version is supposed to be one thing. If you’re stacking up six months of accumulated observations, that’s a different conversation, and the person on the other end will receive it as an attack rather than a check-in.
The third is soft-pedaling to the point where the person doesn’t realize what you said. This is the one I struggled with the most. I’d say something like “I think it’s about time you might start considering how you communicate with the rest of the team,” when what I actually meant was “your communication style is alienating people and they’ve stopped listening to you.” I wrapped the message in so many caveats that the person would walk out thinking they’d received a polite suggestion for some future quarter. I’d walk out thinking I’d addressed it. Both of us were wrong. I’ve gotten better at this out of necessity, but I still catch myself, especially in writing. When I review something I’ve drafted, I now ask whether someone reading it would understand they need to do this now, or whether they’d read it as a suggestion they might think about at some unspecified later point. If it’s the second one, I will rewrite it.
The fourth is routing the conversation through HR or a formal process when it’s actually a relationship conversation. The minute it becomes procedural, you’ve changed what conversation you’re having, and you’ve lost the version that was still recoverable.
As a manager, I rarely regret the challenging conversations I had. My real regrets are the easy conversations I kept postponing, only to watch them become hard.
Why the wait feels responsible
There’s a reason this pattern is so common, especially among newer leaders. Raising something soft feels disloyal. It feels like questioning someone’s competence over a concern you can barely articulate. Most leaders would rather be wrong about timing than risk being wrong about the read. So they choose silence and call it patience.
The person on the other end almost always senses it. They see the workarounds. They notice the topic that doesn’t come up. The silence is louder than the leader thinks, and the absence of feedback is often interpreted in the worst possible way. If someone on your team is upset that other people seem to be routing around them and they don’t know why, that’s often what’s happening. The team adapted before anyone told them that the team was adapting.
One of the most effective ways to break this cycle is to foster a culture where people feel safe raising concerns early. Encourage openness by regularly asking for feedback, and make it clear that surfacing issues is valued, not penalized. Reinforce through your actions that bringing up problems will be met with curiosity rather than punishment. When people see this modeled consistently, it becomes easier for them to speak up, and the leader’s burden to spot and address every issue alone becomes lighter.
The rest of the team is watching too. If they see someone struggling and nothing visible is happening, they will invent a story. The story is rarely flattering to the leader. When the team starts adjusting around someone, your reputation is already adjusting too.
The human cost compounds. Each round of avoidance makes the next round easier to rationalize. After enough cycles, the leader is no longer avoiding a conversation. They’re avoiding their own role.
When waiting reaches the next level up
A few newsletters ago, I talked about a different version of this same mistake from my first CTO job. A conflict was building between the engineering and product teams over prioritization. The engineers kept flagging serious technical debt. The product team kept prioritizing new features. The same tension came up every planning cycle, and the engineers felt they were raising something real only to be overruled every time.
The chief product officer and I both saw it. We had a good working relationship, and we both did the thing that felt responsible. We coached our own sides and told ourselves we were giving the teams room to work it out. What we were actually doing was giving the problem room to grow.
Letting teams resolve their own conflicts is often the right call. The mistake was not noticing that this one had stopped resolving and started compounding.
Eventually, the conflict reached the CEO, not through me, not through my CPO peer. Someone from another part of the organization noticed the issue and mentioned it to him in passing. In our next one-on-one, the CEO said something that has stayed with me. “If it’s getting to me, that means it’s not being handled. And if it’s not being handled, I’m going to feel obligated to step in. I will do a worse job at it than either you or the CPO, because I’m further from the work.”
The cost of waiting wasn’t a number. It was my CEO doing a piece of my job in front of everyone who could see it. The harder cost was with my own engineers. They had been raising the technical debt problem for a long time, to their managers, to their directors, and eventually to me. My silence taught them something. The person who was supposed to have their back on technical health wasn’t visibly in the room on it.
A while after the dust settled, an engineer who’d been part of the original group left. In his exit conversation, he mentioned how long it had taken to resolve this problem, which was one of the reasons he started looking. Once the issue was resolved, he was already in loops with other companies.
The earlier version would have been small. A shared prioritization rule. An agreed-on tech debt budget. One joint planning session with the CPO. Cheap, slightly awkward. The version I actually had came after the CEO was already involved, with both teams dug in and a leader above me watching whether I could run my own organization.
What changed
Earlier in my career, my default was to wait. The reasoning sounded responsible. Get more data. Give people the benefit of the doubt. Don’t overreact. Don’t be the manager who pounces on one bad week.
What’s changed is the recognition that the data is rarely the bottleneck. The bottleneck is usually the willingness to be wrong out loud. Once you stop treating uncertainty as a reason to wait, the cost calculus inverts. The earlier conversation is almost always cheaper, even when you’re wrong.
The shift wasn’t from “wait” to “act fast.” It was from “wait until I’m sure” to “act when I notice.”
I used to think I was being patient. Most of the time, I was just being slow.
A test for this week
Pick the three people or situations you’ve been quietly avoiding bringing something up about. For each one, run the five signals: pattern, distance, workarounds, cognitive load, and inflation. If two or more are present, you’re not waiting for the right moment. You’ve already missed it. The right move is to schedule the smaller version this week, before it becomes the bigger one.
If you’re leading a remote or distributed team, pay extra attention: these signals can be subtler when you are not sharing physical space. Patterns may only show up in tone or delays in messages. Distance might appear as silence in group chats or people dropping off video calls quickly. Workarounds could mean people are excluded from smaller Slack threads or side meetings. Your own cognitive load might show as rereading written updates or hesitating to add an item to an agenda. Inflation often happens invisibly as issues compound without hallway conversations to surface them. Adapting your awareness to the remote context is key: be intentional about checking in, ask more frequently, and actively look for shifts that may not be obvious through a screen.
Performance management is something you do every week. Performance reviews are the box you file at the end. The earlier, smaller conversation is the actual work. Everything else is documentation of what you should have addressed sooner.
The conversation before the conversation is the cheapest version. Every version after gets more expensive.
Have a question about this or something from a previous newsletter or episode? Reach out at contact@itdependsbook.net or find me on Bluesky at @kevingoldsmith.com.
To hear an extended discussion of this topic, please listen to my most recent podcast episode: The Conversation Before the Conversation.
Thanks again for reading! If you find it helpful, please share it with your friends.
Buy on Amazon | Buy on Bookshop.org | Buy on Audible | Other stores