People do not leave for money as often as you think
Every resignation I have received had a reason underneath the reason. Here are the ones that actually showed up, what I could have done about them, and the retention conversation I now have before it is too late.
When an engineer resigns, the reason they give in the exit conversation is usually a better offer. That is true and it is also the least useful fact available. Everyone has a better offer somewhere. The question is why they went looking.
I have kept notes on every resignation from a team I ran. Underneath the offer, the reasons cluster into a short list, and money by itself is near the bottom.
The reasons that actually showed up
They could not see the next year. The most common one. The work had become predictable, the promotion was vague, and nobody had described a future that was different from the present. An offer from outside is a future by definition. That is its whole appeal.
Their manager changed. Sometimes that manager was me, arriving. People are loyal to a manager who invested in them, and when that person leaves, the investment feels like it left too. The window after a manager change is the most dangerous period for retention on a team, and most organizations treat it as routine.
Something broke their trust and nobody fixed it. A promotion promised and not delivered. A reorganization announced badly. A public credit that went to the wrong person. Individually small, and I have watched each of them be the real reason.
The work stopped mattering. Engineers will grind on hard problems for years if they believe the problems matter. A quarter of work that gets cancelled, or ships and is never mentioned again, drains that faster than any deadline.
They were carrying something alone. The only person who understood a system, the only one on the pager rotation who could handle a certain class of incident. That is a compliment for a while and then it is a trap.
The stay conversation
I now have one specific conversation with each person on my team, once or twice a year, before there is any sign of a problem. I ask it directly: what would make you leave, and what would make you stay?
The first time people hear it they are startled, because it sounds like a question you ask on the way out. Asked on the way in, it produces the most useful answers I get all year. One engineer told me she would leave if she was still on the same product in eighteen months. That was fixable and I fixed it. Another told me he would stay through anything as long as the on call rotation did not get worse. I knew, then, what not to touch.
I write the answers down in their 1:1 document. I read them when I plan the quarter. It is the single highest return hour I spend on retention.
What I can and cannot fix
Some reasons are in my power. Assignment, scope, recognition, the next project, the honesty of the promotion conversation. Those I own completely, and when someone leaves over one of them I do not get to blame the market.
Some are not. Compensation bands I do not control. A company strategy that has moved away from what someone wants to work on. A partner who took a job in another city. For those, the best thing I can do is help the person leave well: give them time to interview, write the reference, make the last month a good one. People remember how they were treated on the way out, and their friends hear about it.
The counteroffer
I have made counteroffers and I have mostly regretted them. Money fixes the reason they gave. It does not fix the reason underneath, and it teaches the rest of the team that the way to get a raise is to bring an offer. The exceptions were cases where the underlying reason was genuinely a compensation gap I had been unable to close through the normal process, and the resignation was the lever that finally moved it. Even then, half of those people left within a year.
What the AI rollout taught me
Retention during a big change runs on one question: does this person believe they still have a place here? The year we moved to an AI first way of working, nobody left because of the tools. Two people nearly left because they were not sure the tools left room for them, and neither of them said so until I asked.
Once they said it, it was fixable. Both are still on the team. The version of that year where I did not ask ends with two resignations that the exit interview would have blamed on better offers.