Letting Someone Go: The Last Step, Not a Surprise

From the field: “This is the first time I hear this”

They were a mid-level backend engineer on the reporting team. In the last six months, 7 of the 11 tasks they took did not finish inside the sprint. Two bugs reached production. One of them made the end-of-day transaction report reach the regulator 40 minutes late.

I had talked about all of this in our one-on-ones. At least, I thought I had. What I actually said was: “We need to speed up a bit.” “Let’s look at that reporting work again.” “Let’s keep the end of the sprint a bit tighter.” None of these said “if this continues, you cannot work here”. None of them were written.

The decision formed in my head in the week after the second production bug. The final conversation took 40 minutes. I spent thirty of them defending the decision, because the person in front of me kept asking, fairly, “when did you tell me?” I had no date to give.

Why the surprise is your fault

Something you only hint at is, for the other person, something you never said. Someone on every team hears “we need to speed up a bit” every week. Nobody takes that sentence to mean “my job is at risk”. And they are right not to.

I wrote about how to give feedback in a separate post: event, impact, expectation. That post had one rule: if you are saying the same thing for the third time, you are past the feedback stage. This post starts there. How to measure performance is another topic. Here I assume the measuring is done and the problem is clear.

What I really missed was this: the soft sentences were not protecting them. They were protecting me. For nine months I postponed my own discomfort. They paid the bill all at once, in the final conversation.

Written expectations, a fixed period, check-ins

The second time I reached the same point, I changed the order. After the third conversation I wrote one page, and we both read it. It had four things: the reason, the expectations, the support and the dates.

IMPROVEMENT PLAN  -  6 weeks, check-in every 2 weeks
------------------------------------------------------------
Reason        : 7 of the 11 tasks taken in the last 6 months
                did not finish inside the sprint. 2 prod bugs.
Expectation 1 : A task you take finishes inside the sprint, or
                you say "this will not finish" by mid-sprint.
Expectation 2 : Every PR that goes to production has tests and
                a rollback step.
Expectation 3 : Every Friday, task status in 3 lines, written.
Support       : 1 hour a week pairing with a senior engineer.
Check-ins     : week 2, week 4, week 6. Every result is written
                down and both sides read it.
Outcome       : In week 6 each item is "met" or "not met".
                If not met, the employment ends.
                This line is written in the plan CLEARLY.

The last line is the hardest one to write. But without it, the plan reads like a suggestion for growth. The person only understands how serious it is in week six. So you have not removed the surprise. You have only delayed it.

Why six weeks

Less than two sprints measures noise; one sprint can just be bad luck. More than three months tires both sides, and the team notices. Our sprints are two weeks long, so six weeks is three full sprints. Not too short, not too long.

Check-ins produce dates

The real job of a check-in is to produce a date. In week six, the question “how did we get to this decision?” will come. The answer needs to be a timeline, not an impression. The second time, the timeline looked like this:

My test for myself: if I cannot write the timeline on one page, I am not ready for the decision. In that case, the next step is not the final conversation. It is the next check-in.

The plan must be possible to pass

If you have already decided, the honest path is not a plan. It is a direct conversation. Talk to HR and legal first about what is possible. But do not act as if there is a chance that does not exist.

The final conversation: short, clear, not open for debate

The first time, I talked for 40 minutes, because I had no dates and had to defend the decision inside the conversation. The second time, it took 12 minutes. The only thing that changed was that the conversation was a notice, not a decision. The decision had already been made in week six, on a page both sides had seen.

FINAL CONVERSATION  -  10-15 minutes, someone from HR present
------------------------------------------------------------
1. First sentence is the decision:
      "Today we are ending your employment."
2. One sentence of reason:
      "2 of the 3 expectations in the plan were not met.
       We also talked about this in week 2 and week 4."
3. What happens next: last day, payments, access, reference.
4. Stop and listen: do not open a debate, do not defend
   the decision again.
5. Close: "If anything comes up today, HR is here.
   From tomorrow you can also write to me."

DO NOT: spend the first 5 minutes on small talk.
DO NOT: say "it was not my decision" (it was).
DO NOT: go through the plan items one by one again.

Saying the decision in the first sentence feels harsh. But the person knows why they are there the moment they walk in; seeing someone from HR is enough. Every extra minute is waiting for them and relief for you. Listening is not the same as negotiating. They say what they want to say. You do not reopen the decision in reply.

“How did we get to this decision?”

This question comes every time. The first time, my answer was an impression: “In the last few months things did not go the way we wanted.” The person argued with that, and they had a right to, because you can argue with an impression. The second time, my answer was the timeline: “We set up the plan in the third conversation. In week two, two items were met. In week four, one. In week four we said there were two weeks left.” Nobody argues with a date. The person was still upset, but there was nothing left to argue about.

There is a trap here: moving back into defence while you read out the timeline. Say it once, then stop. If they say “but that week this happened”, listen. The place for that conversation was the week-four check-in, not today.

What to tell the team, and how much

The first time, I said nothing for two days. I thought it was respect for the person. In those two days, rumours filled the gap: was there a reorg, were we shrinking, who was next? The following week, three people asked the same thing in their one-on-ones, in different words: “Am I doing OK?”

The second time, I told the team the same day, at 17:30, in five minutes.

The left column is for the team’s sense of safety. The right column is for the dignity of the person who left. They do not conflict. The team’s real question is not “why did they leave”. It is “could this happen to me, and would I know in time?” The best answer is that everyone already hears a clear picture in their own one-on-ones. How to build that is in the one-on-ones post. Few things shake psychological safety as fast as someone leaving.

Access and handing over the work

The first time, I made two mistakes. Their access to the VPN and to the production database read replica stayed open for six days; nobody owned the removal list. The second mistake cost more. The night job that uploads the report to the regulator’s portal used a certificate tied to their personal account. Getting a new certificate took two working days. For those two days we uploaded the report by hand.

In the security audit post I explained why access removal should be measured by the longest time, not the average. One account left open after someone leaves makes the whole process meaningless.

Respect for the person

A correct decision does not mean the person is a bad person. Most of the time they are a good person in the wrong job, the wrong team or at the wrong time. A few rules:

• They do not pack their desk in front of the team. An early slot the next morning, or we send their things.
• If they want to write a goodbye message, they write it. In the team channel, in their own words.
• The reference is honest. I write what they did well. The second person’s testing discipline was really good, and I wrote that.
• A “what’s next for you” talk is offered. Not forced. The second time, we had coffee two weeks later and talked about where they would do better.

What to track

What did not work for me

• Hinting. For nine months I said “let’s speed up a bit”. They heard it as small talk, not a warning. That was a normal way to hear it.
• Writing a plan after I had decided. Once, I wrote a four-week plan when I had already decided in my head. At the week-two check-in they said: “This is already over, isn’t it?” They were right. The last two weeks were wasted for both of us.
• Softening the final conversation. The first five minutes were “how are you, how was your week”. They knew why they were there. Those five minutes only made the wait longer.
• Staying silent with the team. I thought it was respect. Two days of silence made three people doubt their own jobs.

Checklist

Conclusion

That Tuesday, of the three people in the room, I was the least prepared. The decision may have been right. But every sentence I had not written in nine months came back to me, one by one, in those 40 minutes.

The second time, the decision was still heavy; nothing about it got easier. What changed was that the person knew what they would hear when they walked in. At the end they said: “Yes, I knew this was coming.” That is the best sentence you can hear in a conversation like this.

The test: if someone on your team is underperforming, do they know today that their job is at risk? If not, you are the one preparing their surprise.