Performance and Growth
helping people do their best work
Learning Objectives
- Give specific, actionable feedback using a structured format.
- Run a useful performance review rather than a compliance exercise.
- Help staff identify and pursue realistic career development goals.
Before
Tahia had been submitting pull requests with commit messages like "fix stuff" and "more changes". Jess had been meaning to raise it for two months. She kept not doing it because it felt awkward and because the code itself was fine.
The problem with postponing is that it compounds. By the time Jess did have the conversation, Tahia had a backlog of sixty commits with useless messages, and Jess had two months of low-grade frustration coloring how she interpreted everything else Tahia did. What would have been a two-minute conversation in week one had turned into an uncomfortable one in week nine.
Feedback
Vague feedback like "your communication needs work" sends someone away with nothing they can act on; specific feedback gives them a target. The SBI format makes difficult conversations more manageable by forcing you to be specific:
- Situation: "In the last four pull requests…"
- Behavior: "…your commit messages describe what changed but not why, for example 'fix stuff' and 'more changes'…"
- Impact: "…which means when I need to understand the history of a change six months from now, I have to read the code rather than the log."
Use the same format for positive feedback: "In last week's code review, you caught the edge case in the parser before it was merged, which saved us from a bug that would have affected every user of the wildlife dataset."
When you give negative feedback, do it in a one-to-one meeting. Ambushing someone in a group setting turns the conversation into a status competition rather than a discussion about improvement.
Giving feedback is uncomfortable in both directions. The first time Jess used the SBI format with Tahia, she rehearsed it three times before the meeting. Tahia's response was, "Oh, I didn't know that was a problem. I'll fix it." The conversation Jess had been dreading for two months took four minutes. The discomfort before the conversation is almost always worse than the conversation itself.
Performance Reviews
Annual performance reviews are often described as a management best practice. In practice, most of them are a compliance exercise: a form gets filled out, a box gets checked, and nobody's behavior changes. The review that actually helps someone is the one that surprises them with nothing.
That means the one-to-one meeting has to do the real work. A weekly or biweekly one-to-one with a three-part agenda (what did you do, what are you planning, what is blocking you) gives you and the person you manage a shared record of expectations and progress. When the annual review arrives, you are summarizing a conversation that has been happening all year, not reconstructing one from memory.
A few things that make reviews more useful and less painful:
- Write down expectations before the review period starts, not during it. "Meets expectations" only means something if the expectations were documented in advance.
- Separate the performance conversation from the compensation conversation. When the two happen in the same meeting, people stop hearing the feedback the moment money comes up.
- Document everything. If a performance problem is serious enough to discuss, it is serious enough to put in writing. A pattern of undocumented verbal warnings does not help anyone.
If you are not sure whether a performance issue is serious enough to document formally, ask yourself: if this person were fired in six months, would I be able to explain to HR why it wasn't a surprise? If the answer is no, start writing things down now.
Training and Skill Development
Research software teams tend to underspend on training for the same reason they underspend on documentation: the return is diffuse and delayed, and the immediate cost is visible. The result is that people learn on the job in ways that are slow, uneven, and occasionally expensive.
A more deliberate approach starts with identifying the gap between what someone can do and what the project needs them to be able to do—not in the abstract, but for specific tasks coming up in the next quarter. Training that connects to near-term work is more likely to stick than training that feels like a general self-improvement exercise.
Training can take several forms:
- Pair programming and code review: the cheapest form of skill transfer that already happens in most projects, usually without being treated as training.
- Structured reading: a paper or chapter a week, discussed in a one-to-one. In practice, this rarely works because it is the first thing to be set aside when "real" work needs to be done.
- Preparing a presentation: the best way to get someone to learn something is to tell them that they'll be teaching it to others four weeks from now.
- Workshops and short courses: these are more expensive, but useful for filling specific gaps.
- Conference attendance: also expensive, but useful for exposure and networking.
Whatever you budget, be consistent. If one person on a two-person team gets professional development funding and the other does not, you have created a grievance regardless of the reason.
Career Development Conversations
Most managers wait for staff to bring up career goals. Most staff are nervous to do this because they do not want to seem like they are angling for something or planning to leave. The result is that career conversations do not happen until someone has already decided to go somewhere else.
The conversation is simpler if you make it a standing item rather than a special occasion. "Where do you want to be in two years, and what would we need to be true for you to get there?" is not a threatening question if you ask it every six months. To make the discussion productive:
- Be honest about what is and is not possible in your project and institution. Promising a promotion path that does not exist does more damage than having no promotion path.
- Connect the person's goals to the project's needs where you can. "We need someone to own the data ingestion pipeline, and you want more infrastructure experience—those could be the same thing" is a conversation worth having.
- Document what you agree to.
Career development for research software engineers is complicated by the fact that RSE is still not a standard career track at most institutions. The people you manage may be trying to navigate toward a research position, a staff engineering role, or something that does not have a name yet. Be honest about the limits of what you can offer, and about what you have seen work elsewhere.
Exercises
Feedback Role-Play (8 min)
Prompt the LLM with this:
Play the role of a junior programmer on my research software team. I'm going to give you feedback about a real situation. Respond in character: push back if the feedback is vague, accept it if it's specific and fair. Here's the situation: [describe a real or plausible scenario where you needed to give someone feedback but found it difficult].
Have the feedback conversation using the SBI format. If the LLM accepts your first attempt without pushback, your feedback was too vague, so try prompting it with more specifics.
Career Conversation (7 min)
Think of someone on your team or someone you manage in a volunteer project. Write answers to these two questions as if you were that person:
- Where do I want to be professionally in two years?
- What would need to be true for me to get there?
Then switch perspective: write one thing the project could offer that would help them move toward that goal, and one honest constraint you would need to name.