At around 25 people, something shifts. The team is big enough that you can no longer hold everyone's performance in your head. Someone asks about career progression. A manager comes to you unsure what to do with a struggling team member. And so you do what feels responsible: you build a performance system.
You open a doc and start writing. And what comes out, almost without thinking, is some version of what you sat through at your last job. Maybe your last two. A review cycle, probably annual. A set of rating criteria. A template for the conversation. A score.
It feels like structure. It is structure. What it is not, necessarily, is design.
A Gallup study found that 98% of CHROs are unhappy with how performance management works in their organisations. Gartner named it the most urgent HR reform priority of 2026. These are people whose entire function exists to get this right. They have dedicated teams, consultants, and benchmarking data. And still, they look at what they built and say: not this.
Founders building their first system at 30 people rarely have any of that. What they have is memory.
Memory is a powerful design input. It tells you what a performance conversation looks like, how often it happens, what gets written down and what gets said out loud. It gives you a ready-made template at exactly the moment you need one.
The problem is that memory does not come with a filter for what worked and what did not. It captures the format, not the outcome. And the format of most performance systems was never designed for performance. It was designed for documentation, defensibility, and the kind of consistency that protects large organisations from legal exposure. None of which is what a 30-person company actually needs.
What founders tend to build is a scaled-down version of something that was already a compromise. An annual cycle because that is what a cycle looks like. A rating scale because you need something concrete to say. A form because there should be a record.
The team reads it correctly. Not because they are cynical, but because most of them have sat through it before. They know what this kind of system signals: that performance is something measured at a fixed point in time, reviewed with a number, and filed away. That the conversation in the room matters less than the document produced after it.
The founders who do this differently rarely have more HR knowledge. They are often the ones who have paused long enough to ask what they actually want to know. Not "how is this person performing against their objectives" but "what does this person need to do their best work, and are we giving it to them?"
That is a different question. It produces a different conversation and a different kind of culture. One where performance is continuous, rather than something that arrives twice a year with a score attached.
It is harder to design from scratch than it sounds. But it is considerably easier at 25 people than at 250.
The system you inherit by default is the one you absorbed without realising it. The system you actually want requires a moment of deliberate choice that most founders skip because it does not feel urgent yet.
It always does later.