The Case Against Public Levels

In Dragon Ball Z, the Saiyans wear a scouter, a little eyepiece that reads someone’s power level and puts a number on it. Vegeta lives by that number. He sizes up every opponent by what the scouter tells him, and his whole sense of himself is wrapped up in pushing his own reading higher than everyone else’s. The number stops being a description of the fighter and becomes the thing he obsesses over.

Public engineering levels are like that scouter.

When I joined Pocus I was unsure about a lot of things, but certain about a few. One of them was to delay public levels as long as I possibly could. Some people found it odd. Most startups get this backwards. Seed and Series A companies, twenty or thirty engineers in, stand up a public ladder almost out of the gate. It’s wrong for the stage, and wrong in a way that’s hard to walk back.

Levels

I think of levels as an abstraction over what an engineer actually does. You take the work, and the work you wish they were doing, and spread it across a few axes. Technical depth, collaboration, impact, whatever set you pick. Then you grade against it. That's probably fine as a private exercise.

As soon as a level is public, it becomes the thing people optimize for. It's human nature. Competitive engineers want to show progress, and when progress has a public form, that form is what they chase. This is worst at a startup, where the most valuable thing someone does in a month often lands on no axis at all. Three weeks on-call so someone else can ship. A weekend unblocking a customer. A rewrite nobody asked for that saves the quarter. None of that grades well, so people quietly stop doing it. You wanted better engineers. You get better level-chasers.

Accountability

The usual defense is accountability. Public levels supposedly keep everyone honest about who is operating where. It rarely works at the scale of a young company. At every place I’ve worked that had public levels, there were people sitting at levels they had no business being at. Everyone knew. Their managers knew. Nobody did anything.

The reason is simple. Performance management is hard, and early orgs are terrible at it. Most people doing management are doing it for the first time, and first-time managers are notoriously bad at managing performance. There are a hundred reasons someone might be coasting at a level. Health, motivation, something happening at home. Sorting that out takes a calibration muscle that a young org just doesn’t have yet. We design these systems assuming clean execution and then hand them to people still learning the basics of how to give hard feedback.

A wrong level that nobody corrects doesn’t just sit there. It starts winning arguments. In cross-org discussions, the higher level tends to carry the room. Level becomes a proxy for authority, which is a bad thing anywhere and a worse one at a startup. Staff engineers get deference because they’re staff, and a lot of them arrive with strong opinions they get to act on because of the title, not the idea. Early on you want the best idea to win, wherever it comes from. A level that pre-wins arguments is the opposite of that. The part people resist hearing is that making someone re-earn their standing on the merits isn’t a cost of having no levels. It’s most of the point.

Sequencing

So what should you actually do?

First, make sure money isn’t forcing the issue. Most companies tie compensation to level, on the logic that a higher level means a scarcer skill means more pay. That holds at a big company, where you can swap one L5 for another. At a startup nobody is swappable. Every engineer is a specific bundle of value, and when a level bump is the only way to earn more, people will chase the bump even when it cuts against what the company needs. Decouple the two. Bonuses, equity, wide overlapping bands all do the work. Once money stops routing through level alone, most of the pressure drains out.

While the org is still small enough to hold in your head, run with no levels at all. That doesn’t mean no expectations. It means managers own a clear, written growth plan for each engineer, just not pinned to an abstract ladder. It will feel inconsistent, and you’ll have to sit with that. What you get back, engineers who aren’t thinking about promotions, is worth far more.

Once you’ve built a little of that muscle, you introduce private levels. Now there’s a real framework underneath, but only the engineer and their manager know where someone lands. This scales much further than people think, and plenty of well known companies run for years exactly this way. Levels still leak, because people talk, and that has never once been the disaster everyone worries it will be. What private levels buy you is room to be wrong. You run the framework through a few calibration cycles and watch what it does, whether it actually measures anything, whether two managers grading the same engineer land anywhere close. You make the levels public, if you ever do, only after the framework has earned it.

None of this is forever, and none of it is a substitute for fixing management. There is some scale at which levels start to earn their cost, and you add them then. Early is not that moment. And if your performance management is broken, levels won’t fix it. They’ll make it worse, because now the dysfunction has a number attached and a comp band defending it. Build the muscle first. Add the levels second. Most startups do it in the other order, and everything downstream gets harder.

Today

AI makes the case stronger. The gap between a junior and a senior engineer is closing faster than any ladder can keep up with. For now it's one more reason to stay at no or private levels longer than feels comfortable, until the dust settles on what these tools actually change.

Public levels do more harm than good unless they're added carefully, at the right stage, on top of management that already works. Most companies clear that bar far later than they think. So the right default is to avoid them as long as you can, and to treat the urge to add them early as a sign that something else needs fixing first.

At Pocus, no levels never once stopped us from shipping or from paying people what they were worth. It also kept away the people who show up for a title instead of the work. The overhead we never carried, the promotion theater, the arguing over made-up numbers, the engineers optimizing a level instead of the company, is a big part of what let us do the work at all.