Levels and ladders, later than you think

A career ladder is a description of the company you already are, written down so it can be applied consistently. Most first ladders are a description of a company somebody read about.

That's the whole problem with writing one early. You end up with six levels, a competency matrix with four axes, and language borrowed from an organisation with two thousand engineers and a performance calibration department. Then you spend two years explaining why nobody at your company matches any of the descriptions.

The trigger

Write the ladder when the same fairness argument has come up two or three times.

Not once. Once is a conversation, and it usually has a specific answer that doesn't need a framework. Two or three times means the question is structural: two people doing visibly different work with the same title, or the same title with meaningfully different pay, or someone asking what would have to be true for them to get to the next thing and you giving an answer that you'd have phrased differently last month.

That's the signal, and it's a much better one than headcount, because it fires when the pain is real rather than when a number crossed a threshold. It usually lands somewhere between twenty-five and forty engineers. It can land at fifteen if your first ten hires were unusually senior, and it can wait until sixty in a company where everyone joined within eight months of each other.

There's an earlier version of that same signal, and it's cheaper to watch for because nothing has to reach you for it to fire. People start comparing themselves to each other. Not formally, and usually not as a complaint: it shows up in how they talk about who got which project, who's been here longest, who's doing what. Comparison is the behaviour that precedes the fairness argument. By the time the argument lands on you, the warning already went past once.

The cost of writing it early is not the effort of writing it. It's that a ladder is a promise, and revising a promise costs you more credibility than not having made it.

Three axes, and no more

Every real ladder reduces to three things, and every bloated ladder is these three plus twenty specific behaviours that will be out of date within a year.

Scope. How big is the thing this person is responsible for. A feature. A system. A team's worth of systems. An area of the business.

Autonomy. How much context do they need supplied. Junior people need a defined problem. Senior people take an ambiguous one and come back with a defined problem plus a recommendation. The step from "solves the problem given" to "works out which problem" is the single biggest jump in an engineering career and most ladders bury it.

Impact. What changes because they were here rather than someone else. Note that this is deliberately not "output," and the difference matters at the top of the ladder, where the most valuable thing a person does is often to prevent six months of work from happening.

Three axes, five levels at most, and write each level as a paragraph rather than a grid. A paragraph forces a coherent picture of a person. A grid lets you tick nine boxes and produce someone who doesn't exist.

Staff-plus, and the incentive that looks like a promotion

The reason to have a real senior IC track is not fairness or optics. It's that without one, every ambitious engineer eventually concludes the only route up is management, and you convert your three best engineers into three mediocre managers in the same eighteen months.

I won't promote anyone into management unless there's a genuine IC path they could have taken instead. Genuine means it goes as high, pays comparably, and has actual people on it who are visibly respected. A staff track that exists on paper with nobody on it is worse than no track at all, because it's now visible evidence that the company says one thing and does another.

Genuine also means it spans a whole career. A track that runs two levels and stops isn't a path, it's a courtesy tier, and everyone standing on it can see the ceiling from where they are. The test: could somebody join you at mid-level and still have somewhere to go twenty years later without ever managing a single person. If the answer is no, management remains the only real route up and all you've done is make the signage more elaborate.

The trap at staff-plus is that the level has to be justified by scope the company actually has. A staff engineer's job is leverage across teams, and if there's only one team, there's nothing to be leveraged across. Promoting someone to staff because they're excellent and you don't want to lose them creates a role with no content, and the person feels it before anyone else does.

If you're at that point and the scope doesn't exist yet, pay them properly at the level below and say exactly what you're doing and why. People take that better than a title with nothing behind it, and much better than finding out later.

Migrating a team on without a riot

This is where ladders usually go wrong, and the failure is almost entirely about sequencing.

Map everyone privately first, before the ladder is announced, using the draft. You're checking whether the ladder describes your actual people. If four people land at a level that doesn't match how they're regarded, the ladder is wrong, not the people. Fix it before anyone sees it.

Fix the pay before you announce the levels. The mapping will surface people who are underpaid relative to where they land, and the gap is usually someone who joined early and never got corrected. If you publish levels first, those people find out they're underpaid from a document, which is a bad way to find out.

Nobody goes down. Even when the honest read is that someone is levelled above where the ladder puts them, a demotion delivered by spreadsheet is the fastest way to lose a person and to teach everyone else what the exercise was really about. Freeze them, be honest with them in private, and let it resolve over time.

Announce the ladder as a description, not a change. "Here's how we already think about this, written down." Which should be true, if you waited for the trigger.

And expect the first month to be noisy anyway. Some people will discover they're not where they thought. That conversation is one you owe them regardless, and the ladder didn't create the gap, it exposed it.

What a ladder can't do

It won't make promotion decisions consistent by itself. Two managers reading the same paragraph will disagree about the same person, and the only fix is a conversation between them with examples on the table.

It won't tell you who to promote. It tells you who's already operating at the next level, which is the correct rule (promote to recognise, not to motivate) and a genuinely hard one to hold when someone good is unhappy.

And it won't survive contact with a fast-growing org unchanged. Plan to revisit it annually, in a short session that asks one question: is this still a description of who we are, or has it become aspirational. Aspirational ladders quietly raise the bar on everyone without anyone deciding to.