Career ladders and grades for a small IT company: criteria beat vibes

Promotion as a negotiation lottery

In a company without grades, a raise has exactly one mechanism: asking. Which means pay tracks assertiveness, not contribution — the loudest negotiators compound, and the engineers you most want to keep quietly interview elsewhere instead of negotiating. By the time their counter-offer conversation happens, it's a resignation conversation.

The follow-on damage is just as predictable:

• Every comp decision is bespoke — decided in the moment, against no reference, so two people doing the same work drift apart for no reason anyone can state.
• Titles inflate — "senior" becomes a retention patch instead of a bar, and then means nothing in hiring.
• New hires leapfrog loyal people — market offers get benchmarked, internal pay doesn't, and the gap (pay compression) surfaces at the worst moment: when people compare notes.
• "What do I need to do to grow here?" has no answer — and for strong engineers, no answer reads as "nothing here for you".

None of this requires bad intent. It's the default physics of unstructured pay — and the fix is boring, written structure.

What a ladder needs at 20 people

Not Google's ten levels. At 10–50 people a ladder is a one-page document per track:

• 3–5 grades per track. Junior → middle → senior → lead is plenty. More granularity than you can honestly distinguish is false precision — every boundary you add is a boundary you'll argue about.
• Written criteria per grade, phrased as observable behavior. "Designs and delivers a service end-to-end without supervision" is a criterion; "is senior enough" is a vibe. A short checklist per grade beats a page of prose — you'll assess against it, so make it assessable.
• A salary band per grade. A min and max, in a stated currency. Bands are what turn "can we afford this raise?" into "where in the band are they, and does the move fit?" — and they're the structure pay-transparency rules increasingly expect.
• Separate tracks where the work differs. Engineering, QA and design don't share criteria; and if you can, keep a senior-IC path distinct from the management path so the only way up isn't a job people don't want.

Write it down, publish the criteria to the team, and date it. An unpublished ladder is a rumor.

Running promotions against criteria

A ladder on paper changes nothing until promotions actually run against it:

• Readiness is signals, not feelings. Time in the current grade, where current pay sits inside the band, and the last review outcome — three cheap signals that tell you who to look at first. None of them promote anyone; they prioritize the conversation.
• Evidence is collected against the written criteria. The promotion case is "here's the criterion, here's what they did that meets it" — pulled from review history and goals, not from whoever tells the best story in the room.
• Promotions happen on a cadence. Tie decisions to review cycles — twice a year is typical — instead of ad-hoc, whenever-someone-pushes timing. A cadence removes the penalty for not pushing, which is the whole point.
• A "not yet" comes with the gap named. Against a checklist, a rejection is "these two criteria, here's what evidence would look like" — a growth plan, not a brush-off. That answer retains people; silence doesn't.

The shift is subtle but total: from "convince me" to "show the criteria are met". Same decision-makers, radically different fairness.

The EU pay-transparency angle

The EU Pay Transparency Directive makes this more than good practice. Among the duties that apply to employers of all sizes: pay and progression must rest on objective, gender-neutral criteria, candidates get pay ranges, and employees gain a right to information about pay levels for comparable work.

A written ladder with criteria and bands is, in practice, the mechanism for all of it:

• "Objective, gender-neutral criteria" is precisely a documented grade checklist — levelling you can point to when asked why one person is paid more than another for comparable work.
• Answering right-to-information requests requires knowing your own structure: grades, bands, and who sits where. Companies without one will be reconstructing it under a deadline.
• Posting salary ranges for candidates is trivial when every grade already has a band — and a scramble when it doesn't.

For a growing team the timing logic is simple: building bands and criteria at 20 people is a week of founder time. Retrofitting them at 100 — under reporting obligations, with existing pay gaps to explain — is a project.

How Helia HR does this

The Career paths pack ($1/employee/mo — see pricing) turns the one-pager into a working system:

• Per-department grade ladders — each with ordered grades and a criteria checklist per grade, so "what senior means here" is a list, not lore.
• Salary bands per grade — optional min and max in your currency, kept with the grade definition. Comp itself stays behind the strictest permission gate.
• Each person's current grade and the path forward — employees see where they are, the next grade's criteria, and their full grade-transition history (when, by whom, with notes).
• Promotion readiness signals — a 0–100 advisory score per person from time in grade, position within the salary band, and the latest review rating, with a ranked candidates view for HR. It prioritizes conversations; humans make the call.
• Promotion requests — filed against a target grade with a supporting message, so the case and the decision live on the record instead of in a hallway.
• An AI growth plan the manager edits — drafts a development plan from the person's grade, the next grade's real criteria, their goals and 360 feedback. The manager rewrites and owns it; salary figures never enter the prompt; the workspace owner can switch AI features off globally or one by one.

Reviews connect: the Reviews & Goals history is exactly the evidence a promotion case runs on.

FAQ

How many grades should we start with?

Three to five per track. If two adjacent grades are hard to tell apart in real people, merge them — a boundary you can't defend in a promotion conversation is a liability, not precision.

Should we publish salary bands internally?

Publish the criteria to everyone at minimum — an unpublished ladder can't guide anyone. On bands, teams differ: some publish every band, some show each person their own grade's band. The direction of regulation is one-way, though — candidates already get ranges under the directive, so build bands you'd be comfortable stating.

Is a high readiness score an automatic promotion?

No. It's an advisory signal built from tenure, band position and review outcomes — a prioritization aid for who to evaluate first. Budget, business need and the evidence-against-criteria case stay human decisions.

What about people who manage vs. people who don't?

Keep them as separate tracks with separate criteria — a senior-IC path and a management path. The classic failure is making management the only route up, which promotes your best engineer into a job neither of you wanted.

We're 12 people. Isn't this premature?

Twelve people with two future seniors is exactly when it's cheap. The ladder is a page, the bands are six numbers — and the first "why did they get a raise and not me?" conversation costs more than writing both.