The design ladder I ran had thirteen rows. Six rewarded work that AI has made much faster to produce. None asked who decided what deserved to exist, who could refuse a request or whose standard held after they left.
Both design ladders, in fact. At Grammarly I ran product design and brand and marketing design as one organization, and each had its own ladder. Put side by side they share the same structure: four competency groups, thirteen rows, the same wording in most cells, differing only in which craft skills are listed at the bottom. That is not two ladders. It is one lane with two specializations, and every transfer between them became an argument about whether a BMD4 was a PD4, when the two documents said the same thing in different fonts.
Thirteen rows. Six of them are now measuring something that costs almost nothing to do.
Before the rows, the inheritance, because almost every design ladder in the industry has the same parent.
Where it came from
2016 — Merholz and Skinner
One track, five levels, and management titles appearing inside it rather than beside it. Every ladder below inherits this shape.
Design Manager sits at L3 and Design Director, Creative Director and VP of Design all sit at L5, so two levels carry both an individual and a management meaning at once.
2021 — what we actually ran
Five ladders, three shapes. The two design lanes are identical twins and research is a third, mirroring them level for level; product management is a cousin at the same height; engineering uses the same numbered height but a different shape, with its management track folded into the same line. Every one of them runs to VP.
Product management already scored UI/UX expertise. Product design already scored front-end development and product thinking. Neither ladder mentions the other. Research is a copy of design with the craft list swapped, which was the sensible thing to do at the time and is the thing the fourth essay in this series argues about.
The published ones
Design has one book everybody inherited from and a scatter of company documents published since. They agree on the axes and disagree about almost everything else.
GitLab is the useful operating example here. It names Product Design as the decision owner for how a direction is experienced, alongside Product deciding what and why and Engineering deciding how it is built. Intercom is the stronger example of revising a design ladder after people found its promotion criteria unclear, overlapping and partly outside their control. The UK Government framework contributes something neither does: shared capability language across many adjacent roles, maintained as public infrastructure. These are company- and institution-specific systems, but each makes a seam explicit instead of hiding it in overlapping competency rows. The fork worth knowing is Merholz’s own, and I should say that he taught me most of this in the first place. He has since argued for a trellis rather than a ladder — a structure that admits people move sideways as often as up. That is the right instinct, and it is the one this series builds on. Merholz keeps an org-wide leveling framework underneath his trellis, which is the part people quoting him tend to drop, and it is the part that matters: a sideways move only has an exchange rate if there is one shared core for both sides to be read against.
The six rows that stopped working
Six of the thirteen rows. They apply to all five, because all five inherited them from the same place, but design is where they do the most damage, because design is where the output they measure fell furthest in price.
The meetings row. Attends meetings, then contributes regularly, then drives meetings, then is the person the meeting exists for. This came from 2016 and I copied it without asking what it measures. It measures attendance. The room at Lyft that taught me judgment density had twenty people in it and every one of them would have scored at the top of this row. Not one of them was positioned to end the argument.
The autonomy row. Executes with specific guidance from their manager, rising to leads independently. Distance from your manager as a proxy for trust. That had six meaningful levels when teams were large. On a team of seven senior people it has about three and the rest is padding.
The process row. Understands the standard process, improves it, defines it, shifts to program. Coordination is exactly the cost that collapsed. A level defined by owning how a team coordinates is a level defined by the cheapest part of the job.
The technical expertise row. Strong in one, capable in two, rising to expert in two. The T-shape, drawn across a list of named skills. It assumed those skills sat inside disciplines with edges.
Recruiting, marked optional at every level. Defensible when a portfolio told you something. At Honor the first design team was hired largely on portfolio and a conversation, and it worked, because in 2014 a portfolio was expensive to fake. Now a portfolio can be generated in an afternoon and a take-home mostly measures who had two free weeks. Judgment about people has become one of the scarcer things a senior person owns, and it is sitting in the document with an asterisk beside it.
And one absence. Nowhere does any of the five ask what a person believes should exist. Communication covers whether they can explain a decision. Problem solving covers whether they can frame one. Nothing asks what they would put their name to and be wrong about in public.
What held
The ownership group, which sorts a person into informed, involved and invested, appears in four of the five in almost identical words and has aged better than anything else in any of them. It measures disposition rather than output, and disposition did not get cheaper. It carries over untouched.
So does the promotion criterion: a strong track record at your level, plus impact already visible at the next one. Not potential, not tenure, not a good quarter.
The skills, re-cut
Between them the five ladders name about two dozen skills. Sorted by what five years did to them:
Graphic design. Visual design. Animation. Prototyping. Illustration. Document production. Still visible when done badly; being expert in them no longer levels anybody up.
Front-end development. Product thinking. Information architecture. Basic analytics. The UI and UX literacy the product ladder used to footnote as a bonus. Assumed from the second level, in every lane. Absence is a problem; presence is not a credit.
Writing. Interaction design. Service design. Concepting and art direction. Quantitative research and experiment design. Systems architecture. The proficiency scale itself is fine and stays; what changes is that expert in two only means something when the two are drawn from this band.
Use craft as a gate for the role. It cannot be the only route upward: a person may be excellent in three scarce skills and still make decisions only inside their own work.
Boris Cherny’s original post and replies describe prototyper, builder, sweeper, grower and maintainer as working temperaments in teams that use coding agents. Design adds another recurring job: editing forty plausible outputs down to the one worth building. That judgment has always mattered. Cheap prototypes make it easier to see.
The six axes, in design’s language
What replaces the thirteen rows. Three of the six levels; each cell describes judgment rather than an artifact, which is what lets one shared core serve five functions.
The D4 column is where the old ladder had nothing to say. Every cell in it describes a refusal, and a document that levels on scope cannot score one. Notice also what is absent from the table: not one cell names a tool, a deliverable or a screen.
The failure mode to watch
Privilege can distort every axis. Range depends on who received cross-functional work. Opinions held favors people who were safe enough to be wrong in public. I have written about that mechanism elsewhere. Put both checks on the calibration agenda and inspect the assignments managers handed out. Section 11 of the workbook shows how.
The other caution is about scope. A ladder serves ranking. Development and money are two different jobs, and running all three through one document ruins all three. Whatever you do with this, keep it doing the one thing.
Take the design argument into calibration
If craft got cheaper, the ladder has to grade what became scarce.
The Leveling workbook replaces output-shaped rows with a shared core built on decisions, judgment, opinions, range, ownership and what survives the person. Its design guide translates those six axes level by level, then supplies the evidence log, scorecard and calibration mechanics needed to use them with a real team.
One of five essays taking a function through the same framework; the system they share is set out in full in the workbook. The others cover engineering, product, research and marketing. Written from having inherited, run and rebuilt leveling and review cycles at three companies of very different sizes, and against the public frameworks that preceded them: the levels framework in Org Design for Design Orgs by Peter Merholz and Kristin Skinner, Rent the Runway’s engineering ladder, and Ravi Mehta’s product competency toolkit. Figures from the AI in Design Report 2026 and Figma’s State of the Designer 2026.
© 2026 Renato Valdés-Olmos