Research has public job families, but no field-wide ladder with the standing Merholz has in design, Fournier in engineering or Mehta in product. GitLab runs Research from Intermediate through Principal; Intercom and the UK Government publish substantial paths too. Teams still lack one common reference they can argue with.
The research teams I ran at Grammarly and Pitch had levels. We borrowed the product design ladder: same count, same compensation bands, and the same expectations about scope and autonomy translated into research language.
That mirroring was the sensible thing to do at the time. Research reported into design, design had the nearest usable standard, and building a fifth document from scratch for a team of that size would have been ceremony. But a borrowed ladder inherits the assumptions of the discipline it was borrowed from, and design's assumptions are not research's.
Research has public job families. It still lacks a field-wide standard.
What the field has to work from
The short shelf
There is no research equivalent of Fournier or Mehta — nothing the field argues about — but there are serious company frameworks. They show the work clearly even when they have not become a common standard.
The earlier version of this essay missed GitLab. Its Research family runs through Principal and separates foundational and strategic research from Product Design’s solution validation. Intercom adds six IC stages and a separate management craft. The UK Government publishes six maintained levels with assurance, inclusive practice and strategic influence. Together they make the public shelf much stronger, even though none has become the common standard. Many company ladders still stop early, so the complaint that there is nowhere left to grow persists.
I have some sympathy for the borrowing, because I know what the alternative costs. At Honor, where I was the founding designer, the first research program was one I ran myself, in the homes the product was for, before the product had a name. Nobody leveled it, because there was nobody to level it against. When the team grew to twenty across product, brand, design ops and research, the researchers were leveled against the designers, because the designers had a ladder and the alternative was writing one in a week I did not have.
What the borrowed ladder got wrong
Mirroring design was a defensible shortcut and it was also wrong in a specific way. A design ladder measures the progression from executing a screen to owning a surface to shaping a strategy, and each level is anchored in artifacts somebody can look at. Transposed to research, that progression becomes: runs a study, owns a research area, shapes what the company believes.
The first two translate. The third does not, because the thing that separates a strong senior researcher from a strong staff one is not the size of the area they cover. It is whether the organization changed its mind.
At Pitch the research team produced the finding that turned the company’s plan. It was not a large study. It was the right question, asked once, with the nerve to bring back the answer nobody had ordered, and it changed what the company believed about itself. Under the borrowed ladder that work scored as a single study in a single quarter. A researcher can run flawless studies for two years, produce reports everybody praises, and change nothing. Under a borrowed design ladder that person looks promotable: their scope grew, their studies got more complex, their stakeholder list got longer. Under a research ladder that measures the right thing they are stuck, and everyone senior can feel it without being able to point at the row that says so.
The design ladder, with the nouns changed
What happens when a research ladder is transposed from a design one. Four of the rows survive the translation. Two arrive meaning something they cannot measure, and one of them is the row the discipline actually turns on.
The shaded rows are the ones that do not survive the transposition, and they are the two that separate a strong senior researcher from a strong staff one. Everything above them is a measure of activity.
The border is not with design
The other thing the mirror got wrong is where research actually sits. Because research usually reports into design, the assumed adjacency is with design. In practice the live boundary is with data science and analytics, and it has been for years.
Research and data science often start with the same question: why did this cohort stop returning, what are people doing with the feature, which explanation fits the number? One side may bring interviews and the other a query. Too often they meet only at the readout.
What has changed is that the tooling collapsed. A researcher who can write the query is no longer unusual. An analyst who can run six conversations to work out what the drop-off means is no longer unusual either. Synthesis, the thing that was supposed to be the researcher's protected craft, is now the thing most obviously assisted by a machine, which is uncomfortable and worth saying plainly.
In Boris Cherny’s original post and replies about agent-shaped teams, the scout gathers signal from the world. A tool can already scan and summarize much of that signal. The level-defining decision is which question deserves the company’s attention.
Research and Data science
The live border, and the one no research ladder is drawn against. Synthesis moved into the shared column in about two years, which is why the middle of the ladder is where this hurts.
Research and Product
Research shares methods with Design. Its harder boundary with Product is who decides when the evidence is sufficient and what happens next.
What survives
Models can cluster forty transcripts quickly and consistently. Synthesis remains a method step. The level comes from choosing the question, judging what the evidence can carry and changing the decision that follows.
Three things survive it, and they are what I would build a research ladder on.
Question quality. The most consequential act in research is deciding what to ask, and it happens before any method is chosen. A researcher who reliably converts a vague stakeholder anxiety into a question worth twelve interviews is doing something no tool does, because the tool has no stake in the answer.
Evidentiary honesty. Knowing what a finding can carry. Six interviews do not support a roadmap decision and everybody in the room wants them to. The researcher who says so, when saying so kills a plan somebody has already staffed, is operating at a level the borrowed ladder has no row for.
Belief change. Not reports delivered, not studies run, not stakeholders served. Which of the company's beliefs is different because this person worked here, and does it still hold. This is the hardest to measure and it is the only one that separates the top two levels.
The six axes, in research’s language
Three of the six levels. The row a borrowed ladder cannot produce is the third one down.
The borrowed design ladder had no ancestor for opinions held. Adding it makes a change in belief something calibration can inspect.
Why the field never settled one
Research teams are often small enough to borrow a standard and large enough for the mistake to surface late. At six researchers, the design ladder with its nouns changed can pass. At twelve, the missing rows show up as a promotion decision nobody can explain, usually years after the document was copied.
The other reason is that writing one down means committing to belief change as the measure, and that measure is uncomfortable for exactly the people who would have to write it. It makes a researcher's level partly dependent on whether the organization listened, which feels unfair, and is unfair, and is also true.
Score the opportunity and the response together. If the company withholds access or ignores sound evidence, record that failure in the review of the leader who controls the decision. Do not quietly turn it into a ceiling for the researcher.
Take the research argument into calibration
Research needs its own expectations inside a company-wide standard.
The Leveling workbook gives Research the same six-level core as Engineering, Product, Design and Marketing while preserving the decisions only researchers keep. The discipline guide, evidence log and calibration scorecard make a sparse public precedent into a system a team can actually run.
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 design, engineering, product 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