The Title Is Not the Job
Every workforce analysis starts by rebuilding the same missing object — a frame for what a job actually is. None of the rebuilds reconcile, which is why none of the answers do either.
A recruiter, a compensation analyst, and a finance director are looking at the same person in the same seat, and each of them is holding a different job.
The recruiter has a title, because that is what the requisition carried. The comp analyst has a survey code, because that is what the benchmark required, and the survey code is not the title — it is the closest thing the vendor's taxonomy had. The finance director has a cost center and a level, and the level came out of a reorganization in 2019 that nobody has fully explained since. When the three of them disagree about whether this person is paid correctly, they will spend the meeting arguing about the answer. The disagreement is upstream of the answer. They are not describing the same job.
This is the most boring problem in workforce analytics and it is underneath almost every expensive mistake in it.
The rebuild tax
O.K. Watch what happens when a company asks a genuinely reasonable question — are we paying our engineers competitively? — and follow the first two weeks of work.
Nobody starts on the pay question. They start on the frame. Which of these forty-one engineering titles are the same job? Is a Senior Engineer here a Senior Engineer at the companies in the benchmark, or is it a level below? Do we count the managers? What about the three people whose titles were negotiated at hire and describe nothing? Every one of those is a modeling decision, every decision is made under time pressure by whoever is holding the spreadsheet, and none of them get written down as decisions. They get written down as the data.
Then the next question arrives — where is attrition worst? — and the frame gets rebuilt, differently, by someone else. Now the two analyses cannot be laid on top of each other, because the objects underneath them are not the same objects. Nobody notices, because both analyses are internally coherent and both produce a chart.
Call it the rebuild tax. It is paid on every analysis, it is invisible on every invoice, and it compounds — because each rebuild becomes the precedent the next one inherits. The tax is not the analyst's fault. It is the predictable cost of the object being missing.
The reason it stays missing is that building it is unglamorous. A job taxonomy has no demo. It does not appear in a board deck. It is the plumbing, and plumbing gets funded after it fails.
What the frame has to do
So we built it. The shape it had to take matters, because the naive version — a big list of titles — is what most organizations already have, and it does not work.
A frame has to make a job addressable: one stable key that survives a title change, a reorg, and a market cut. Ours is a coordinate — family, focus, level — which currently resolves 14,948 job profiles across 533 families and 44 superfunctions.
Levels are where most frames quietly break, so it's worth being specific. Ours runs 25 of them across four parallel tracks — professional, support, management, executive — rather than one ladder everybody climbs. That structure isn't decoration. A principal engineer and a director are not the same rung wearing different hats; they are different tracks at comparable altitude, and a frame that flattens them into a single ordinal will mis-price both, every cycle, in opposite directions. The ladder shape is itself a claim about how work is organized, and most taxonomies make it by accident.
Those are counts, not accomplishments; a taxonomy's size is the least interesting thing about it. What matters is that the coordinate is the same object every consumer joins on. Pay joins to it. Skills join to it. Postings join to it. The guides join to it.
It also has to be crosswalked outward, or it is just another private dialect. The coordinate resolves to SOC, which is what lets public occupational data and public wage data reach an internal role at all.
And it has to be honest about the modeled layer, which is where most job-and-pay products quietly stop being honest. The pay figures on a profile are a model's output, and the surface says so in the place a reader will see it — a twelve-month outlook labeled as a directional projection from labor-market pressure, not as observed history, with medians priced at a national base. That labeling is not modesty. It is the difference between a number a reader can interrogate and a number they have to take on faith, and the market-data industry has spent thirty years selling the second kind.
The gap you can only see from a frame
Here is the part I did not anticipate, and it is the argument for building the boring object.
Once a role is a stable coordinate, you can attach two independent descriptions of it and compare them.
The first comes from a field's own canon — the books a discipline wrote to explain itself, decomposed into what a role is for. The second comes from the market: what postings ask for, what the wage data pays, which tasks and skills the occupational data weights as important. Both describe the same coordinate. They are produced by completely different processes, and neither one knows about the other.
They do not agree. The distance between them is a real, measurable object, and it is the most interesting thing on the page — what the discipline believes the job exists to accomplish, set against what the market is actually rewarding someone for doing.
I can't yet tell you what a big gap means for a given role — whether the field is ahead of the market, the market has repriced the work, or the books are describing a job that no longer exists. That's the research. But you cannot even ask the question until both descriptions hang off the same coordinate, and until now nobody had put them there.
A generic job site cannot produce that number. Not because it lacks engineers, but because it only has one of the two descriptions. It has the market side — postings, salaries, titles — and no canon side at all, because assembling the canon side means reading the field's literature and turning it into structured claims about the role. The gap is the thing you get for free once you have both, and you cannot get it any other way.
The first job we could compute it for honestly was Compensation Analyst, because it was the one role where all three links were real on the same day: a guide grounded in eighteen books, a family-to-SOC crosswalk that resolved, and SOC-keyed market data underneath. One job. That is not a product launch; it is an existence proof. The interesting work is that the same three links now exist for a great many more.
Why this is the tent pole
I have been asked, more than once and reasonably, why so much effort goes into a job taxonomy when the sellable things are the analyses on top of it.
Because the analyses on top of it are only as good as the object underneath, and every one of them is currently paying the rebuild tax. Pay benchmarking needs the frame. Leveling needs the frame. Attrition-by-role, internal mobility, span-and-layer work, the whole workforce-planning family — each of them either has a shared frame or is quietly inventing a private one. Build it once, properly, addressably, and the rebuild tax goes to zero for everything downstream.
There is a second reason, less noble and worth admitting. A public reference surface for jobs has a long tail that behaves like nothing else in the portfolio: people search for their own role, constantly, forever. That is the most durable inbound path we have, and it happens to be produced by the same object that makes the analytics correct. It is rare for the marketing incentive and the engineering incentive to point at the same artifact. When they do, build the artifact.
To be fair to the incumbents, the reason job data stayed bad is not stupidity — it is that the frame is expensive to build, impossible to demo, and pays off only in things that are hard to attribute to it. That is exactly the profile of infrastructure. It also means it stays unbuilt until somebody is willing to do unglamorous work for a while, which is a low bar that almost nobody clears.
The title is what the requisition happened to say. The job is what the coordinate resolves to.
Which brings it back to the recruiter, the comp analyst, and the finance director, still in the meeting, still arguing about the number. They are not disagreeing about pay. They are disagreeing about what a job is — and not one of them knows that is the argument they are having.