A Number, Not an Argument

The question sounds like it should have an answer. It rarely does.

A compensation analyst is leveling a role. Is the "Senior Engineer" at the company down the street the same job as her "Staff Engineer," or a notch below? She asks three people who do this for a living and gets three answers, each defended with real expertise and none of them checkable. So the leveling decision — which sets the pay band, which sets the offer, which sets what a person earns for years — comes down to whose judgment carries the room that afternoon. It is not a measurement. It is an argument, and it is an argument every single time, because there is no shared reference that would let it be anything else.

That's the subject of this essay: the move that turns "match this job" from an argument into a number. It has been made before, in a different industry, by a small company that understood something most people miss about where value lives.

"Match this color" used to be an argument too

Before 1963, getting a color right across a supply chain was a negotiation. A designer in one city picked a red; a printer in another mixed ink to match it by eye; the two disagreed, by mail, slowly. Then Lawrence Herbert, who had bought a small commercial color-card company, had the insight that built an empire: the problem wasn't the ink. It was that there was no shared language for color. He created the Pantone Matching System — a number for every color, plus a calibrated reference book so the number meant the same thing in New York and Shenzhen.¹ "PMS 185" stopped being a conversation and became a fact.

Notice what Pantone chose to own. Not the ink — ink is a commodity, mixed by anyone. Pantone owned the reference: the standardized language both sides trusted, and the calibrated guides that made the number reliable. Ink makers paid to be Pantone-certified. The margin and the moat lived in the shared vocabulary, not the material underneath. Sixty years later the whole industry still specs in a number a printing shop invented, because being the reference everyone maps into is a more durable position than making the best of anything.

Jobs have the same problem, and no Pantone

Work is exactly the pre-1963 color industry. Every organization titles, levels, and scopes its jobs its own way; the surveys, the HRIS, the applicant tracker, and the job description all disagree about what a given role even is. "Match this job" is a negotiation conducted by experts, and like color-matching by eye, it produces confident answers that don't reconcile. The cost isn't abstract — it's the leveling decision, the pay band, the equity audit, the workforce plan, all resting on a comparison nobody can actually compute.

The reason it stays this way is that no one occupies the position Pantone occupied: a neutral, shared reference that the side which defines a job and the side which consumes the data both map into. There are reference systems, but they sit at the extremes — the public occupational codes are open but far too coarse to carry a pay decision, and the private survey frameworks are rich but locked, each its own incompatible dialect. The middle — granular enough to decide with, open enough to compute on, neutral enough to belong to no one — is empty.

Coordinates: the part that turns names into numbers

Here is where Pantone went further than most people remember, and where the analogy stops being cute and starts being a blueprint. Pantone didn't stop at giving each color a name. It gave each color coordinates — a position in a measurable space — so that the difference between two colors became a computed number, not an opinion. Designers call that difference Delta-E: how far apart two colors are, objectively, on a scale.² The name made colors referable. The coordinates made them comparable.

A job can carry coordinates too. Place every role in a measurable space and "how similar are these two roles" gets a real answer — a distance, between identical and unrelated — instead of an expert's shrug. The discipline that makes it trustworthy is the discipline Pantone modeled and most analytics ignore:

Keep the spaces separate. A color can be described in more than one system for more than one purpose; a job is richer, so you hold several coordinate spaces side by side and refuse to mash them into one number — where the role sits structurally (level and function), what work it involves (skills and tasks), where it lands on the pay surface, what its description means. Two roles can be close in pay and far in content; collapsing that into a single "similarity score" throws away the only information worth having.
Let it say "I don't know." Every comparison returns one of three verdicts — confident, needs review, or no confident match — against a tolerance, exactly the way a Pantone match is either close enough or it isn't. Below the bar, it refuses to guess. A system that always returns an answer is a system that fabricates, and a fabricated match is worse than an honest blank.
Freeze it in editions. A coordinate belongs to a dated, versioned edition, so a number computed today still means the same thing next quarter. When the method improves, you publish a new edition; you don't silently move the old ones out from under everyone who built on them. Pantone reissues its swatch books for the same reason — a standard that quietly drifts is not a standard.

That is the difference between a catalog of job titles and a measurable space. One lets you look a role up. The other lets you compute with it — level it, dedupe it, price it, defend it — with a number instead of an argument.

Own the system, not the ink

The strategic core follows directly. The valuable position is not being the system where jobs are stored and transacted — that's the ink, the commodity, and it's already occupied. The valuable position is being the reference every other system maps into, and owning the crosswalk: the translation that lets any taxonomy — a survey's, a government's, a single company's homegrown ladder — resolve into one shared space. Owning the translation between everyone's vocabulary is worth more than owning any one vocabulary, and it is the thing a competitor cannot cheaply copy, because by the time they want it, everyone is already speaking through you.

The fence that keeps them in

And now the part that, to me, decides the game.

The private survey houses copyright their job structures and enforce it. That is their right, and from inside their business it looks like prudence — the leveling framework is the product, so you protect it. But look at what the protection does at the level of the whole market. It fences each house permanently inside its own dialect, which means the incompatibility that creates the need for a neutral reference is not a temporary accident anyone is racing to fix — it is load-bearing to their business models and therefore permanent. And it guarantees that none of them can ever become the shared standard, because a reference everyone trusts cannot be one competitor's proprietary, enforced asset. The hilltop stays empty not in spite of how well it's defended on every side, but because of it. It is Chesterton's fence inverted: the wall they built to keep rivals out of their structure is the same wall that keeps them out of the neutral position none of them can occupy.

Here is the technical move that turns that structural opening into something buildable — and it is the part the enforcement can't reach. You do not need their codes to locate their jobs. Their copyright protects their expression — their titles, their codes, the specific structure they authored — and we don't reproduce it. But the position of a job in a measurable space is computed by universal math that their structure merely happens to fit, the same way any point fits into a coordinate system. We never warehouse their architecture; we compute where the work sits, independently, in ours. The reference is ours by construction, because we built it out of universal math rather than out of anyone's protected library.

What we keep, and what we don't

That principle decides exactly what is stored where, and it's worth stating plainly because it's the line that keeps the whole thing clean.

What is owned, permanently and across everyone, is the universal structure, the math, and the open mappings — none of which is anyone else's expression. What is not stored for permanent cross-client reuse is a licensed source's copyrighted material — its titles, its codes, its relationships. A crosswalk that depends on a proprietary, licensed dataset — and the client's own adjustments and feedback on it — lives at that client's level, gated by that client's license. The customer who has paid for a given survey gets their mapping between that survey and the canon, and their tuning of it; the next customer, who hasn't, gets none of it. We resell no one's structure. The universal coordinates travel; the licensed joins stay home.

It's the same discipline twice: own the neutral reference, never take or resell what isn't yours.

Never pull an Adobe

There's a cautionary chapter in Pantone's own story, and it's the guardrail on all of this. In 2022 Pantone moved its color libraries behind a separate subscription; Adobe removed built-in support, and files that used those colors began opening with the affected colors replaced by black, behind a dialog telling users to pay to get their own work back.³ Sixty years of accumulated goodwill — the trust that made the standard a standard — spent in a news cycle.

The lesson, as a rule: you may own the vocabulary everyone depends on, but you monetize the services and the data flywheel around it — never the right to name a job, and never a customer's own mapped work, held hostage when they stop paying. A standard is only a standard while it is trusted, and the fastest way to lose the position is to behave like a tollbooth on work people have already done. The reference everyone maps into has to stay dependable, or it stops being the reference.

Measurable, not arguable

Go back to the analyst with the role to level. In the world we have, she takes a vote among experts and hopes. In the world this builds, she gets a position in a shared space, a distance to comparable roles, a verdict that's honest when the data is thin, and an answer she can show her work on — the same answer anyone else querying the same reference would get. The leveling stops being a negotiation about whose judgment wins and becomes a measurement anyone can check.

Pantone proved that the durable fortune in a fragmented market isn't in making the thing everyone makes — it's in becoming the reference everyone agrees to point at, and giving that reference coordinates so agreement becomes arithmetic. Work is a far larger fragmented market than color, it has no such reference, and the incumbents are busy fencing themselves out of the position. The number that ends the argument is sitting there, unclaimed. The move is to compute it.

This is a cross-portfolio thesis in the People Analyst program — that "job" is the undefended join key of the economy of work, and that giving it a shared reference with computable coordinates turns leveling, similarity, and pricing from expert opinion into measurement. It extends The Join Key (the join-key thesis) and shares its posture with the Measurement Meets AI essays: the honest "no confident match" verdict is the same refuse-to-fabricate discipline that runs through Themes Aren't Evidence and You Can't Compute With a PDF. This is thesis, not announcement — it makes no product claims; the architecture it describes is the work.

Lawrence Herbert created the Pantone Matching System (PMS) in 1963, standardizing the specification of color — a number plus a calibrated reference guide so a color means the same thing across vendors and geographies — rather than manufacturing the ink. Pantone licensed the standard and sold the (periodically reissued) reference guides. ↩
The CIE L*a*b* (CIELAB) color space assigns each color coordinates in a perceptually-organized space; the distance between two colors in that space ("Delta-E") expresses their difference as a computed number rather than a subjective judgment. The analogy here: a canonical job profile carrying coordinates across several deliberately-separate spaces (structural, content, pay, semantic), with similarity computed as a distance. ↩
Beginning with updates after August 2022, Adobe removed bundled Pantone color libraries following a licensing change; files using the removed colors opened with those colors replaced by black and an error dialog, and continued access required a separate Pantone Connect subscription (about $15/month). Widely reported, late 2022 — the canonical example of squandering the trust that makes an owned standard valuable. ↩