What Are You Paying For?
Ten people, one job, ten different numbers. The reflex is to call it a measurement failure and go hunting for the real market value. There is no real market value. Pay is a reconciliation of three kinds of value — and when the reconciliation is wrong, the bill arrives somewhere you weren't looking.
Pull the comp file for any job with ten people in it and you will almost certainly find ten different rates of pay. Not wildly different, usually, but different — and different in ways nobody in the room can fully explain on the spot. Widen the lens to the same job across a hundred companies and you get a hundred numbers. We say we believe in equal pay for equal work. Then we look at the data, and equal pay for equal work turns out to be almost nowhere.
The standard move, when someone notices this, is to treat it as a problem of not-yet-knowing. We just haven't found the real market value yet. Buy a better survey, match the jobs more carefully, and the true number will resolve out of the noise like a photograph developing.
I want to argue that this move is the original sin of compensation, and that almost everything that goes wrong downstream — the good people who leave, the raises that buy nothing, the fairness complaints from a team you actually paid carefully — starts here. There is no real market value waiting to be found. A rate of pay is not a fact you look up. It is the output of a decision, and the decision is reconciling three completely different kinds of value at once.
"We pay market"
Start with the sentence everyone says: we pay market. It sounds like a position. Usually it's gobbledegook, and the reason is worth slowing down on, because it's the whole game in miniature.
A market is not a point. It's a list. You define a job, you collect what a set of employers actually pay for it, and you line those numbers up from highest to lowest. The middle of that list is the median — the 50th percentile. So when someone says "we pay market," what they generally mean, if they mean anything, is that they target the median. Which means they are competitive to the half of the market making less than the median, and not competitive to the half making more. "We pay market" is, on inspection, a decision to be uncompetitive to the better-paid half of your own talent pool, stated as if it were a neutral fact.
And there isn't one market, either. Line up every employer in the world and you get one answer. Narrow to your country, your industry, your city, the specific handful of companies you actually lose people to and poach people from, and each narrowing gives you a different number. The relevant market is the set of opportunities genuinely available to the person in the chair. Which of those lists you choose is itself a judgment call — a decision dressed up as a lookup.
This is external value: where you sit, on purpose, in a list you chose. It's the kind of value everyone fixates on, because it's the one you can buy a chart for. It is also only one of three, and the other two are where the real trouble lives.
The fairness you can prove, and the fairness people feel
The second kind is internal value — what a job is worth relative to the other jobs in your own building, set without regard to who happens to hold it. And buried inside it is the single most counterintuitive finding in the whole field, the one I've watched sink careful, well-meaning compensation programs.
Internal fairness has two parts, not one.
The first part is the one analysts work on: are people doing substantially the same work in fact paid about the same, with every difference traceable to something job-related — performance, scope, experience? You can grind on this for a year and get it genuinely, defensibly right.
The second part is whether people feel it's fair. And the classic research on pay — this goes back decades — is blunt about which part wins. The perception of fairness predicts how people feel and what they do better than the actual dollar amount does. People don't grade their pay in isolation. They grade it against the person in the next chair, against the friend at the other company, against what they figured they'd be making by now. Satisfaction with pay is a function of how much you get, how much you think others get, and how much you thought you should get — and two of those three terms are perceptions.
Sit with the implication. You can solve the fairness you can prove and still fail completely, because you never made the work legible to the people it was for. I have seen organizations do months of honest, technically sound work on their pay structure and reap almost none of the benefit, because the work stayed inside the comp team and the felt experience on the floor never changed. Transparency isn't a virtue you bolt on at the end. It's the second half of the actual deliverable. And you can't be transparent if you don't have anything you'd be proud to show — which is the quiet reason a lot of pay programs stay opaque.
What individuals actually weigh
The third kind is personal value — the utility a specific human being assigns to this company, this job, this package, given their needs, their history, what they're optimizing for this year. Personal value is what decides which jobs someone chases, which offers they sign, and how long they stay before they start taking the recruiter's calls. Two people paid identically can value the arrangement completely differently, and that difference isn't noise to be averaged away. It's the thing that actually moves whether they come and whether they stay.
Three kinds of value. External: where you sit in a market you chose. Internal: whether the structure holds together and whether people believe it does. Personal: what it's worth to the individual on their own terms. A defensible pay number has reconciled all three. Most pay trouble is a failure to hold them apart long enough to get each one right.
The errant equation
Here's why this isn't an academic taxonomy. When the value equation is wrong, the damage almost never shows up on the line labeled "compensation." It shows up as three problems you'll spend a year misdiagnosing: a hard time attracting people, a workforce that's present but not activated, and the wrong people walking out the door.
I'm not going to put a percentage on what that costs you, because I don't have one I can defend, and this field is crowded with numbers nobody can defend. Here is what I can tell you after more than twenty years of doing this work: I have never once gone looking for waste in a compensation system and failed to find it, and the waste has always cost more than the fix. Every time. That's not a statistic — it's a pattern I've watched hold across organizations of every size and shape.
And it isn't an indictment of anyone's management. It's closer to a law. Without active review, every system drifts toward entropy, and compensation is no exception — it is one of the most active, most continuously perturbed systems an organization has. Organizations run on the resources they have, inside a set of constraints, and over time those constraints stop looking like choices and start looking like the weather: fixed, natural, simply how things are. They are nothing of the kind. Every one of them was set by someone, at some point, for some reason that may no longer hold — and almost none of them are as immovable as they have come to feel. The waste isn't a failure of effort. It's what accumulates when no one is positioned to notice that the walls can be moved. Which is why, today, when seeing has finally become cheap, declining to look is itself a decision. It just isn't the best one on offer.
What I can't tell you in the abstract is how much it's costing you. That number is real and it is knowable, but only against your specific facts — which is the kind of thing a short, honest diagnostic answers far better than a headline ever will.
The question I keep coming back to, the one that organizes the whole practice, is this: to what degree are your attraction, activation, and attrition problems a function of an errant value equation?
The honest answer is that it depends, and that the dependence is knowable. Pay matters more or less depending on conditions you can actually measure — how satisfied someone is overall, how much they personally weight money, how big a gap they could close by leaving, how much of the value their role can produce they're actually delivering. Pay dissatisfaction does drive turnover, but the link runs through general satisfaction and through how much the person cares about pay in the first place. So "pay fixes retention" is wrong and "pay doesn't matter" is wrong. What's right is narrower and more useful: for these people, in this segment, right now, here is how much the errant equation is driving the outcome you care about — and here is where fixing it pays back the most.
And the errant equation produces damage even when nobody is underpaid in aggregate. Take the organization that, following perfectly clean market data, pays managers well above its senior individual contributors. The market justifies every number. But the differential quietly tells your best engineers that the way up is out of the work they're brilliant at and into a management job where their specific talent is worth less to you. A defensible pay decision, manufacturing an errant incentive that drains value from exactly the roles your strategy depends on. The fix — parallel technical and management tracks with equivalent rewards — isn't something the market data would ever suggest. It's something you do because the market-justified default produced an outcome you didn't intend. You have to watch the incentives your own pay program creates, because the market won't watch them for you.
Measuring it
None of this is manageable as philosophy. It becomes manageable the moment you decide to measure it — what gets measured gets managed, and what doesn't, simply won't. You don't solve the value equation once; you run a loop. Know where you stand: where you sit against the market, where the money is actually going, where the structure does and doesn't hold up against your own standard. Plan: make the call, argue it from data, get real agreement because you settled the objectives up front. Do: decide, write down why, and communicate it — that's the second half of fairness again. Then start over, because the market moved, the strategy moved, and the people moved while you were deciding.
This is the part of our compensation platform I find people most want and least expect: not the number, but the loop. The value stack the platform runs on — lifetime value, activation, the value you're actually capturing, and the opportunity left on the table by segment — is the instrument. The three kinds of value, reconciled or errant, are what the instrument is reading. When the system asks where the next dollar of pay buys the most retained value, it's asking, segment by segment, where your value equation is most wrong relative to something you care about, and where setting it right returns the most.
So: what are you paying for? Not a market rate. You're paying to reconcile three kinds of value well enough, often enough, that the people you want to keep stay activated and the bill never quietly arrives somewhere you weren't looking. The mechanics of compensation — the surveys, the ranges, the merit matrices — are just the tools. The moment they become the point, you've stopped paying for value and started paying for the machinery. The questions worth asking are the plain ones, of every technique you run: what does this actually do for us, by what mechanism, under what conditions would it stop working — and how would we know if it were working at all?