The value equation

The principal-issues thesis shows the workflow running — structure first, calculation downstream, a value stack that ends in segmented opportunity. This piece is about the part of that machine that reasons about pay. Compensation looks like a lookup: find the market number, apply it. It is actually a reconciliation of three different kinds of value, and when the reconciliation is wrong, the damage does not show up on the pay line. It shows up as an attraction, activation, and attrition problem you spend a year misdiagnosing.

---

Here is the problem, stated the way it actually presents. Take ten people doing the same job in the same company and you will often find ten different rates of pay. Widen the lens to a hundred people in that job across a hundred companies and you will find a hundred numbers. We say we believe in equal pay for equal work, and then we look at the data and equal pay for equal work is almost nowhere to be found.

The usual reaction is to treat this as a measurement failure — we just have not found the real market value yet. That reaction is the mistake. There is no single correct market value waiting to be discovered. A rate of pay is not a fact you look up. It is the output of a decision that reconciles three different kinds of value, each of which answers a different question, and each of which can be right or wrong on its own terms.

Most compensation trouble is a failure to keep the three kinds of value separate long enough to get each one right.

Three kinds of value

The foundation of the framework is the recognition that "value" in compensation is not one thing. There are three perspectives, and a defensible pay decision has reconciled all three.

External value is the comparison to the outside — the competitive position you pay for a given job in a job market. The trap here is the phrase "we pay market." A market is not a point; it is an array. You define a job, collect what different employers pay for it, and line the values up from highest to lowest. The median describes the middle of that list — which means paying "the market" puts you at the 50th percentile, competitive to the half of the market making less than the median and not competitive to the half making more. "We pay market" can be a real statement of position or it can be gobbledegook, and the only way to tell is to have actually purchased the data and located yourself in the array. External value also has no single "market." Narrow the list by occupation, by geography, by the set of employers you actually compete with for talent, and each narrowing gives you a different answer. The relevant market is the set of opportunities genuinely available to someone in that job — which makes defining the market a decision, not a lookup.

Internal value is the comparison inside the organization — the worth of a job relative to the other jobs, set without regard to who happens to hold it. It answers two questions: how do these jobs differ, and how much does each contribute to what the organization is trying to do? Internal value is built through a sequence — job analysis, then evaluation, then a job structure that lays out the relationships among jobs. And internal value carries the most under-appreciated finding in the whole field. Internal equity has two parts, not one. The first is whether people doing substantially similar work are in fact paid similarly, with every difference explained by job-related criteria. The second is whether employees feel their pay is equitable. You can get the first exactly right and still fail, because the classic research on pay is unambiguous: the perception of fairness predicts attitudes and behavior better than the absolute dollar amount does. People do not evaluate their pay in isolation — they evaluate it relative to others. An organization that has done careful, defensible work on actual equity and never made that work legible to its people has solved half the problem and may reap none of the benefit.

Personal value is the comparison each individual makes on their own terms — the utility a particular person assigns to this company, this job, and this pay package given their needs, preferences, circumstances, and history. Personal value is what determines which jobs someone pursues, which offers they accept, and how long they stay before they start answering the recruiter's calls. Two people can be paid identically and value the arrangement completely differently, and the difference is not noise — it is the thing that actually moves attraction and retention.

Getting these three forms of value right more often than your competitors do is, more or less, the point. A flaw in the value equation can look inconsequential at present scale and still be the crack that widens as more weight is applied.

Objectives come first, mechanics come second

You do not pay people a particular way "just because." A compensation system exists to produce specific results, and the discipline is to start with the end in mind. The common objectives are familiar — attract, motivate, and retain the talent that drives performance; control labor cost; influence the thoughts, feelings, and behaviors of employees; reward differentiated contribution with differentiated pay; maintain visible and internal consistency; comply with law. Most organizations would nod at all of them. The work is in prioritizing them, because there are never enough dollars to optimize for all of them at once, and the priority depends on the problem you actually have to solve right now.

This matters because the objective dictates the mechanics, and a mismatch is expensive. Suppose you decide the system should reward exemplary behavior. The theory underneath that choice is doing four jobs at once: it makes the rewarded person more likely to repeat the behavior, makes peers more likely to imitate it, lifts the strong performer's pay out of a competitor's range so they stay, and leaves weaker performers below market so that if they are poached you get a fresh draw. To actually collect those four effects you need a wide enough pay range to separate average from excellent, a way to define and measure the behaviors, and — critically — a way to communicate the difference. Get the range too narrow or leave the logic uncommunicated and the individual and their peers never notice the differential, and you lose the first two effects entirely and may miss all four. The objective was sound; the mechanics did not match it; the money was spent and bought nothing.

Between the concepts and the objectives sit the techniques — job analysis, market surveys, job evaluation, merit planning, range design, the rest of the technology of compensation decision-making. Techniques are the verbs that tie the concepts to the objectives. They are not ends in themselves, and the moment the mechanics of compensation become an end in themselves is the moment the system starts drifting away from the results it was built to produce.

Know, plan, do — and then do it again

Because the world, the organization, and the people in it change daily, a value equation is never solved once. The framework runs as a recurring cycle. You know where you stand — competitive analysis against the market, an analysis of where compensation spending is actually going, an equity analysis against your own stated standard. You plan — you make recommendations, and because you have agreed the objectives up front and you are arguing from data, you can seek genuine agreement rather than win an argument. You do — you decide, you document the decision, and you communicate it. Then you go back to the beginning, at least once a year, because the inputs have already moved.

The throughline of the cycle is measurement. What gets measured gets managed; what does not, simply will not. The objectives are not slogans — they are the standard against which the whole system is evaluated, which means every one of them has to be measurable or it is not worth the breath spent stating it. If the objective is to attract and retain highly competent staff, and your skilled people are leaving for higher pay while your less-skilled people stay, the system is not performing as intended, and the measurement is what tells you so before the trend hardens into "worst of your kind."

The errant value equation

This is where the compensation model meets the value stack, and it is the most important sentence in the whole framework. Stated as a question:

To what degree are attraction, activation, and attrition problems a function of an errant value equation?

A note on the cost of getting this wrong. It is tempting to attach a headline number — organizations waste N% of their compensation spend — and the field is full of such numbers, almost none of them defensible. This piece will not offer one. What more than twenty years of practice does support, as an observation rather than a statistic, is a pattern: a compensation system examined carefully has reliably turned out to contain waste, and that waste has reliably cost more than correcting it would.

The reason this pattern holds is structural, not a matter of competence. Absent active review and management, systems drift toward entropy, and compensation — continuously perturbed by hires, exits, promotions, market moves, and one-off exceptions — drifts faster than most. Organizations operate inside a set of constraints, and over time those constraints come to feel natural and fixed, simply the way things are. They are not. Each was set by a decision, at some point, for a reason that may no longer apply, and most are more repositionable than they have come to seem. Waste is what accumulates when no one is positioned to notice that the constraints can be moved. The corollary is that, in a period when the analysis is cheap and available, the absence of awareness is itself a choice — and not the best one available. The magnitude in any specific organization is nonetheless real and estimable, but only against that organization's actual facts — which is precisely the work a diagnostic does, and precisely what a generic claim cannot do honestly.

There is no general answer to the errant-equation question, and pretending there is one is its own error. Pay matters more or less depending on knowable conditions — general job satisfaction, how much a given person weights pay, the gap to the market that person could actually command, where they sit in the value the role can produce. Pay dissatisfaction is related to turnover, but the relationship is moderated; the structure of pay differentials drives the desire to leave far more than it drives day-to-day motivation, which is a necessary-but-not-sufficient condition for performance. So the honest position is not "pay fixes attrition" or "pay does not matter." It is that the degree to which an errant pay value is driving an outcome you care about varies with the facts, and the facts change — which means you need an analytics that brings these things together and the vigilance to keep looking.

The errant value equation also produces problems that have nothing to do with being underpaid in aggregate. Consider an organization that, following clean market data, pays managerial roles well above expert individual-contributor roles. The market justifies it. But the differential induces the best engineers to leave the bench for a management track where their particular talent is worth less to the firm — a defensible pay decision generating an errant incentive that drains value from exactly the roles the strategy depends on. This is why some research-driven organizations deliberately build parallel technical and managerial tracks with equivalent rewards: not because the market told them to, but because the market-justified default produced an outcome they did not intend. You have to watch for the errant incentives your own pay program creates.

This is the link back to the value stack the platform runs on. Employee Lifetime Value, activation, Net Activated Value, and segmented opportunity are the measures; the value equation is the theory of pay those measures reason with. When the command center asks where the next dollar of pay buys the most retained value, it is asking, segment by segment, where the value equation is most errant relative to an outcome the organization cares about — and where correcting it pays back the most. The value stack is the instrument. The three kinds of value, reconciled or errant, are what the instrument is reading.

What this does and does not claim

The compensation command center referenced in the principal-issues thesis runs this theory on structured synthetic data — a generated population built to exhibit the patterns real compensation data exhibits, carrying none of any real organization's confidential information. Pay varies in that data the way pay actually varies (level dominates, then function, then geography, then tenure), and exit hazard is driven by the same forces that drive it in the world, so the analyses surface structure that means something. But every number on that surface is a property of a generated population. None of it is a client result, and none should be read as a measured return on any real intervention. The framework tells you where the value equation is most likely errant and what correcting it would address. What that correction returns, in dollars, in a specific real organization — that is a separate, sourced claim, and the demonstration does not make it.

The mechanics of compensation should never be allowed to become an end in themselves. The questions worth asking of any technique are the plain ones: what does this do for us, by what observable mechanism, under what conditions would it stop working, and how would we know if it were working at all. The value equation is just the discipline of holding three kinds of value distinct, measuring each against an objective you chose on purpose, and staying honest about how much — right now, for these people — the equation is actually driving the outcomes you care about.

---

Companion to the principal-issues thesis. The framework summarized here is Mike West's Business-Value Compensation-Analytics Model; the attraction/activation/attrition framing is the Triple-A program from People Analytics for Dummies (West, 2018). The value stack — Employee Lifetime Value, activation, Net Activated Value, and segmented opportunity — is documented in the thesis and runs live on the compensation command center on structured synthetic data.