There is a conversation I have had enough times that I can hear the shape of it coming. Someone who is the best in the building at a hard, specific thing sits down and asks what comes next for her. She is not bored and she is not angling for anything. She wants to know whether staying good at the work has a future in this company, and she has already looked at the pay grid, so she has a guess. Whoever she asks tells her about the technical track. Principal engineer, or fellow, depending on the year and the employer. She nods and goes back to her desk. Some months later she is running a team and has not touched the work in a while.

Nobody in that story made a bad decision. She followed the money that was actually on offer, and her boss filled a management opening with the most capable person available. The company recorded a promotion. What made the outcome inevitable was a decision taken years earlier by people who never met her, when somebody set the bands and nobody asked what the spread between them would teach. The loss lands on no scorecard anywhere, which is a large part of why it keeps happening.

Forty three payrolls, read all the way through

In April, four economists put evidence under an argument that has run on anecdote for decades. Nicola Bianchi, Lydia Cao, Benjamin Friedrich and Kieu-Trang Nguyen pulled personnel records from 43 medium sized and large firms, rebuilt the complete reporting chains and organizational layers so they could see exactly who sat where, and then compared what the two tracks were paid. Their paper is a working paper rather than a refereed publication, so treat it as strong early evidence rather than a settled finding.1

Individual contributors turn up throughout those hierarchies, well up into them, so the expert track is real in the sense that people occupy it. It is also paid substantially less than management at comparable positions, and the penalty grows toward the top.1 The higher you go, the more it costs a person to stay in the craft.

Two other findings travel with that one. Crossing from an IC role into a manager role brings a large pay gain, which tells your people what the ladder rewards more plainly than any competency framework ever will.1 And although pay inside each track does not vary much across demographic groups, women and workers from minority groups are disproportionately on the IC side, more so at the senior ranks, which makes track assignment one of the routes by which a company ends up with an internal pay gap nobody chose and nobody can account for.1

What the money is actually buying

Luis Garicano wrote the model that explains the mechanism back in 2000. A knowledge hierarchy forms because matching a problem to someone who can solve it is expensive. Production workers learn the common problems, the harder and rarer ones get passed upward to specialists, and communication lets one specialist's knowledge cover a lot of people.2

Read that as a pay design and the shape of the gap makes sense. What scales with each layer you climb is reach. A principal engineer may hold the hardest problem in the company in her head and still apply that knowledge once, in one place. A director applies thinner knowledge across forty people. Depth is paid for the problem it solves. Reach collects on everything sitting underneath it, and there is more sitting underneath it every layer up.

The document says one thing, the payroll says another

Everything we have built in delivery over the last twenty five years assumes the opposite arrangement. Cross functional teams work because the person who understands the system best is in the room while the decision about that system is being made. Product ownership works when someone with real technical judgment can say what a request will cost before it gets committed. Then the compensation structure reaches in and removes precisely that person, one promotion at a time, and everybody involved describes it as career growth.

The organization compounds it on the way across. I have written before about how onboarding support thins out as the role gets more consequential, and the IC to manager crossing is where that thinning starts. She was excellent at a job with fast feedback and clear right answers, and she is now doing a job with slow feedback and no right answers.

I will leave alone the separate problem of managers measured in ways that reward holding on to their best people, which I argued a couple of weeks ago and do not need to relitigate. The two structures stack. One of them makes leaving the craft the only route to a bigger number. The other makes it hard to move anywhere at all without leaving the company.

Your people read the career ladder document once, on the day HR published it. They read the payroll every two weeks, and that is the version they believe.

Go and look at your own numbers

The most useful thing in the paper for anyone running an organization is that the penalty is not uniform. The estimates sort firms into distinct types by the shape of their hierarchy and the makeup of their workforce, and those types produce differently sized IC penalties.1 The size of your own gap is therefore a design choice your company has already made, probably without ever putting it on an agenda.

Checking it takes an afternoon and one person from compensation. Match your ICs against your managers at the same level, then look hard at the top two levels rather than the blended average, because the average is where this hides. If the spread is wide up there, your technical track is a set of titles with nothing behind them, and the people you built it for worked that out a long time ago.

The harder question is what you would have to give up to fix it. Paying a principal engineer like a director means conceding that a person can be worth that much with nobody reporting to them, and most compensation committees have never been asked to say that out loud. So what would it cost you to make the expert side of your ladder real, and what is it costing you now that it is not?