01 · THE RANKING
What actually predicts base salary
We fit base salary against nine characteristics of the seat, then removed each in turn to see how much unique explanatory power it held. The bars show that unique contribution — what is lost when the factor is taken away and the other eight remain.
Unique explanatory power, each factor with the other eight held
The headcount trap
Employee headcount looks like a real driver on its own — 8.7% of the variation — and collapses to 0.2% once revenue and span of control are known. It is a proxy for scale, not a price signal.
Span of control carries roughly 51× its unique information; company revenue about 28×. Benchmarking a role to “a 20,000-person company” is benchmarking to the wrong number.
OLS on the natural log of base salary, n=235, R² = 0.47, HC3 robust standard errors. These are associations in observational survey data, not causal effects. A factor ranking high carries information about pay; it does not follow that changing it would change pay.
02 · SPAN OF CONTROL
The strongest single factor
+53%
Median base for executives managing 100+ people, against those managing 1–5
9.1%
Unique explanatory power of span of control — the highest of any factor
5.0%
Company revenue, the second strongest
COMPENSATION BY NUMBER OF REPORTS MANAGED
SVPs are not paid more because they are SVPs. They are paid more because SVP seats in this sample tend to be larger jobs. Where they are not larger — as in the sub-$1B market — the premium disappears entirely.
03 · WHAT TURNS OUT NOT TO MATTER
Reporting line, headcount, work model
01
Reporting line
Ranks seventh of nine. No reporting relationship carries a meaningful adjusted difference. Descriptively it looks worse still — the reports-to-CEO group posts the lowest median base of any reporting line at $255K, because those seats skew to smaller organizations.
02
Company headcount
Eighth of nine. Meaningful on its own (8.7%), near zero once revenue and span are known (0.2%).
03
Work arrangement
Last of nine. The raw on-site premium of $25K does not survive controls — the adjusted difference is inside the noise. On-site executives work at bigger companies.
04 · WHERE RAW AND ADJUSTED DISAGREE
Two views, published together
Descriptive medians and scope-adjusted models answer different questions, and on several findings they disagree. Our editorial standard for this edition is to publish both and explain the difference rather than picking whichever produces the stronger headline.
Are women paid more in comparable seats?
DESCRIPTIVE
Pooled, +3%. Private sector only, +5%. By level the direction flips: women lead by $37.5K at VP, level at SVP/EVP, $20K behind in the C-suite.
SCOPE-ADJUSTED
Between +9% and +12% on base depending on specification. Significant in a sparse model, not significant under the full one (p = 0.20). A 13–18% premium does not reproduce at any specification we tested.
WHAT WE PUBLISH
We publish the level split and the sensitivity, state that the estimate is positive but not robust, and attach no causal story to it.
Do public companies pay less for the same seat?
DESCRIPTIVE
No visible difference. Within revenue band, public and late-stage private sit within a few percent in all three bands.
SCOPE-ADJUSTED
About -11.0% (p = 0.06) — marginal.
WHAT WE PUBLISH
Lead with the pay mix, which is unambiguous: late-stage private grants run 0.93× base against 0.58× at public companies.
Is a competing offer worth about 10%?
DESCRIPTIVE
No difference at all. Median base is $300K with none, one, or several offers.
SCOPE-ADJUSTED
+6.8% on base (p = 0.27) and +10.5% on total cash (p = 0.17). Negotiating shows a similar-sized, similarly insignificant effect on the same test.
WHAT WE PUBLISH
We cannot separate leverage from the act of asking at this sample size, so we make no recommendation and show both point estimates.
Has AI scope been repriced?
DESCRIPTIVE
Executives whose scope expanded significantly report a lower median base than those reporting no change — though that group is small (n=18) and skews senior.
SCOPE-ADJUSTED
+8.6% on base (p = 0.15) and +12.1% on total cash (p = 0.11). Directionally positive, not reliable.
WHAT WE PUBLISH
AI scope is not yet consistently priced. Neither “rewarded” nor “penalised” is supported.