rouka · compensation intelligenceBenchmark set 6.3 · October 20265,226 roles · 143 regions · 10 sectors
rouka

Methodology

How a number becomes a figure you can act on.

Compensation in the private market is reported inconsistently and rarely published. We do not resolve that by averaging it away. Each figure carries the state it was produced in.

VERIFIED

Facts we can publish outright: counts of our own dataset, meaning roles, regions, sources and the benchmark version, plus figures from genuinely public sources such as BLS or SEC filings. The source and its year appear beneath the figure.

MODELED

Anything derived from licensed survey data, plus percentile, rarity class, upside to median and any regional adjustment. This is every compensation band on the platform. The line names the surveys and the adjustment, because naming a source is attribution, not republication.

ESTIMATED

No role-specific grounding. A sector median stands in, labeled so you treat it as directional rather than as a figure to negotiate against.

Regional adjustment

National survey bands are adjusted to the metro where the work happens. The adjustment is named in the source line rather than folded silently into the number.

National band adjusted to metro
Aspen, CO
+45%
Vail, CO
+40%
Manhattan, NY
+40%
Greenwich, CT
+30%
National Average (US)
baseline

A selected sample. 143 regions carry a multiplier, and the full set is applied to every band rather than shown here.

MODELED
rouka methodology, regional multipliers, version 6.3, October 2026

What we will not do

Average across surveys that measure different populations to produce a smoother band.

Present a modeled figure without saying it was modeled.

Show a role-specific band where only a sector median exists.

Publish a figure whose source cannot be named.