Methodology preprint · Version 0.1 draft · Updated July 2026
Methodology for International AI Compensation Data.
United States, United Kingdom, and Canada. A peer-reviewable preprint documenting Orbyt Intelligence's cross-jurisdictional reconciliation across seven government data sources. Published CC BY 4.0.
Abstract.
The AI compensation data market segments into three product categories: enterprise compensation survey products (Mercer, Aon Radford, WTW) at $25,000 to $100,000 per seat per year; consumer freemium products (Levels.fyi, Glassdoor, Payscale) with opaque methodology and US-dominant coverage; and payroll-feed products (Pave, OpenComp) at $50,000+ per company with individual-record live data. None occupies the developer-API segment with a transparent methodology contract across multiple jurisdictions.
This paper documents the methodology behind Orbyt Intelligence's Tier 1 international coverage. Tier 1 reconciles seven government- sourced data inputs into per-country weighted estimates updated as each government source publishes. Reconciliation weights are locked per country and disclosed on every estimate. Currency normalization uses ECB- backed daily rates. Cross-country sanity bounds quarantine implausible values. Every estimate carries its methodology version, source breakdown, sample size, confidence interval, and disagreement flag in the locked response envelope.
Source weights, locked.
Per-country weights sum to 1.0 independently. The reconciliation engine never blends across countries. A US measurement and a UK measurement for the same role and city slug land in separate tuples and reconcile with their own weight sets.
| Country | Source | Weight | Cadence |
|---|---|---|---|
| US | BLS OES | 0.57 | Annual (May) |
| US | DOL H-1B LCA | 0.43 | Quarterly |
| UK | ONS ASHE | 0.60 | Annual (October) |
| UK | HMRC PAYE RTI | 0.20 | Monthly |
| UK | Skilled Worker Visa | 0.20 | On update |
| CA | ESDC Open Government | 0.60 | Annual (November) |
| CA | StatCan WDS (planned) | 0.40 | Not yet ingested |
Contents.
- Introduction. The state of AI labor data. The gap in the developer-API segment.
- Data sources. Per-source documentation for all 8 Tier 1 sources with permanent URLs.
- Role taxonomy and occupation crosswalk. The honest disclosure that no official concordance exists between US SOC 2018 and UK SOC 2020 or NOC 2021.
- Reconciliation methodology. Per-country weighted average. Renormalization. Disagreement detection at the locked 25% threshold. Cross-country sanity bounds.
- Currency normalization. The fx_rates table. Frankfurter API. Weekend FX handling. FX outage policy.
- Sample size and confidence. The disclosure floor. Per-country practicality. The HMRC RTI fan-out honest disclosure. Confidence interval calculation.
- Limitations and known gaps. Crosswalk uncertainty. Geographic granularity. Selection bias. Structural shifts within the survey window.
- Open invitations. The Role Taxonomy is yours. The Methodology is open for critique. The API is priced like Stripe.
- References.
License + citation.
Published under CC BY 4.0. You may adopt, modify, and redistribute with attribution. No restriction on commercial use.
Bartak, J. (2026). Methodology for International AI Compensation Data: United States, United Kingdom, and Canada. Orbyt Intelligence. https://www.orbytjobs.ai/orbyt-intelligence/methodology/international