Intelligence Glossary

A query parameter that lets you request a salary estimate as of a specific historical date. Every Intelligence endpoint that returns numbers accepts as_of and replays the same calculation against the snapshot from that date. Allowed range starts at 2024-07-01 (the first stable snapshot) and ends today plus one day.

Every salary estimate returns three numbers: low (25th percentile), median (50th percentile), and high (75th percentile). Together they describe the middle 50 percent of comparable roles. A tight spread between low and high means the market is converged on a number. A wide spread means meaningful variation exists across companies, levels, or geographies.

A numeric score from 0 to 1 indicating how much we trust this specific estimate. Higher confidence comes from larger sample size, recent data, and agreement across sources. A 0.95 confidence means the estimate is well-supported. A 0.40 confidence means the estimate is directional, not definitive. Always inspect this before quoting a number.

A boolean field on every estimate that flags when our sources materially disagree about the true salary range. True means at least one source is reporting values more than 20 percent outside the consensus. The estimate is still returned, but the disagreement is surfaced so you can decide whether to trust it.

A YYYY.N string (currently 2026.2) that identifies the exact calculation rules used to produce an estimate. Stable releases carry a clean version. Variants like 2026.2-synthetic-cagr or 2026.2-community add a suffix that honestly discloses what kind of methodology produced the number. Every response carries this tag.

Percentile markers in a salary distribution. P50 is the median (half of comparable roles pay less, half pay more). P25 and P75 bracket the middle 50 percent. P90 and P99 describe the long tail of high outliers. The Intelligence API returns P25, P50, and P75 by default; percentile breakdowns are available via the /range endpoint.

The count of underlying data points that contributed to a given estimate. Returned on every numeric response. A sample of 500 is well-supported; a sample of 12 is thin and the confidence level will reflect that. Sample size is the primary input to the confidence level score.

An array on every estimate response showing what fraction of the underlying sample came from each upstream source (BLS OES, H-1B LCA). Lets you see whether an estimate is balanced across sources or dominated by a single feed. Helpful for sanity-checking outliers.

The U.S. Bureau of Labor Statistics Occupational Employment and Wage Statistics program. Publishes annual wage data for 800 occupations across every metropolitan area, sourced from employer surveys. The most authoritative public salary dataset in the country. Refreshed annually each spring; current snapshot is 2024 data published 2025.

Salary disclosures filed under state pay-transparency laws (California SB 1162, New York Pay Transparency Law, Colorado Equal Pay Act, Washington, etc.). Every public job posting in covered states must include a salary range. Our connector for this source is built and wired, and it holds zero rows, so no Orbyt figure is derived from a disclosed posting today.

Two leading applicant tracking systems used by mid-market and growth-stage companies. We optionally ingest public job posting feeds from companies that publish through these platforms to capture compensation ranges that haven't yet been filed under pay-transparency laws. Gated behind a legal review per RFC-003.

Labor Condition Applications filed by employers sponsoring H-1B visa workers, published quarterly by the Department of Labor's Office of Foreign Labor Certification. Each filing includes the employer's prevailing wage offer, the specific job title, the work location, and the level. The most granular salary dataset in the country for technology roles.

A third-party service that scrapes and normalizes pay-transparency-required job postings across all 50 U.S. states, providing a single feed of every posting that legally must include a salary range. Operators choose which aggregator to wire into ingestion via the PAY_TRANSPARENCY_DATA_URL environment variable. No aggregator is wired in production today, so this source contributes nothing to any Orbyt figure.

Every quarter we freeze the full Intelligence dataset into a versioned snapshot. The snapshot is the basis for the as_of time-travel parameter and the source of all historical lookups. Snapshots make calculations reproducible: the same query with the same as_of always returns the same number regardless of when you call it.

The U.S. Securities and Exchange Commission's Electronic Data Gathering, Analysis, and Retrieval system. Public-company filings include named executive officer compensation, equity grant schedules, and CEO-pay-ratio disclosures. Orbyt has an EDGAR connector built but not yet active; the current dataset does not include EDGAR-sourced data.

The fixed annual cash compensation a role pays, paid out across a year. Base salary excludes bonus, equity, stock, and benefits. Most negotiations focus on base because it is the foundation for every other compensation component. The Intelligence API's default salary number is base.

The range a company has internally approved for a given role and level (e.g., Senior Engineer in San Francisco: $180K to $240K). Bands are how companies enforce internal pay equity and limit ad-hoc negotiation. Knowing the band lets you negotiate within it without overshooting.

Equity is ownership in the company, granted as stock units. Restricted Stock Units (RSUs) at public companies vest into shares on a schedule. Incentive Stock Options (ISOs) at private companies give you the right to buy shares at a fixed strike price. Both can dwarf base salary at high-growth companies; both can be worth zero if the company fails.

On-Target Earnings: base salary plus the variable cash component you receive when you hit 100 percent of your annual targets. Common in sales, account management, and customer success roles. The Intelligence API returns base and variable separately so you can compute OTE yourself based on your expected hit rate.

The sum of every cash and equity value a role pays in a year: base salary plus annual bonus plus equity vested in that year plus sign-on bonus amortized over the vest schedule plus benefits valued at market rate. TC is the apples-to-apples number to compare offers, not base salary alone.

The portion of compensation paid based on performance, separate from base salary. May be annual (target percentage of base, paid yearly) or quota-based (sales commission paid quarterly). Variable comp is at-risk: hitting 100 percent of target pays the full amount; missing target pays proportionally less. Some roles pay above 100 percent for over-performance.

An engine that calculates how compensation shifts across a company's funding stage. Seven canonical stages from pre-seed through public. Surfaces the equity-vs-cash tradeoff: early-stage offers tilt cash-light equity-heavy, growth-stage and public roles invert. Pro tier and above.

Five derived signals per company: hiring intent (open roles vs trailing average), compensation pressure (pay-band shifts), attrition (departure rate), AI maturity (AI-role share), and transparency (public-disclosure quality). Available filtered by signal strength. Ultra tier. Beta status.

Four metrics per role-and-city: posting volume trend, time-to-fill estimate, salary drift week-over-week, and competitive intensity (postings per qualified candidate). Outputs a Hot/Warm/Cool/Cold classification per market segment. Pro tier and above.

A canonical list of 598 software, AI, and product roles mapped into 8 role families and 5 seniority bands. Includes BLS SOC code mapping so estimates can be cross-referenced against public wage data. The foundation for every other Intelligence engine. Available on every tier including Free, via the /engines/ai-role-taxonomy endpoint.

Estimates how long a technical skill retains market value before depreciating, modeled as one of three decay shapes: exponential (fast obsolescence), logistic (S-curve adoption then plateau), or plateau (lifelong relevance). Helps you decide whether a skill is worth investing in given your career horizon. Pro tier and above; preview status.

The complete record of how a single salary estimate was produced: which raw data points contributed, what weights were applied, which methodology version computed the number, and when the snapshot was frozen. Available via the /api/v1/intelligence/lineage/{data_point_id} endpoint. Pro tier and above; intelligence:lineage scope required.

A unique identifier returned on every Intelligence estimate. Pasting this id into the /lineage endpoint returns the full provenance trail: which sources contributed, which methodology version was applied, when the snapshot was frozen, and which corrections (if any) have been applied. end-to-end traceability for every number.

A scheduled job that fetches the latest data from each registered upstream source (BLS, H-1B, DOL, etc.), parses it into our canonical row format, reconciles it against the existing snapshot, quarantines any anomalies, and commits clean rows to the active snapshot. Runs daily at 04:00 UTC; surfaced via /api/cron/ingestion-run.

A staging area for inbound data points that fail automatic validation: salary outside plausible range for the role, location not in our canonical city list, employer not previously seen, or values that materially disagree with the existing dataset. Quarantined rows wait for human review and do not affect estimates until released.

The step that compares newly-ingested rows against the existing snapshot, identifies overlaps and contradictions, and decides how to merge them. The reconciliation policy is documented in RFC-003. Disagreements between trusted sources increment the disagreement flag on downstream estimates rather than silently picking one source.