Understand job matching

Job matching is the process Pave uses to map employee records from connected HR platforms to Pave's job catalog of standardized job families and levels. Accurate job matching is what makes compensation benchmarks reliable. When every company's roles are mapped to the same catalog, you can compare compensation across thousands of companies with confidence.

Pave automates job matching using machine learning models trained on over one million employee records. This eliminates the manual matching process required by traditional compensation surveys, where analysts map roles to survey job codes by hand, often introducing inconsistencies across companies.

What signals Pave analyzes

Pave's matching algorithm analyzes 20+ signals across four categories for each employee record. No single signal controls the outcome, which prevents individual data errors from distorting results.

Job data: Job title, job family, department, function, internal level, job code, and job description

Organizational context: Reporting relationships, hierarchy depth, span of control (direct and indirect reports), and team composition

Compensation patterns: Base salary, bonus eligibility, equity grants, and pay mix. These help distinguish between levels. For example, compensation patterns differ significantly between an entry-level and a senior individual contributor, even when titles are ambiguous.

Company context: Company size, stage, ownership type, industry, location, and organizational complexity

How the matching process works

Job matching follows four steps:

1. Data ingestion: Pave receives employee data from connected HRIS, ATS, and EMS platforms. Data flows to Pave continuously from these integrations.
2. Multi-signal analysis: For each employee record, the algorithm runs multiple classification tasks against Pave's job catalog, analyzing the signals described above to determine the best-fit job family and level.
3. Consensus voting: The results of all classification tasks are combined into a single match. This consensus approach means that no individual classification task can override the others, reducing the impact of noisy or incomplete data.
4. Aggregation: Matched records are de-identified and aggregated into the compensation database that powers Market Data benchmarks.

How Pave validates job matches

Pave's matching models are built on over one million employee records from thousands of companies. The algorithm has been validated by both internal and external compensation professionals to confirm that results are consistent with how experienced practitioners would match roles manually.

The algorithm is reviewed and refined on an ongoing basis. Because Pave collects data continuously, the models improve over time as more records flow through the system.