Understand Calculated Benchmarks

Calculated Benchmarks are compensation benchmarks generated by Pave's machine learning models. They fill gaps where raw benchmark data has insufficient sample size, so you can get usable market data for roles, levels, locations, and company stages that would otherwise return no results.

Raw benchmarks aggregate compensation data pulled directly from HRIS records across Pave's dataset. Calculated Benchmarks use the same underlying data but apply regression models trained on patterns across the full dataset (level progressions, geographic pay differentials, job family relationships, and company stage trends) to produce estimates where raw data alone is not sufficient.

All standard Market Data filters work with Calculated Benchmarks. You can filter by location, company stage, industry, and company ownership the same way you would with raw benchmarks.

How Pave selects which benchmark to display

For every benchmark you view, Pave computes both a raw and a calculated benchmark and compares them. Pave displays whichever has a higher consistency level. This means:

• When raw data is robust, you see the raw benchmark
• When raw data has a small sample or high variance, the calculated benchmark may replace it
• When no raw data exists for a given combination, the calculated benchmark fills the gap

This selection happens automatically. You do not choose between raw and calculated data, and you cannot filter to view only raw benchmarks.

How to identify Calculated Benchmarks

You can tell whether a benchmark is raw or calculated in several ways:

In the table view: Calculated benchmark values display with a dashed underline. Hovering over a calculated value shows a tooltip explaining that the benchmark was computed using Pave's models.

In consistency labels: The consistency label for each benchmark indicates whether it is raw or calculated, along with the sample size behind the data. For more on how consistency labels work, see Understand consistency labels.

In exports: For Market Data Pro customers exporting data to Excel, the report includes a "Data Method" column that identifies whether each benchmark is raw or calculated.

Supported compensation types

Calculated Benchmarks are available across all compensation types in Market Data:

• Base salary
• Total cash
• Variable pay
• New hire equity
• Refresh equity
• Unvested equity
• Total equity

Calculated Benchmarks cover the same percentiles as raw benchmarks (10th, 25th, 40th, 50th, 60th, 75th, and 90th).

How Pave validates Calculated Benchmarks

Pave's models are trained on over one million employee records and tested against held-out raw data across thousands of role, level, and location combinations. The methodology has been reviewed by compensation consulting firms to validate that outputs are consistent with how experienced compensation professionals interpret and normalize market data.

FAQ

Can I view only raw benchmarks?

No. Pave automatically displays whichever benchmark (raw or calculated) has a higher consistency level for each data point. You cannot filter to view only raw data.

Does using Calculated Benchmarks mean Pave's raw data is unreliable?

No. Calculated Benchmarks supplement raw data. When raw data has a sufficient sample size and low variance, Pave displays the raw benchmark. Calculated Benchmarks address the specific case where raw data is too sparse or too variable to produce a reliable number on its own.