Skip to main content
All CollectionsMarket Data
How Figures ensures benchmark quality
How Figures ensures benchmark quality
Updated today

Overview

At Figures, we've implemented several mechanisms to ensure our benchmark data remains accurate, reliable, and representative. This article explains how we maintain data quality through freshness monitoring, balanced sampling, and regular updates.

Data collection

We receive compensation data through two main channels:

  1. HRIS Integration

    • Automatic daily synchronization with your HR system

    • Data is continuously updated

    • Provides real-time reflection of market changes

  2. Spreadsheet Import

    • Manual data imports with varying refresh frequencies

    • Refresh rate depends on individual client update schedules

Data freshness

Since spreadsheet updates can occur at irregular intervals, we've implemented a data degradation system to prevent outdated information from impacting our benchmark accuracy. This ensures that older data has a decreasing influence on market insights over time.

How freshness works

  • All data points are tracked based on their last update date

  • Weight degradation begins after one year

  • Each month beyond the one-year mark reduces the data's weight by 15%

  • Data is removed from the benchmark after 18 months without updates

Weight degradation timeline

Time since update

Data weight

0-12 months

100%

13 months

85%

14 months

70%

15 months

55%

16 months

40%

17 months

25%

18 months

10%

18+ months

Removed

Balanced sample distribution

Why balance matters

Large companies with numerous employees can significantly impact benchmark calculations. For example, if a tech giant with 1000 engineers contributes to a sample alongside smaller companies with 50 engineers each, without adjustment, the large company's compensation practices would disproportionately influence the benchmark results. This could create a skewed view of the market that doesn't accurately represent diverse company practices.

Our balancing approach

To prevent this, we've implemented a sophisticated weighting system that:

  • Maintains data point value from all companies

  • Prevents market dominance by larger organisations

  • Ensures diverse company practices are represented

  • Creates a more accurate reflection of the overall market

Weight thresholds and adjustments

Sample size rules

We apply different thresholds based on the number of companies in a sample:

  • 3 companies: No single company can represent more than 34% of the sample

  • 4+ companies: No single company can represent more than 25% of the sample

Adjustment process

When a company exceeds these thresholds, we follow a three-step process:

  1. Initial Assessment

    • Calculate each company's initial weight based on their data points

    • Identify companies exceeding thresholds

  2. Weight adjustment

    • Take the square root of the exceeding company's percentage

    • This mathematical approach reduces the impact while preserving significance

  3. Weight normalisation

    • Redistribute the remaining weight among other companies

    • Ensure all weights still sum to 100%

    • Apply new weights to individual employee data points

Monthly release cycle (starting 2025)

Starting in 2025, we're implementing a monthly release cycle for our benchmark data instead of real-time updates. This change allows us to enhance data quality through thorough validation while providing more stable and reliable market insights.

The monthly process explained

Each month follows a structured process:

  • Throughout the month:
    We continue collecting data as usual. Your HRIS integrations keep syncing daily, and spreadsheet imports are processed as they arrive. However, this new data doesn't immediately impact the benchmark calculations.

  • Validation:
    Our team conducts thorough quality checks on all new data:

    • We analyse market trends and identify any unusual patterns

    • We verify the consistency of new compensation data

    • We ensure all calculations (freshness weights, company balancing) are working correctly

  • Monthly release:
    Usually once a month all validated data from the previous month becomes active in the benchmarkAll users see the same, consistent market data

Did this answer your question?