Table of Contents:
- How does Pave calculate benchmarks?
- How does Pave validate its calculation methods?
- Which dataset is better to use, raw or calculated?
- How do consistency labels apply to calculated benchmarks?
- Why does a calculated benchmark have low consistency? Does this mean Pave isn’t confident in it’s calculation?
- How can I tell whether a benchmark is calculated?
- Will I still need to analyze and manipulate the calculated data when using the data?
- Does this mean Pave’s raw data isn't reliable?
- Can I view only the raw (non-calculated) benchmarks?
- What compensation types does Pave provide Calculated Benchmarks for?
How does Pave calculate benchmarks?
Compensation data has a number of strong patterns. US jobs pay more than equivalent roles in the UK. Higher levels get paid more than lower ones. Our team of data scientists leverage machine learning to identify these patterns across our entire dataset and then combine this deep understanding of the trends we see across the market with the raw data in and around a specific benchmark to produce a calculated benchmark in cases where robust market data is not available.
How does Pave validate its calculation methods?
Prior to launch, our algorithm was tested extensively against real market data and by industry experts at multiple compensation consulting firms to validate outputs. In many ways, our approach emulates the manual data “smoothing” (or normalization) process already used by most compensation professionals. Moving forward, we are continuing to review and refine our algorithm in partnership with customers, consulting firms, and other industry experts.
Which dataset is better to use, raw or calculated?
Pave is committed to providing customers with the most accurate and usable compensation data set, which includes a combination of both "raw" and "calculated" benchmarks. We analyze all of our data to provide customers with the best dataset available, so you don't need to choose between different datasets.
How do consistency labels apply to calculated benchmarks?
Our consistency labels apply to both raw and calculated benchmarks and take into account various factors, including sample size in order to give a view into how well this data represents the market via a margin of error.
Sample size is just one of the indicators of how consistent the market data is. Two benchmarks may have the same number of employee records, but if one is distributed tightly around the median and the other is widely distributed from the median, they may have different consistency labels. Learn more about Consistency Labels here.
Why does a calculated benchmark have low consistency? Does this mean Pave isn’t confident in it’s calculation?
Consistency labels are intended to give customers context into the underlying data and how much variation there is in pay for that benchmark. When there is high variation in how the market compensates a given role, the consistency level of the benchmark will be lower in order to provide customers with that context. This is true regardless of whether the benchmark is raw or calculated.
In the case of a calculated benchmark being labeled as "Low Consistency," we're providing you with a benchmark and piece of data to reference, but we do want to be transparent that there is a higher level of variation in how the market compensates this role.
How can I tell whether a benchmark is calculated?
We want customers to deeply understand the data they are referencing. As we combine both raw and calculated benchmarks into our benchmarks, customers will be able to distinguish between them both in-app and in export. Please see the article above to understand how to differentiate between the types of benchmarks.
Will I still need to analyze and manipulate the calculated data when using the data?
The introduction of Calculated Benchmarks is intended to make Pave market data easier to use than typical survey data. However, customers typically have their own unique job definition, level structure, and compensation philosophy that may require additional manipulation and translation on top of calculated benchmarks.
Example: If a customer has 8 IC levels as opposed to Pave's P1-P6 levels, the customer will need to translate and map market data between the levels. If a customer places outsized value on their P6 levels and wants to reward them with equity beyond the market norm, then they will need to make adjustments to the market data to reflect this.
Does this mean Pave’s raw data isn't reliable?
No! Calculated benchmarks are intended to supplement raw survey data and give customers data that is easier to use and interpret than raw data in isolation. This does not mean that Pave’s raw data is unreliable.
There is more to the consistency and accuracy of a benchmark than just sample size, particularly in the world of equity data. When comparing cash and equity benchmarks, we found that it in an equity benchmark, it typically takes 10 times the number of samples of a base salary benchmark to reach the same level of consistency. This is mainly due to the widely different equity programs across companies and the high-level of variation in equity grant values across employees.
Even if a benchmark has a sizable number of raw data points, the data can still include a high degree or variation and require manipulation and normalization on the customer's end to make this data usable.
Can I view only the raw (non-calculated) benchmarks?
Calculated benchmarks are introduced alongside raw benchmarks to give customers access to the best combination of data we have. Customers will still see raw benchmarks when searching, and they will be supplemented with calculated benchmarks. We do not currently offer the ability to view only raw benchmarks.
What compensation types does Pave provide Calculated Benchmarks for?
Currently, Pave offers Calculated Benchmarks for the following compensation types: Base Salary, New Hire Equity, Refresh Equity, and Unvested Equity.
Pave does not currently support Calculated Benchmarks for Total Equity, Variable Pay, or Total Cash.