I built CompData as pay data and ready-made pay structures for people setting compensation without a full team behind them. A normal market-data purchase often leaves the buyer holding a percentile they cannot explain. The uncertainty does not disappear because a vendor leaves it off the page. It moves into the buyer’s judgment. I wanted the estimate to carry its own range and source, so the person using it can say what the evidence supports. I also wanted the work after the estimate to be buyable. Many teams need ranges, bands and grades more than they need another table or a consulting project. CompData therefore sells those structures at posted prices, and its templated systems label what came from the model and what came from compensation policy. The website serves the person making the decision; the API and agent tools let the same work fit into software. The aim is straightforward: make a sound pay answer and a usable structure available to teams that do not have a compensation department.
Who it is for
CompData is for HR leaders, founders, compensation managers and consultants who need a reliable pay number without a compensation team behind them. They use it to price jobs, compare modeled market pay with observed occupation wages, check minimum-wage requirements, and buy ranges, bands and grades that can be put to work.
The problem
Survey data often reaches a smaller team as a locked percentile table: useful enough to quote, weak when someone asks what changed by job level, function or region. The buyer still has to judge how certain the estimate is and turn raw figures into ranges, bands and grades. Without a compensation analyst, they can spend heavily on data and still be left explaining why a job should pay that amount, how reliable the answer is, and which parts of the final structure came from evidence or policy.
What I built
CompData is a catalog of pay data and ready-made pay structures for people who set compensation. It sells modeled market pay, observed occupation wages, minimum-wage data, and structures with ranges, bands and grades at posted prices. The modeled market-pay dataset is $490. Its model combines three commercial survey sources; on held-out log-pay data, it explains 89% of the variation (R² 0.8907). Each estimate’s uncertainty comes from the model’s own residual error. Buyers can use the website, a REST API for pay systems and structure design, or three MCP tools that let an AI assistant compose a structure. After checkout, buyers collect their purchase from a CompData access page reached through a bookmarkable private link.
What is new in it
- Each market estimate includes an interval derived from the model’s residual error. The range is evidence about how much the data can support, rather than a separate judgment added for presentation.
- The product works in a browser, through a REST API, or through three MCP tools for AI assistants. The API covers pay-system retrieval and pay-structure design.
- Industry scope selects the functions included in a structure. It does not change the pay estimate because the model contains no industry differential, keeping that limitation visible.
- Market-anchored ranges, bands and grades are sold as products at posted prices. Smaller teams can buy the analyst’s output directly instead of receiving raw data that still needs a structure.
- With a company identifier, CompData can reuse shared industry, size and lifecycle context across products. The buyer does not have to create the same company profile again.
Where it stands
CompData is built so a smaller team can get an explainable market estimate and buy a working pay structure without assembling either from scratch. It is live with posted prices for observed wages, modeled market pay, minimum-wage data and pay structures. The planned consortium would add anonymized peer contributions under minimum group-size protections to narrow estimates as each niche grows.