What data Pave Agent can see

What data does Pave Agent use?

Pave Agent answers questions from two kinds of data: your company's data in Pave, and Pave's own market data. It also reads files you upload to it.

It does not search the web and does not pull in outside sources. Knowing which data an answer came from explains most surprises, so this article lists what is available and what is not.

Your company's data in Pave

Pave Agent can see the data your company has loaded into Pave:

• Employee compensation and organizational structure, including who reports to whom. This is what makes questions about team rollups and span of control possible.
• Your job architecture: job families, levels, and the jobs you have defined.
• Pay bands and band sets from Market Pricing, including pay zones and which employees match to which band.
• Compensation history, including past raises and promotions.
• Merit cycles from Compensation Planning, both active and finalized, including the performance ratings and other data loaded into a cycle.
• Third-party survey data you have uploaded into Market Pricing, along with your survey mappings.
• Your Guidance and your Skills. These shape how it answers rather than what it analyzes. See the articles on setting up Guidance and setting up Skills.

Pave market data

Pave Agent can see Pave's market data alongside your own:

• Market data benchmarks: salary, bonus, and equity by role, level, location, industry, and company stage.
• Your peer groups, as configured for your company.
• Market insights: how companies operate, described below.
• Pave Data Lab: Pave's published research and community discussion, described below.
• What makes up a benchmark, including which companies participate in Pave's dataset and how its dimensions are defined.
• Public job postings, described below.

For how Pave builds and cuts its benchmarks, see Understand Market Data methodology.

Market insights

Market insights covers how companies run their organizations, not just what they pay. Use it when your question is about practice rather than pay level: how often peers give raises, how deep their org charts run, how much of the option pool they burn in a year.

Raises and merit cycles. Raise rates as a share of eligible employees and of total employees, raise percentages broken out by merit, promotion, and equity, promotion rates, and equity refresh participation rates.

Turnover and tenure. Annual turnover rates and average tenure.

Organizational structure. The number of reporting layers, direct reports and total reports per manager, and how headcount is distributed across an organization.

Equity. Equity burn rates, new hire and ongoing participation rates, grant counts, and vesting practices including duration, interval, cliff, and structure.

Geo differentials. Location ratios for base salary, total cash, and annual bonus.

Most of these can be broken out by job family, level, location, industry, company stage, and headcount size, though which of those apply varies by metric. You can combine up to two breakouts in one question. If a breakout is not available for the metric you asked about, Pave Agent tells you rather than guessing.

Three things worth knowing about how these are built:

• They are computed from Pave's live data across participating companies, so they reflect current practice rather than a point in time.
• They are aggregated. You get market-level figures, never an individual company or employee.
• Every cut has to draw on a minimum number of companies. A narrow slice may come back empty rather than give you an unreliable number, and a thin one is flagged as low confidence. You can ask how many companies and employees sit behind any figure, and that count comes back with the result.

Pave Data Lab

Pave Data Lab is Pave's published research alongside what compensation leaders are discussing with each other. Pave Agent can search it:

• Pave's own compensation analysis and research.
• Pulse surveys and polls.
• Community discussions.

"Search Pave Data Lab for conversations about how I should think about off-cycle promo equity grants."

This searches published material and tells you what has been written, rather than calculating a figure from data. Use it when you want to know how other practitioners approach a question, not what a number is.

Public job postings

Pave collects and curates public job postings so you can look at what companies are advertising, which is useful for roles that do not map cleanly to a survey job family.

For each posting, Pave Agent can see the job title, the posting date, whether the listing is still live, the location (city, state, country, metro area, and whether the role is remote), and the pay the employer disclosed, which may be a salary range, an hourly rate, or total cash.

Two things to keep in mind:

• Job postings are a curated dataset, not a live search of the web. Pave Agent is not browsing the internet when it answers these questions.
• The mapping from a posting to a Pave job family is inferred, not official. Pave Agent labels it as inferred whenever it uses one. Treat posted pay as a separate signal from Pave market benchmarks rather than a substitute for them.

Files you upload

Pave Agent can read files you upload to it, either into your file library or as an attachment on a single message. See the article on uploading files and exporting results.

What Pave Agent cannot see

• Anything that is not in Pave. This is the most common reason an answer comes back empty. If your bonus targets are not loaded, Pave Agent cannot analyze them, and if a system is not integrated with Pave, its data is not available.
• The open web. It cannot look up a company's careers page, read a news article, or check a figure online.
• Public and third-party sources Pave does not license, including SEC filings, government labor statistics, and cost of living indexes.

If you are not sure whether something is loaded, ask. Pave Agent can tell you what it has for a given team or job.

What Pave Agent will not change

Pave Agent does not change your compensation data. It cannot save a range into Market Pricing, edit a pay band, change a merit cycle, or update an employee record. Everything it produces is either an answer in the conversation or a file you download.

If you want to act on an answer, make the change in Pave yourself.

What your data needs for useful answers

Pave Agent is only as good as what it has to work with. Answers improve when:

• Your job architecture reflects the jobs you actually have, rather than placeholder or stub jobs.
• Your pay bands are live rather than left in draft.
• Your employees are matched to bands.
• You have a market data source you trust, whether that is Pave benchmarks or survey data you have uploaded.

If answers look wrong, these are the first things to check. See the article on troubleshooting Pave Agent.