For Compensation and Benefits Analysts ·
What you'll accomplish
This guide sets up Claude to extract and organize salary benchmarks from dense survey PDFs. Instead of spending 3–4 hours manually scanning a 400-page Mercer or Radford report for the 20 job families you care about, you upload the relevant sections and get a clean comparison table in 20 minutes.
What you'll need
Go to {{tool:Claude.url}} and log in to your account. Click New Chat to start a fresh conversation. You'll start fresh each time you work with a new survey, as Claude doesn't retain files between conversations.
What you should see: A clean chat window with a paperclip or attachment icon in the message input box.
Click the paperclip icon (or drag and drop) to attach the survey PDF. If the full survey is too large, upload only the relevant section (e.g., the Technology & Engineering section of Radford). Claude can handle files up to 100MB in Claude Pro.
What you should see: The filename appears above the message box indicating the file is attached and processing. Troubleshooting: If the file is too large, split it into sections using Adobe Acrobat (File → Split Document) or a free PDF splitter. Upload only the pages containing the job families you need.
In the message box (with the file attached), type your prompt. Be specific about exactly what data you want extracted. Here's a template:
This is a [survey name] salary survey for [year]. Please extract all benchmark data for these job families:
- [Job Family 1] at levels [level range]
- [Job Family 2] at levels [level range]
For each benchmark, extract: Job Family name, Level, P25 Base Salary, P50 Base Salary, P75 Base Salary, P50 Total Cash Compensation, and sample size (n=).
Format the output as a table. If a data point is missing or suppressed due to small sample size, note that in the cell.
What you should see: Claude reads the document and begins extracting. This may take 30–60 seconds for large documents.
When Claude returns the table, spot-check 2–3 data points against the original PDF. Navigate to the source page in the survey and confirm the numbers match. This verification step takes 5 minutes but is essential. Survey PDFs can have complex formatting that occasionally causes misreads.
Troubleshooting: If numbers don't match, tell Claude: "The P50 base for [Job Family] Level [X] shows $[what Claude said] but the survey shows $[actual number]. Please re-read page [X] of the survey and re-extract that row." Claude will usually self-correct.
Copy the table Claude produced (Ctrl+A, Ctrl+C on the output) and paste it into Excel. Use Paste Special → Text to get clean data. Then apply your normal age-and-blend methodology on top of the extracted benchmarks.
What you should see: Clean rows of benchmark data in your Excel workbook, ready for aging factors and blending with other survey sources.
Extract compensation by geography: "From this survey, extract P50 base salary for [job family] at [levels] for these regions: US National, California, NYC Metro, Texas. Format as a comparison table."
Compare to internal salary structure: "I've extracted these survey benchmarks: [paste your table]. Our current grade midpoints are: Grade 7 = $95K, Grade 8 = $110K, Grade 9 = $128K. Add a column showing the difference between our midpoint and the survey P50 as a dollar amount and percentage."
Summarize survey methodology: "Summarize the methodology section of this survey: sample size, industries represented, data collection timeframe, how outliers are handled, and any geographic or industry adjustments applied. Flag any limitations that would affect how I should use the data."