Prepare business data for a suitability discussion.
Begin with a description. A clean-looking spreadsheet does not establish ownership, privacy or demand.
1. Establish authority.
Identify who created the material and which contracts govern it. Check customer, employer, platform and third-party restrictions. Obtain appropriate rights review before sharing. Do not copy OEM manuals, paid datasets or customer files into a pilot.
2. Define the buyer’s task.
Specify the period, granularity, units, fields, exclusions and intended evaluation. Agree a minimum useful scope before preparing hundreds of records.
3. Exclude prohibited information before upload.
Remove names, email addresses, exact customer/site locations, credentials, bank/card details, device serial identifiers and sensitive information. Review whether combinations of remaining fields can identify someone. Automated redaction is imperfect and does not certify anonymity.
4. Make a documented table.
Use UTF-8 CSV or structured JSON with stable column names. Current limits are 512 KB, 10,000 rows and 100 columns, including limits on normalized output. Do not split or rename fields to bypass controls. Record missing values and duplicate counts instead of silently inventing or changing information.
5. Review the processed output.
Approved uploads send plaintext over HTTPS for cleaning; Google Gemini reviews the normalized table. The result is encrypted and DataVault retains keys for controlled delivery. Check the full cleaned file, schema and description for remaining identifiers and lost useful content.