Packaging & Delivering Your Data for Maximum Value

How data is packaged affects:

• Buyer first impression: Well-packaged data signals professionalism
• Integration speed: Buyers can integrate faster, potentially paying more
• Data quality perception: Clear documentation reduces buyer risk
• Value multiplier: Good packaging can add 20-50% to perceived value

Tabular Data (Most Common)

CSV (Comma-Separated Values)

• ✅ Pros: Universal compatibility, simple, human-readable
• ❌ Cons: Large files, no native compression, character encoding issues
• Use when: Data is standard tables, broad buyer base expected
• Example: CRM records, customer behavior, financial transactions

Parquet

• ✅ Pros: Compact, efficient, preserves data types, good for big data
• ❌ Cons: Requires special tools to read, less universal
• Use when: Large datasets (>1GB), technical buyer, speed important
• Example: 1M+ log events, behavioral tracking data

JSON (JavaScript Object Notation)

• ✅ Pros: Flexible, hierarchical data support, API-friendly
• ❌ Cons: Larger file size than CSV, character encoding issues
• Use when: Complex nested data, API-centric buyer
• Example: Support tickets with nested fields, product catalogs with variants

Excel (XLSX)

• ✅ Pros: User-friendly, widely supported, multiple sheets possible
• ❌ Cons: Row limits (1M), not ideal for large data, slow to load
• Use when: Smaller datasets (<100k records), non-technical buyer
• Example: Pricing data, aggregated statistics, summaries

SQL Database Dump

• ✅ Pros: Complete schema, relationships preserved, powerful for analysis
• ❌ Cons: Requires database knowledge, larger file size
• Use when: Relational data, sophisticated buyer, maximum flexibility
• Example: CRM with related tables, workflow systems with hierarchies

Recommendation

Best practice: Provide in 2-3 formats

• Primary format for expected buyer type (CSV for general, Parquet for data engineers)
• Secondary format for flexibility (JSON for API compatibility)
• Small sample in Excel for quick review

Directory Structure

data-sale-package/
├── README.txt                  # Start here
├── DATA_DICTIONARY.csv         # Field definitions
├── ANONYMIZATION_REPORT.pdf    # How data was anonymized
├── data/
│   ├── customers_202401.csv
│   ├── customers_202402.csv
│   └── ...
├── samples/
│   ├── sample_100_records.csv  # For quick review
│   └── sample_structure.xlsx   # Data structure example
└── metadata/
    ├── collection_methodology.txt
    └── quality_metrics.json

Logical Organization

• By time period: data_2024q1.csv, data_2024q2.csv (if multiple periods)
• By category: support_tickets.csv, crm_records.csv, etc.
• By sensitivity: public_data/, anonymized_data/, etc.
• Samples separate: Keep sample data in /samples/ for easy review

Good Naming

• crm_contacts_2024_anonymized.csv
• support_tickets_2023_2024.parquet
• product_catalog_20241001.json
• financial_transactions_monthly_202301_202401.csv

Bad Naming

• data.csv (unclear content)
• export_final_FINAL_v3.csv (version chaos)
• john_export.csv (includes name, privacy risk)
• secret.csv (too vague, concerning tone)

Naming Format

[data_type]_[identifier]_[anonymization]_[date].format

• data_type: crm_contacts, support_tickets, invoices, etc.
• identifier: year range, version, or frequency (optional)
• anonymization: "anonymized" or "hashed" if relevant
• date: YYYYMMDD format (20241001 = Oct 1, 2024)

Essential Information Per Field

Additional Metadata

• Null handling: What does NULL mean? Was it missing at source or excluded during anonymization?
• Value ranges: Min/max/distribution info if not obvious
• Categorical values: List all possible values for categorical fields
• Calculated fields: How are computed fields calculated? What's the formula?
• Data collection: When was data collected? Update frequency?

README.txt Template

DATA SALE PACKAGE - [Your Company Name]
============================================================

OVERVIEW
This package contains anonymized [data type] from [your company].
Total records: [X]
Time period: [start] to [end]
File format: CSV
Total size: [X GB]

CONTENTS
1. README.txt - This file
2. DATA_DICTIONARY.csv - Field definitions and types
3. ANONYMIZATION_REPORT.pdf - Anonymization methodology
4. data/ - Main data files
   - [filename].csv (description, record count, size)
5. samples/ - Sample files for quick review

QUICK START
1. Read DATA_DICTIONARY.csv to understand fields
2. Review samples/sample_100_records.csv
3. Load data/[filename].csv into your analysis tool

QUALITY METRICS
- Total records: [X]
- Complete records: [Y]% (records with all fields)
- Null handling: [Describe]
- Duplicates: [0] (after de-duplication)
- Data quality score: [X]/100

ANONYMIZATION
- Method: SHA-256 hashing of direct identifiers + aggregation
- See ANONYMIZATION_REPORT.pdf for detailed methodology
- Re-identification risk: <5% (third party re-identification)

DATA COLLECTION
- Collection period: [dates]
- Collection frequency: [daily/weekly/monthly]
- Last update: [date]
- Update frequency: [after sale] or [ongoing]

LEGAL
- GDPR compliant: Yes
- CCPA compliant: Yes
- Data purchase agreement: Required (provided separately)

QUESTIONS
Contact: [your email]
Support: [support contact]
Documentation: [link if available]

ANONYMIZATION_REPORT.pdf

Include:

• Executive summary of anonymization approach
• List of fields and anonymization technique per field
• Re-identification risk assessment results
• Methodology references (GDPR, NIST, academic standards)
• Residual risks or limitations

Why Provide Samples

• Buyers can validate structure before full download
• Builds confidence in data quality
• Shows good faith and professionalism
• Reduces buyer hesitation

Two Types of Samples

Small CSV Sample (100-1000 rows)

• Actual data, fully anonymized
• Representative of full dataset distribution
• Can include all fields or key fields only
• Filename: sample_100_records.csv

Structure Excel (show format only, no sensitive data)

• Shows all fields with empty/example values
• Data types shown (date, integer, string, etc.)
• Includes sample values for each field
• Filename: sample_structure.xlsx

Small Datasets (<1GB)

Option 1: Encrypted File Service (Recommended)

• Tresorit, Virtru, Sync.com
• Advantages: Encrypted, traceable, time-limited access
• Send link to buyer with password (sent separately)

Option 2: Cloud Storage with Presigned URL

• AWS S3, Google Cloud Storage, Azure Blob
• Advantages: Large file support, fast downloads
• Use time-limited presigned URLs
• Enable access logging

Option 3: Secure FTP/SFTP

• SSH-based file transfer
• Advantages: Encrypted, professional, audit trails
• Requires buyer to have FTP capability

Large Datasets (>1GB)

Option 1: Direct Cloud Bucket Access

• Buyer gets read-only access to S3 bucket
• Time-limited IAM role (expires after contract)
• Advantages: Fast, can stream directly to buyer's systems
• Disadvantages: Requires buyer AWS/GCS account

Option 2: Physical Media

• External hard drive or SSD shipped via courier
• Data encrypted on drive
• Advantages: Very fast for multi-TB datasets
• Disadvantages: Shipping delays, physical security risk
• Track package carefully; ensure secure receipt

Option 3: Escrow Service

• Third-party holds data until payment clears
• Protects both buyer and seller
• Advantages: High security, dispute resolution
• Disadvantages: Cost ($500-$5000+), slower

File Integrity Checks

Provide checksums so buyer can verify file wasn't corrupted:

crm_contacts_2024_anonymized.csv
SHA-256: a1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1b2

MD5: 5d41402abc4b2a76b9719d911017c592

Buyer can verify: shasum -a 256 crm_contacts_2024_anonymized.csv

Delivery Confirmation

• Require buyer to confirm receipt and validate checksums
• Only then trigger payment (if escrow) or invoice (if prepaid)
• Document download dates/times for audit trail

✅ File Format

• Primary format chosen (CSV/Parquet/JSON based on buyer type)
• Sample provided in another format for compatibility
• All encoding is UTF-8
• No proprietary format dependencies

✅ Organization

• Logical directory structure
• Clear, consistent file naming
• Samples in separate /samples/ folder
• Metadata in separate /metadata/ folder

✅ Documentation

• README.txt covers all essentials
• DATA_DICTIONARY.csv complete and accurate
• ANONYMIZATION_REPORT.pdf included
• Collection methodology documented

✅ Verification

• Sample data provided for review
• Checksums calculated and provided
• File sizes documented
• Record counts verified

✅ Delivery

• Encryption in transit (encrypted file service or SFTP)
• Access is time-limited
• Audit trails enabled
• Buyer can verify integrity
• Delivery confirmation process planned

Well-packaged data: 1.3x - 1.5x Structure & Cleanliness multiplier (vs. 1.0x for raw export)

Professional packaging:

• Increases buyer confidence
• Speeds integration and value realization
• Enables repeat business
• Justifies premium pricing

❌ "I'll just send a ZIP of exported files" → No documentation, format unclear, unprofessional

❌ "I'll email the CSV directly" → Unencrypted, no access logging, privacy risk

❌ "Field names are self-explanatory" → Buyer wastes time figuring out what fields mean

❌ "I'll just export from the database as-is" → Internal IDs, unnormalized data, hard to use

❌ "I'll provide data in our proprietary format" → Buyer can't use without our tools

1. Choose file format(s) based on expected buyer type
2. Create data dictionary for your dataset
3. Generate samples
4. Write README and anonymization report
5. Set up secure delivery method
6. Package everything in logical structure
7. Calculate checksums
8. Submit through our Sell My Data form