Granular Data Collection
What You'll Learn
You'll learn how to collect data at the right level of detail—avoiding excessive precision when coarser information meets your needs—as a core privacy engineering practice.
The Core Idea
Granular Data Collection means choosing the appropriate level of specificity when gathering information. Not every use case requires pinpoint accuracy. Collecting a user's exact GPS coordinates down to the meter when you only need their city or region violates data minimization principles and creates unnecessary privacy risk.
Think of it like ordering a package. If the courier only needs to know your city and postal code to route the shipment, why would you give them your exact GPS coordinates, bedroom floor plan, and daily schedule? The extra detail doesn't help them—it just exposes you to more risk if that data leaks.
Practical Examples
Instead of collecting:
- Exact timestamp → collect hour or day of week
- Precise age → collect age range (18-25, 26-35)
- Full GPS coordinates → collect ZIP code or neighborhood
- Exact income → collect income bracket
- Complete browsing history → collect categories of interest
- Fine-grained sensor readings → collect aggregated summaries
Why It Matters
Excessive granularity creates multiple problems:
- Privacy exposure: Precise data enables re-identification and profiling
- Security liability: More detailed data is more valuable to attackers
- Compliance risk: Regulators scrutinize unnecessary precision
- Storage costs: Finer-grained data consumes more resources
The key question: "What's the coarsest level of detail that still serves our legitimate purpose?"
Implementation Approach
Before collecting any data field, explicitly document:
- The specific purpose requiring that data
- The minimum granularity needed for that purpose
- Why coarser alternatives won't work
If you can't justify the precision level, dial it back.
Key Takeaway: Collect data at the coarsest granularity that still fulfills your legitimate business purpose—excessive precision creates privacy risks without adding value.