📥Raw Data

NEW

Also known as: Source Data, Unprocessed Data, Primary Data

Original, unprocessed data captured directly from sources before transformation, cleansing, or modeling.

DESCRIPTION

Raw data is the initial form of captured information, preserved as-is prior to cleaning, normalization, or modeling. It is essential for reproducibility, auditability, and future reprocessing with improved methods. In modern architectures, teams store raw data in a landing or bronze zone to maintain an immutable system of record. Handling raw data requires schema management, metadata, and safeguards for sensitive information. It enables exploration and feature discovery but can be noisy or inconsistent. Proper lineage and versioning ensure traceability from raw inputs to refined outputs.

ALTERNATIVE DEFINITIONS

Wikipedia

Raw data, also known as primary data, is data collected from a source.

Establishes terminology equivalence with primary data and emphasizes origin.

Source

IBM

Raw data is data in its original, unprocessed form, before it has been cleaned, transformed, or analyzed.

Explicitly contrasts raw with cleaned/transformed states.

Source

Microsoft Learn

A raw or landing zone in a data lake stores data in its original format as it comes from source systems.

Places raw data within data lake zone architecture.

Source

RELATED TERMS

Data Ingestion Extract-Transform-Load Scrubbing Data Lake Data Schema Metadata Data Staging Data Quality Bronze Layer Lineage

COMPETING TERMS

Processed Data

RELATIONS

• Broader: Data

KNOWLEDGE AREAS

Data Engineering Data Management Data Storage & Operations Data Integration & Interoperability Data Science

TARGET AUDIENCE

Data Engineer Analytics Engineer Data Scientist Data Architect MLOps Engineer

EXTERNAL SOURCES

• Wikipedia - Raw data
• Databricks - Bronze (raw) layer in medallion
• Microsoft - Raw zone guidance for data lakes

TOP QUESTIONS

Why is Raw Data important in analytics?

It preserves original signals for reproducibility, auditing, and future reprocessing. Analysts can revisit raw data as business rules evolve, avoiding irreversible information loss.

Category: Relevancy

How should Raw Data be stored?

Use an immutable landing or bronze zone with strict access controls, schema/version metadata, and lifecycle policies. Prefer open formats and partitioning for performance.

Category: Management

What are risks of using Raw Data directly?

It may contain errors, PII, and inconsistent schemas. Direct use can produce unreliable insights and compliance violations. Always cleanse and validate before analysis.

Category: Risks

What is a common mistake with Raw Data?

Overwriting or mutating raw files. Raw data should remain append-only and versioned to ensure traceability and reproducibility.

Category: Mistakes

How does Raw Data relate to medallion architecture?

Raw data maps to the bronze layer, which feeds refined silver and gold layers through governed transformations.

Category: Example

How do you manage schema drift in Raw Data?

Implement schema evolution policies, contract testing, and validation at ingestion. Record schema versions with lineage metadata.

Category: Requirements

What hot topics impact Raw Data today?

Data contracts, privacy-by-design in ingestion, and open table formats for incremental processing are prominent themes.

Category: Topic

What’s the difference between Raw and Processed Data?

Raw is unaltered input; processed has been cleaned, transformed, and modeled for specific use cases.

Category: Differences

Where should teams start handling Raw Data?

Define landing zone standards, implement automated validation, and establish metadata capture from day one.

Category: Getting Started