# Delivery Format We provide flexible access to our data, so your team always gets the information in the right format, at the right time. You can choose between 4 different flat file datasets format. | Format | Pros | Cons | | ---------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------- | | **JSON** | - Human-readable and widely used- Nested structures supported- Great for APIs and web data | - Large file sizes (not space-efficient)- Slower to parse on big datasets- Harder to process in batch (not line-delimited) | | **JSONL** (JSON Lines) | - Line-delimited, easy to stream and process- Works well with big data pipelines (can split files easily)- Still human-readable | - Less widely adopted than plain JSON- No built-in schema enforcement- Slightly less convenient for nested arrays than JSON | | **CSV** | - Very simple and widely supported- Compact for tabular data- Easy to open in Excel/Sheets | - No support for nested/complex data- No types (everything is text)- Ambiguities with delimiters, quoting, encoding | | **Parquet** | - Highly compressed and columnar (great for analytics)- Efficient for large-scale queries (only reads needed columns)- Schema + data types preserved- Optimized for big data tools (Spark, Hive, Snowflake) | - Not human-readable- Higher overhead for small datasets- More complex libraries needed to read/write | *** ### What does the data look like? When you choose to export or receive data, you can select different formats. Each format represents the data in a different way. Here’s what you can expect: #### **1. JSON (JavaScript Object Notation)** ```json { "id": 1, "name": "Alice", "email": "alice@example.com", "skills": ["Python", "SQL"] } ``` **Key points:** * A **human-readable text file** used almost everywhere. * Stores data as objects with keys (`"name"`, `"email"`) and values. * Can handle lists, nested objects, and complex structures. **When to use it:** * Great for structured data that needs to be read or shared easily. * Widely used in APIs and integrations. #### **2. JSONL (JSON Lines)** **What it looks like:** ```json {"id": 1, "name": "Alice", "email": "alice@example.com"} {"id": 2, "name": "Bob", "email": "bob@example.com"} {"id": 3, "name": "Charlie", "email": "charlie@example.com"} ``` **Key points:** * Each **line in the file** is its own JSON object. * Easy to process line by line, especially with large files. * Lighter and faster than a big JSON array for bulk data. **When to use it:** * Perfect for **large-scale data processing** (Spark, Elasticsearch, etc.). * Useful for **logs, events, or streaming-style data**. #### **3. CSV (Comma-Separated Values)** **What it looks like:** ```csv id,name,email 1,Alice,alice@example.com 2,Bob,bob@example.com 3,Charlie,charlie@example.com ``` **Key points:** * A **simple text table**, with each row as a record and each column separated by a comma (sometimes semicolon). * Opens easily in Excel, Google Sheets, or any spreadsheet tool. * Doesn’t support nested or complex data—everything is flattened into rows and columns. **When to use it:** * Best for **simple tabular data**. * Ideal if you want to quickly view and manipulate your data in Excel. #### **4. Parquet** **What it looks like:** * You **cannot open a Parquet file in a normal text editor**. * It’s a **binary, compressed format** optimized for storage and analytics. **Example (using Python/Pandas):** ```python import pandas as pd df = pd.read_parquet("data.parquet") print(df.head()) ``` Output (table view): ``` id name email 0 1 Alice alice@example.com 1 2 Bob bob@example.com 2 3 Charlie charlie@example.com ``` **Key points:** * Stores data with proper types (numbers, strings, dates, etc.). * Very efficient in terms of **size** and **query performance**. * Supported by all major data tools (Snowflake, BigQuery, Spark, etc.). **When to use it:** * Best for **large datasets**. * Ideal for **analytics pipelines** and **machine learning workloads**. ***