Should I store my data as CSV or JSON?

Use CSV when your data is genuinely tabular — every record has the same flat set of fields — and the consumer is a spreadsheet, a data analyst, or a bulk import tool. Use JSON when your data has hierarchy (nested objects, arrays of varying length) or the consumer is a REST API, since JSON is the standard format for both.

What happens to nested objects when I convert JSON to CSV?

CSV has no concept of nesting — it's strictly flat rows and columns. When you convert JSON containing nested objects or arrays to CSV, those values typically get serialized as a JSON string inside a single cell rather than being split into separate columns. If you need real flattened columns, you have to restructure the JSON to be flat before converting, or use a flattening step that turns nested.field into a dotted column name.

Why do my numbers turn into text after converting CSV to JSON?

CSV is a plain-text format — there's no type information anywhere in a .csv file. Every value, whether it looks like a number, a boolean, or a date, is just a string of characters until something on the receiving end decides to interpret it. A naive CSV-to-JSON converter often leaves everything as a string; a good one attempts to detect and convert numbers and booleans, but you should always double-check the output types against what your API or database actually expects.

Can I import a CSV file directly into MongoDB or Firebase?

Not usually without conversion first — both are document databases that expect JSON-shaped documents, not flat rows. The standard workflow is to export or receive data as CSV (common from spreadsheets and legacy systems), convert it to a JSON array of objects, and then import that into the document database.

CSV vs JSON: When to Use Which (and How to Convert Between Them)

Two Formats, Two Different Jobs

CSV and JSON get compared constantly, but they're not really competing for the same job. CSV is a tabular format — rows and columns, like a spreadsheet. JSON is a hierarchical format — objects that can contain other objects and arrays, nested as deep as you need. Picking the wrong one for your use case is where the pain starts.

When CSV Is the Right Choice

CSV wins when:

Your data is genuinely flat — every record has the same fixed set of fields, no nesting
The consumer is a spreadsheet — Excel, Google Sheets, or a data analyst who wants to open the file directly and pivot/filter it
You're doing a bulk import into a system that expects tabular input — many admin tools, CRMs, and legacy systems accept CSV uploads specifically
File size matters — CSV has far less structural overhead than JSON (no repeated key names, no brackets/braces), so it's smaller for the same tabular data

When JSON Is the Right Choice

JSON wins when:

Your data has hierarchy — a user object containing an array of addresses, an order containing an array of line items, anything with real nested structure
The consumer is an API — JSON is the de facto standard format for REST APIs, and virtually every modern backend framework parses it natively
You're feeding a document database — MongoDB, Firebase Firestore, and similar databases store JSON-shaped documents directly, no flattening required
Data types matter — JSON has real types (number, boolean, null, string, array, object); CSV has exactly one type (text), and everything else is an interpretation layered on top

The Common Pitfalls When Converting

Nested Data Doesn't Flatten Automatically

CSV has no way to represent a nested object or array natively. When you convert JSON containing something like {"address": {"city": "NYC", "zip": "10001"}} to CSV, a straightforward converter typically serializes the whole nested object as a JSON string inside one cell — {"city":"NYC","zip":"10001"} — rather than splitting it into address.city and address.zip columns. If you need real flat columns, you need to restructure the source JSON first, or use a converter that explicitly flattens nested keys with dot notation.

Everything in CSV Is Text Until Proven Otherwise

This is the pitfall that causes the most downstream bugs. A CSV file has zero type information — 30, true, and "active" are all just character sequences on disk. When you convert CSV to JSON, a naive converter leaves every value as a string: "age": "30" instead of "age": 30. This quietly breaks anything downstream that expects a real number — a sort, a comparison, a schema validator. Always check the converted output's types against what your consumer actually expects, and convert explicitly (Number(value)) rather than assuming a converter guessed correctly.

Inconsistent Columns Across Rows

When converting a JSON array to CSV, if some objects have a field that others don't, the CSV needs a consistent set of columns across every row — typically the union of every key seen across all objects, with empty cells for whichever objects don't have a particular field. This is usually the right behavior, but it's worth knowing about before you're confused by unexpected empty cells in your output.

Commas Inside Values

A CSV field containing the delimiter character itself (a comma, in a standard comma-delimited file) needs to be quoted — "Smith, John" — or it silently corrupts the column alignment for that row and every field after it. This is exactly why CSV parsing isn't as simple as line.split(","); a real parser has to understand quoting rules, not just split on the delimiter blindly.

A Concrete Example: CSV to MongoDB

This is one of the most common real conversions developers do. MongoDB (and Firebase Firestore, and most document databases) store JSON-shaped documents — they don't understand a flat CSV row directly. The standard workflow:

Receive or export the data as CSV — common from a spreadsheet, a legacy export, or a partner data feed
Convert it to a JSON array of objects, one object per row, with the header row becoming the object keys
Spot-check the converted types — numbers, booleans, and dates are prime candidates for being left as strings by the conversion step
Import the resulting JSON array into the document database

Steps 1 and 2 are exactly what a CSV-to-JSON converter handles.

Try It

ToolNinja's CSV ↔ JSON Converter → converts in both directions instantly — paste CSV, get a JSON array of objects; paste a JSON array, get CSV back. It correctly handles quoted fields containing commas or embedded quotes, supports comma, semicolon, tab, or pipe delimiters for regional spreadsheet formats, and lets you upload a .csv file directly instead of pasting. Everything runs in your browser — nothing in your spreadsheet or API data is ever uploaded anywhere.

For inspecting or reshaping the converted JSON further, pair it with the JSON Formatter.

Sources: