Dataset Transform — JSON to CSV, flatten, filter, rename any Apify dataset

For anyone whose scraper output is nested, messy JSON and who needs a clean table — a CSV for Excel or Google Sheets, a flat file for a database, or tidy rows for the next step of a workflow. Point this Actor at an Apify dataset (or paste rows), say which columns you want, and get clean rows plus a CSV / JSONL / JSON file with a download link. $0.50 per 1,000 output rows. Reading, flattening and filtering are free; rows removed by your filters are not charged. No scraping, no external requests, no API keys.

What it does

• Flatten nested JSON: {"address": {"city": "Austin"}} becomes the column address.city (up to 20 levels, your own separator)
• Arrays your way: keep as JSON, join values (a; b; c), one column per item (tags.0, tags.1), first item only, or item count
• Select, order, omit and rename columns (selecting address keeps every column under it)
• Convert types: text to number ($1,234.50, (12), 1.234,50), integer, boolean (yes/no/true/1), ISO 8601 date (from ISO text or Unix seconds/milliseconds), or to text
• Filter rows: equals, contains, greater than, in a list, exists and more, combined with AND or OR
• Sort (numbers first, empty values last), skip and limit
• Save a file next to the dataset: CSV (delimiter, UTF-8 BOM for Excel, formula-injection protection), JSONL or JSON, with a link in the run statistics
• Works on any dataset: Google Maps, jobs, SERP, e-commerce, your own runs

Input example

{
  "datasetIds": ["WkzbQMuFYuamGv3YF"],
  "select": ["title", "address.city", "rating", "reviews"],
  "rename": {"address.city": "city"},
  "coerce": {"rating": "number", "reviews": "integer"},
  "filters": [{"field": "rating", "op": "gte", "value": 4.5}],
  "sortBy": "reviews",
  "sortOrder": "desc",
  "limit": 500,
  "exportFormat": "csv",
  "csvBom": true
}

Empty input runs a free demo on sample orders (not charged, rows carry _demo: true) so you can see flattening, type conversion, filtering, sorting and the CSV link before using your own data.

Order of processing

1. Flatten (depth, arrays) and optional text trimming
2. Convert types (coerce)
3. Filter rows, then sort
4. Keep / omit columns, then rename, then cut over-long cells
5. Skip and limit; write the dataset rows; save the file

Column names in coerce, filters, sortBy, select and omit are the flattened names before renaming (for example address.city).

Output

• Dataset: the transformed rows. Use Apify's own export buttons for Excel/XML/RSS, or the API.
• File: saved in the run's key-value store as <fileName>.csv (or .jsonl / .json). The link is in the STATS record as fileUrl. It uses Apify's signed public link, so you can share it without sharing your token; if no link can be created, the file is still in the key-value store under fileKey.
• STATS (key-value store):

{
  "rowsRead": 12034, "rowsFilteredOut": 9120, "rowsWritten": 2914,
  "coercionFailures": {"price": 3}, "cellsTruncated": 0,
  "columns": ["title", "city", "rating", "reviews"],
  "fileKey": "export.csv", "fileFormat": "csv", "fileRows": 2914, "fileUrl": "https://api.apify.com/v2/key-value-stores/…/records/export.csv?signature=…"
}

Limits (cannot be raised by input)

• 1,000,000 output rows per run, 1,000 columns per row, 50,000 pasted rows (use a dataset for more)
• Sorting holds matching rows in memory: up to 500,000 rows (the run stops with a clear message, before any charge, if there are more)
• Files are capped at 150 MB (the dataset still gets every row; fileTruncated tells you)
• Text cells are cut at 32,000 characters by default (Excel's limit is 32,767); set maxCellChars to change or 0 to disable
• Empty rows (all values empty after your column choice) are skipped and not charged
• Rows that are not objects (a plain number or text) become {"value": …}

Use cases

• Turn a Google Maps / jobs / e-commerce scraper dataset into a flat CSV for a spreadsheet
• Keep only the 6 columns your CRM import needs and rename them to its field names
• Convert price and date text into real numbers and ISO dates before loading to a database
• Filter and sort a dataset (top-rated, cheapest, newest) and hand the top 100 to the next Actor

Pricing

Pay per event: $0.0005 per output row ($0.50 per 1,000). Set a maximum charge per run to cap spend; the Actor stops cleanly when it is reached.

FAQ

Does it change my source dataset? No. It only reads it and writes a new dataset. Why are my numbers still text? Add the column to coerce as number or integer. Values that cannot be read become empty and are counted in coercionFailures. Excel shows broken characters. Turn on csvBom. Is the file link private? Anyone holding the signed link can download the file; the link does not give access to anything else in your account.

More tools from the same developer

• Google Sheets Export — export an Apify dataset to Google Sheets
• Dataset Merge & Dedup — merge Apify datasets and remove duplicates by the fields you choose
• Dataset Change Detector — compare two runs of any scraper: only added, removed and changed items
• Job Postings Normalizer — unify fields, salaries and dates from any job scraper and remove duplicates
• Google Maps Leads Cleaner — dedupe Google Maps scraper results, format phones (E.164), CRM-ready leads