Data Ingest collect and normalise from every source

Incoming data moves through a defined sequence, mostly automatic, with editors stepping in only for exceptions:

1. Collect from any source. The platform continuously monitors and pulls data from wherever it lives:

• broadcaster portals and websites
• FTP/SFTP servers
• broadcaster APIs (REST and SOAP)
• email inboxes
• cloud storage (Amazon S3, Google Cloud Storage, Azure Blob)

2. Classify and route. Each incoming file is identified by type (a full weekly schedule, an incremental update, a description file, an image package) and routed to the correct channel. A human-readable preview (XML, CSV, DOC/DOCX, PDF) is available, and import can be set per channel as fully automatic or requiring approval.

3. Import and map. A dedicated importer for each source and format converts the input into one unified internal structure, applies broadcaster-specific field mapping, and runs an initial QA pass: required fields present, chronological consistency (no overlapping programmes), encoding and format corrected.

4. Analyse the text. Natural-language processing extracts what is buried in free-text descriptions, such as guest names and episode numbers, infers missing data, and segments it into the right fields.

5. Apply rules and link. A business-rules engine cleans up systematic source errors automatically (lowercase titles, unwanted prefixes, misplaced data), then links each entry to the right record in the content archive.

When a source sends an update, the platform compares it against the current state and highlights exactly what is different, field by field, in a side-by-side view. Trusted, low-risk changes can be applied automatically; anything unusual or ambiguous is flagged for an editor to accept, correct or reject, with a full audit trail of every decision.

A key principle keeps downstream systems efficient: a change is only passed on when it actually affects the data you receive, not for every internal edit. You are not flooded with updates that change nothing on your side.

Sources send data in many different shapes: XML, JSON, CSV, spreadsheets, PDFs, broadcaster-specific formats and more, over 2,500 input formats in all. Each is converted to a single internal format and mapped to the right fields, so the rest of the platform sees one consistent structure no matter where the data came from. This is what lets Media Press take on a new source without disrupting everything downstream.