> ## Documentation Index > Fetch the complete documentation index at: https://docs.brightdata.com/llms.txt > Use this file to discover all available pages before exploring further. # How Bright Data Managed Services works > Bright Data Managed Services runs in five stages, from scope definition to ongoing operations, with two delivery modes and output in JSON, CSV or Parquet. A Bright Data Managed Services project moves through five stages, from scope definition to ongoing delivery, and most projects are operational within days to a few weeks after requirements are defined. ## What are the stages of a Managed Services project? Each Managed Services project follows the same five stages. After the first delivery, site changes, new sources and scaling send the project back to collector development, so the pipeline keeps running as targets change. **1. Scope definition** Before Bright Data builds any collectors, you and Bright Data agree in writing on: * Target sources, with priority and geography * Output schema: field names, types and which fields are required * Collection cadence and freshness requirements * Delivery format and destination * Edge cases and business rules This shared specification is the foundation of the Managed Services engagement and is maintained throughout the project. **2. Collector development** Bright Data engineers build collectors for your sources, fields and collection logic. Projects range from a handful of domains to thousands. For large source lists, delivery is phased. Collectors are built and validated in batches, so the first domains start delivering data while the remaining ones are still being built. Phasing also lets you review early output and confirm the data and schema before the full rollout. **3. Validation and QA** Before the first delivery, every record passes automated schema validation, deduplication, completeness scoring and content fingerprinting. For AI training data engagements, optional compliance screening is available at the domain, page and asset level. **4. First delivery** Bright Data delivers a sample dataset for your review. Once the sample is validated against your requirements, recurring delivery starts on the agreed schedule. **5. Ongoing operations** The Managed Services team monitors pipelines continuously. When a target site changes its structure, tightens anti-bot measures or introduces new page patterns, the team updates the collectors and re-validates the output before the next delivery. Issues are flagged, investigated and communicated to you without waiting for you to report them. ## Which delivery modes does Managed Services support? Managed Services supports two delivery modes. Bright Data owns the collectors, infrastructure, monitoring and issue resolution in both. | Mode | How data arrives | Your involvement | | - | - | - | | Hands-off | Bright Data runs the pipeline on a defined schedule and delivers a recurring feed to your destination, such as Amazon S3, Snowflake, Databricks, SFTP or a webhook | None | | API-driven | You trigger collections on demand through an API | Triggering collections | ## Which delivery destinations and output formats are available? Managed Services delivers to cloud storage, an API, a webhook, SFTP or BI tools, in JSON, NDJSON, CSV, Excel, Parquet, a direct database integration or a custom format on request. | Method | Examples | | - | - | | Cloud storage | Amazon S3, Google Cloud Storage, Azure Blob, Snowflake, Databricks, BigQuery | | API | Pull results programmatically after each collection run | | Webhook | Push delivery to your endpoint on each run | | SFTP or direct download | File-based delivery to your infrastructure | | BI and analytics tools | Snowflake, Tableau, Salesforce and custom integrations | ## Which specialized packages does Managed Services offer? Beyond standard web data collection, Managed Services offers three packages for specific use cases. ### What is the Web Data for AI Model Training package? The Web Data for AI Model Training package delivers structured, model-ready data from public web sources to AI and ML teams. The package covers source identification, collector development, AI-powered normalization and enrichment, and delivery in the format and destination your training pipeline needs. The package supports live collection, historical data retrieval and multimodal collection (text, images, video, audio and PDFs) across many sources at once. For large AI training engagements, the pipeline includes: * **Compliance screening** at the domain, page and asset level, aligned to a legal playbook defined with you. Screening covers terms of service analysis, robots.txt interpretation, brand and keyword deny lists, PII flagging and NSFW classification. * **Deduplication** that detects and removes exact, near-exact and semantic duplicates before delivery. * **Asset-level quality validation**, including minimum file size, format validity and content quality checks. * **Cloud tenancy integration** with OCI, AWS, GCP and Azure, so AI validation models can run inside your own cloud tenancy. ### What is the Mass Web Data Ingestion and Enrichment package? The Mass Web Data Ingestion and Enrichment package collects from hundreds or thousands of public sources where building an individual collector per source is not practical. AI agents discover, extract, ingest and enrich the data. The package supports parsing, normalization, sentiment analysis, tagging, translation and categorization across many diverse sources, and suits data teams building or expanding large multi-source data operations. ### What is the Product Matching package? The Product Matching package identifies and matches products across e-commerce platforms for retailers, marketplaces, brands and data providers. The package supports exact, attribute-based and fuzzy matching with confidence scoring. Product Matching is delivered as a recurring feed with no manual matching. ## How do you start a Managed Services project? A Bright Data data expert scopes every Managed Services project with you before collection begins. For common questions about quality, compliance and setup time, see [Bright Data Managed Services FAQs](https://docs.brightdata.com/products/managed-services/faqs.md). > [card] [Talk to a data expert](https://brightdata.com/products/managed-service) > > Contact Bright Data to scope a Managed Services project.