Nothing close enough? Start from a blank flowchart → Describe it in one paragraph.
How to use a flowchart template.
- 01Identify sources and destinations
List your data sources (Postgres, S3, Kafka, API) and the target system (Snowflake, BigQuery, Redshift, data lake). These become the start and end nodes.
- 02Describe transformation stages
Enumerate each processing step: extract, decode, clean/deduplicate, join, aggregate, enrich. ChatDiagram creates a process rectangle for each stage.
- 03Add validation and quality gates
Specify checks like row-count reconciliation, schema validation, null-rate threshold, or referential integrity. ChatDiagram adds decision diamonds with pass/fail branches.
- 04Define error handling and retries
Describe what happens on failure — dead-letter queue, alert, retry with backoff, or abort. ChatDiagram adds error-path branches with appropriate annotations.
- 05Export for documentation
Download as PNG or PDF to embed in your data catalogue, runbook, or architecture decision record (ADR).
Questions about flowchart templates
What is the difference between ETL and ELT in a pipeline flowchart?
In ETL (Extract-Transform-Load) the transformation step appears between extraction and loading — the data is processed before it reaches the warehouse. In ELT (Extract-Load-Transform) the raw data is loaded first, and transformation happens inside the warehouse using SQL (dbt, Spark SQL). In a flowchart, the order of the Transform and Load rectangles reflects which pattern you are using.
How do I show a branching pipeline where different record types take different paths?
Use a decision diamond labelled with the routing condition (e.g., 'Record type = event?'). Each branch leads to its own transformation and load step before rejoining at a downstream stage, or continuing to separate destinations. Label each branch arm clearly (yes/no, or the specific type name).
How should I represent a streaming pipeline vs a batch pipeline in a flowchart?
For streaming, use a loop arrow back to the ingestion step after the process/output nodes to show the continuous nature, and annotate it with the trigger (e.g., 'per message' or 'per micro-batch'). For batch, include a schedule trigger at the start node (e.g., 'Daily 02:00 UTC') and a completion/success notification at the end.
What should a data pipeline flowchart include for an on-call runbook?
A runbook-oriented flowchart should prominently show: the failure detection point (monitoring alert or pipeline task failure), the decision tree for classifying the failure (source outage vs transformation error vs load rejection), the remediation action for each class (skip/retry/reprocess), and the escalation path if automated remediation fails.