
βοΈ SaaS & Cloud Β· Data Flow
How data from apps, databases and events lands in a lakehouse and becomes dashboards and ML features: bronze, silver and gold layers.
Opening a template is free and never counts toward your daily diagrams. Asking the AI to change it counts as one.
Drawing diagramβ¦
Data flow for a lakehouse: operational databases stream changes through CDC (Debezium) and app events through Kafka, SaaS tools are loaded by a batch connector (Fivetran). Everything lands raw in the bronze layer (Delta Lake on object storage), Spark jobs clean and de-duplicate into silver, then build business models (orders, customers, revenue) in gold. BI dashboards and ML feature pipelines read gold; a data catalog tracks lineage and access. Number the processes and label every flow.
flowchart LR
subgraph EXT ["External entities"]
op_db["Operational Databases"]
app_evt["App Events"]
saas["SaaS Tools"]
end
subgraph PROC ["Processes"]
p1("1.0 Stream Database Changes")
p2("2.0 Ingest App Events")
p3("3.0 Batch Extract SaaS Data")
p4("4.0 Clean and Deduplicate Data")
p5("5.0 Build Business Models")
p6("6.0 Track Lineage and Access")
end
subgraph STORE ["Data stores"]
d1[("D1 Bronze Layer Raw Delta Lake")]
d2[("D2 Silver Layer Cleaned Delta Lake")]
d3[("D3 Gold Layer Business Models Delta Lake")]
d4[("D4 Data Catalog Storage")]
end
op_db-->|"Database CDC changes via Debezium"|p1
app_evt-->|"Application events via Kafka"|p2
saas-->|"SaaS data via Fivetran"|p3
p1-->|"Raw database CDC logs"|d1
p2-->|"Raw event payloads"|d1
p3-->|"Raw SaaS records"|d1
d1-->|"Raw data records"|p4
p4-->|"Cleaned and deduplicated data"|d2
d2-->|"Silver cleansed datasets"|p5
p5-->|"Business models data for orders, customers, and revenue"|d3
p1-->|"Streaming ingestion metadata"|p6
p2-->|"Event ingestion metadata"|p6
p3-->|"Batch ingestion metadata"|p6
p4-->|"Silver transformation lineage"|p6
p5-->|"Gold business model lineage and access logs"|p6
p6-->|"Catalog lineage and access metadata"|d4A data flow diagram with 9 processes and outside parties, 4 data stores and 16 data flows.
Describe your own system in plain English and CloudSketch AI draws it in seconds. See all AI diagram generators.
How a microservices app runs on Kubernetes in the cloud: ingress, namespaces, services with multiple replicas, managed databases and monitoring.
A typical CI/CD pipeline from a pull request to production: build, tests, security scans, staging deploy, approval and a canary release.
The authorization code flow with PKCE used by most modern web and mobile apps to log users in through an identity provider.
A B2B SaaS product serving many customer organisations from one platform: tenant-aware services, per-tenant data isolation, billing and admin.
The classic three-tier web app on AWS: CloudFront and a load balancer, an autoscaling app tier in private subnets, and a Multi-AZ database.
Services that talk through events instead of direct calls: an order service publishes events that inventory, payments, notifications and analytics react to.