Writing SQL in Runway

​

Identifying a Fivetran integration

1. During setup – The integration will display the Fivetran label when being installed.
2. After installation – If already installed, you can check by looking at the URL in the query editor.

​

Adding a new query

​

Writing SQL for data warehouses

• Google BigQuery – Uses GoogleSQL
• Snowflake – Supports standard SQL
• Amazon Redshift – Uses PostgreSQL

​

Writing SQL for Fivetran integrations

SHOW TABLES

1. Data is pulled from your source system into Runway’s Snowflake data warehouse every morning (US Pacific Time).
2. Runway then automatically executes integration queries shortly after to refresh the data in your model. If you need an immediate manual refresh:
3. Manually trigger a sync from your source system.
4. Once the source data is refreshed, rerun your SQL queries manually to update your model. This step ensures the newly synced data appears in your model.

​

Best practices & tips

• Structure your queries to return a long dataset — a list of transactions, accounts, or metrics with a daily or monthly amount. This format ensures flexibility for modeling. Example:

  dimension_1	dimension_..x	date_1	date_..x	value_1	value_..x
  Account 1	Asset	01/01/2025	03/01/2025	10	20
  Account 2	Liability	01/02/2025	04/01/2025	30	40

• Include all dimensions, dates, or values needed — not all of them need to be used in the final database configuration.
• Aggregate/roll up data to the minimum level required for modeling and drill-downs. For example, if pulling general ledger data, and only monthly spend per account is needed, write your query to pre-aggregate the data:
  • Avoid this
    select
    	account,
    	transaction_amount,
    	transaction_date
    from general_ledger
    
    
  • Do this
    select
    	account,
    	sum(transaction_amount) as transaction_amount,
    	date_trunc('month',transaction_date) transaction_month
    from general_ledger
    group by 1,3
    
    
• Filter out unwanted data — some Fivetran tables contain columns like _fivetran_deleted. Filtering out these records helps prevent duplicate records in your dataset.