Start at the edges

Before building a recipe, write down the trigger, the source of truth, and what a successful run means. Decide whether each event represents a new record or an update. These small decisions make the rest of the integration easier to reason about.

Make retries safe

Use a stable external identifier for upserts. A timeout does not always mean a write failed: check the destination before repeating an action. Retry transient failures with a bounded delay, and send validation failures to a review queue.

Leave a useful trail

Record the source event ID, destination ID and outcome. Keep credentials and sensitive payloads out of logs. Give an operator enough context to diagnose a failed job without having to replay it.

Before you switch it on

Test duplicate events, expired credentials, missing required fields and rate limits in a non-production workspace. Confirm that an alert reaches the person responsible for the workflow.

A starting point for your own integration work. Adapt and validate it against your platform version and project requirements.

Working on a similar integration or automation? Get in touch with Aldo.