Photo: free stock photography (Unsplash licence) — see imprint
Flows worth building on day one
Notify purchasing when stock drops below the reorder point. Tag customers who abandoned a cart over €500 so sales can call them. Escalate an order that has been unpaid for ten days. Each of these is fifteen minutes of configuration and replaces a recurring human check.
Where Flow Builder stops
Flow Builder is event-driven and stateless. It cannot loop, it cannot wait for an external system, and it cannot do a nightly reconciliation against your ERP. The moment you need those, you need a scheduled task or a message queue worker — not a flow with a clever workaround.
Custom flow actions are cheap
Writing your own flow action is a small plugin: a class, a config form, done. Now your non-technical team can wire 'push this order to the ERP' into any trigger they like, without a developer in the loop each time.
Log everything, or debug nothing
A silent flow that stopped firing three weeks ago is worse than no flow. Log every execution and alert on failures — automation you cannot observe is a liability, not an asset.
- Stock, abandoned carts and dunning are the fastest wins.
- No loops, no waiting — that is a worker's job.
- Unobserved automation is a liability.
Frequently asked questions
The boring recurring checks a human currently does from memory. Notify purchasing when stock drops below the reorder point. Tag customers who abandoned a large cart so sales can call them. Escalate an order that has been unpaid for ten days. Each is minutes of configuration, and it replaces a task that was being forgotten roughly as often as it was done.
No. It is event-driven and stateless: something happens, a flow runs, it ends. It cannot loop, it cannot wait, and it cannot do a nightly reconciliation against your ERP. The moment you need any of those, you need a scheduled task or a queue worker — not a flow with a clever workaround wrapped around it. Spotting that line early saves you a rebuild.
Only if you built that in. A flow that quietly stopped firing three weeks ago is worse than never having had one, because the team stopped checking manually the day you switched it on. Log every execution, alert on failures, and look at the numbers occasionally. Automation you cannot observe is a liability wearing the costume of an asset.
Check the Flow Builder first — you already paid for it, and most shops use a fraction of it. A surprising number of requests that arrive as 'we need a small plugin' turn out to be fifteen minutes of configuration. Reach for code when you hit a real limit: looping, waiting, state, or a verb the shop does not know yet. Not before that.
We do this for a living — Shopware, Node.js, React, ERP integration and automation for B2B.
Talk to an engineer