Reading and building a node graph
A flow you can read six months later is worth more than a clever one. Layout is not decoration here.
Automation is a node-based editor: triggers connect to actions, nodes sit on a grid, flows run on demand or on their trigger, and any flow can be saved as a template.
The mechanics are quick to learn. What takes longer is building flows that somebody else — including you in six months — can understand without tracing every line.
Layout is documentation
The grid exists so that a graph can be read left to right. Use it that way: trigger on the left, the happy path straight across, branches dropping below.
A flow whose nodes are scattered wherever they were dropped takes ten minutes to understand and gets rebuilt rather than edited. That is a real cost — the rebuilt version loses whatever edge case the original had quietly handled.
Name nodes for what they do in your process, not what they are. "Notify the account manager" beats "Send message" when you are scanning nine nodes looking for the one that is wrong.
Templates, for the shape you repeat
Saving a flow as a template pays off when you have the same shape with different values — the same three-step notify-and-assign for four different projects.
What it does not solve is duplication. Four copies of a template are four flows to update when the process changes, and the one you forget is the one that keeps doing the old thing. Before duplicating, check whether one flow with a condition would do — a single branch is easier to maintain than four near-identical graphs.
Understanding a flow you did not build
Three questions, in this order, and you will have it:
- What fires it? Read the trigger before anything else — most confusion about a flow is really confusion about when it runs.
- What does it write? Follow the actions that change data. Those are the ones that can loop, and they are the ones with consequences.
- What happens on the branch nobody drew? The empty field, the missing person, the previous step returning nothing. This is where inherited flows break.
Run it on demand once before you change anything. Watching it execute answers more than reading it does.
Questions people actually ask
How should I arrange nodes?
Left to right on the grid: trigger at the left, the main path straight across, branches dropping below. Scattered graphs get rebuilt instead of edited, and the rebuild loses whatever edge case the original quietly handled.
When is a template better than a new flow?
When you repeat the same shape with different values. But four copies are four things to update later — check first whether one flow with a condition would do the job.
How do I understand somebody else's automation?
Read the trigger first, then the actions that write data, then ask what happens on the branch nobody drew — empty field, missing person, previous step returned nothing. Then run it on demand and watch.
