The most damaging automation failures are the quiet ones: a form field is renamed, a connection expires, and records stop arriving with no error anyone sees. Designing for failure from the start is cheaper than discovering it from an unhappy client.
Common causes of silent failure
Most failures come from changes outside the automation itself:
- A form, spreadsheet column or field is renamed or reordered.
- An app connection or password expires.
- An input arrives in an unexpected format (a date written differently, an empty required field).
- A usage limit on a tool is reached.
- The step after the automation changed, and nobody updated the automation.
Give every workflow a failure route
Decide in advance what happens when a step fails. At minimum: the failed item is kept (not lost), a named owner is notified with enough detail to act, and there is a documented way to re-run or finish it manually.
Send failure notices to a shared place the team already checks, rather than to the personal inbox of whoever built the workflow.
Add a simple health check
Silent failures are failures of absence: nothing happens. A weekly count is often enough to catch them — "we usually process around this many enquiries; this week shows zero". Put the count somewhere visible, such as the weekly operations review.
Test with bad inputs, not just good ones
Before launch, run the workflow with a missing field, a duplicate, an oversized file and a wrongly formatted date. Confirm each one either succeeds correctly or lands in the failure route. Write the results down; they become the test checklist for every future change.
Key takeaways
- Assume failure and decide where failed items go.
- Notify a shared place and a named owner, not the builder's inbox.
- A weekly volume check catches failures of absence.
