Retries and continue-on-fail

Make a single step resilient – automatically retry it on a transient failure, or carry on down the normal path instead of stopping the whole run.

Help CentreBuilding Processes

Even a well-built process hits the occasional hiccup – a slow page, a brief network blip, an element that loads a moment late. Per-step retry and continue-on-fail settings let a step recover from these on its own, so a momentary problem doesn't end the whole run. These settings sit alongside the exception connection and Try-Catch, which handle errors that can't simply be retried.

1. The per-step settings

On a step you can set:

  • Max retries – how many times the bot re-attempts the step if it fails.
  • Retry delay – how long to wait before each re-attempt.
  • Backoff – how the delay grows between attempts: fixed (the same wait every time) or exponential (the wait roughly doubles each time, easing pressure on a struggling system).
  • Continue on fail – if the step still fails after its retries, log the problem and carry on down the normal (DEFAULT) path instead of stopping the run.

By default these do nothing – a step runs once and, if it fails, the run follows the step's exception path or stops.

Tip: Only enable retries on steps that can succeed on a second attempt – a slow page, a brief network blip, an element that loads late. Retrying a step that failed for a real reason just wastes time and credits.

2. Which failures are retried

Retrying only helps for transient problems. The bot automatically retries failures that are likely to pass on a second attempt:

  • An on-screen element that wasn't found yet
  • Server errors (HTTP 5xx) and timeouts
  • Other unexpected, unclassified errors

It will not retry failures that won't fix themselves – retrying would just waste time and credits:

  • Bad data or bad configuration
  • A business rule that rejected the work
  • Access denied (HTTP 401 / 403)
  • A captcha challenge
  • Missing credentials

This is why a step sometimes retries and sometimes fails immediately – the bot looks at why it failed before deciding. You can see the attempts and the outcome in the run log and history.

3. Continue-on-fail vs error handling

  • Continue on fail – best for a non-critical step where you'd rather note the problem and keep going.
  • Exception connection / Try-Catch – best when a failure needs its own recovery path (log it, notify someone, clean up).

Related articles

See it working on your own data

Everything documented here ships with the platform – try the document tools free, or go live in 7 days.