What happens when a run fails? Is it re-run?

Page timeouts, CAPTCHAs, expired sessions and UI changes all cause individual tasks to fail. Miaowa GEO applies three rules: failures leave a record, failures stay out of the denominator, and failures are never presented as results.

Inside the same cycle

1. A task hitting a technical exception is retried automatically under the same task ID; retries do not consume prompt quota.
2. A task that still fails after retries is recorded as failed or blocked, with the exception type, timestamp and any screenshot captured.
3. Those tasks are grouped under runtime exceptions and excluded from the denominators of mention rate, recommendation rate and similar metrics.
4. An explicit refusal by the platform, or Google not triggering an AI Overview, is a valid observation (refusal / not triggered) and is kept strictly separate from failure.

In the next cycle

We do not re-ask outside the cycle once it closes; doing so would move that cycle's sampling window. The next 72-hour batch runs the same frozen prompt set again, so a single failure costs at most one observation and never removes a prompt for the whole month.

Where to see it

• The monitoring log lists, per cycle, planned tasks, finalised tasks, valid observations, runtime exceptions (failed / blocked), unconfirmed results, cancelled and pending tasks, plus an evidence-completeness percentage and a runtime-exception rate.
• Every task drills down to its stable task ID, sampling time, status and screenshot.
• The public sample report at /report uses the same status vocabulary; failed and not-triggered tasks are shown as such.

What if one platform keeps failing cycle after cycle?

Repeated failures usually mean the platform changed its entry point or anti-automation behaviour. We flag it in the workspace and contact you by email or WeCom about next steps; meanwhile that platform's data stays marked as failed rather than being backfilled.

Are failed cycles refunded or credited?

Self-serve plans are a fixed monthly fee, not a fee per successful task; failed tasks do not consume quota and are not refunded individually. Project contracts can set an evidence-completeness threshold.

Last updated 2026-09-10. The capabilities page and the pricing page are authoritative where they differ.