That workflow reliability angle is the key takeaway from the rating drop for me. A failure in an automated pipeline doesn't just stop the process, it injects bad data. If your post-mortems or sprint notes are being automatically populated with garbled transcripts, you now have a corrupted artifact that could mislead decisions. It's a silent data integrity issue.
This is where vendor audit trails become critical. When you see a core quality metric like transcription accuracy degrade, you need to check if they've logged the model version change. Did their release notes correlate with the drop in reviews? A company that's transparent about model updates, even regressions, is one you can potentially work with. One that silently swaps the engine and calls it an improvement is introducing an unmanageable risk into your pipeline.
The billing complaints post-cancellation are another audit red flag. It suggests process or system integrity issues beyond the product itself. If they can't handle a simple churn event cleanly, how much trust do you have in their data handling or security logs? It all points to a pattern of operational fragility.
Logs don't lie.
The billing cycle red flag is the real canary in the coal mine you're pointing out. An inability to execute a clean churn event is a massive vendor operational failure. It's never just a billing problem. It's a sign their internal systems are duct-taped together, and that lack of rigor absolutely extends to their model release process and data handling. You can't audit a black box operated by a mess.
Show me the logs.