Status: Resolved
On 14 September 2026, some customers saw scores and alerts take longer than usual to appear against their calls, for a period of a few hours, because of a higher than usual volume of calls needing analysis at the same time. Call transcription itself wasn't affected. We increased processing capacity, monitored the backlog until it fully cleared, and confirmed there was no recurrence before closing the incident. No data was lost.
Some customers experienced a delay of a few hours before scores and alerts appeared against recently completed calls. Calls were still being transcribed and made available as normal throughout; only the analysis stage, scoring and alert generation, was delayed. All delayed analysis completed once the backlog cleared, so nothing was lost or skipped.
10:37: An unusually high processing queue was noticed and flagged for monitoring.
11:26: With the queue continuing to grow and affecting how quickly calls were being analysed for some customers, the incident was declared so we could let affected customers know we were aware and on it.
12:27: The queue was confirmed to be steadily decreasing.
13:52: The backlog had cleared. We continued monitoring for a couple of hours to make sure it didn't build up again.
15:51: No recurrence confirmed, and the incident was closed.
The delay was driven by a higher than usual volume of calls needing analysis at the same time, which built up a backlog faster than it was clearing under normal processing capacity.
We increased processing capacity for call analysis so the backlog could clear faster, and monitored it closely until it reached zero. We then continued watching for a further couple of hours to confirm there was no recurrence before closing the incident.
Everything is processing normally. We'll keep monitoring for similar spikes in volume and adjusting capacity as needed.
