1. At 00:30 on Saturday the 17th of August our automated monitoring notified our engineering team that targeting rates had dropped below a predefined threshold for a single customer. Our engineers investigated the issue and confirmed that there was an error being produced for that customer under a limited set of conditions 2. Looking through our logs we realised this issue had started occurring at 03:00 the previous day (Friday the 16th of August). It had originally occurred with minimal impact and had slowly grown over the intervening period until it triggered our alert once it reached predefined threshold. We identified that it had been related to an anomalous traffic pattern we had identified at 18:00 on Friday the 16th 3. Further, at 00:50 we identified a second customer where this issue had started manifesting, but were still below the threshold that would trigger our alerts 4. At 09:20 our engineers identified the root cause of the issue, which occurred when those two customers used a particular type of segmentation query in combination with a recent change we'd made to our Query Language and deployed that week 5. A fix was put in place to our Query Language at 9:52AM London time and rolled out to all effected customers by 11:30am London time on Saturday the 17th of August
Most of our customers were unaffected by this issue. If you were affected your Customer Success Manager will reach out to you with further information.