Hi Everyone, My company is considering Druid for a fraud detection use case and are keen to hear opinions on our particular usage scenario. We will be ingesting 400k+ events each day. Immediately after ingesting each of those events we want to run approximately 600 queries. We are looking for various kinds of counts across time-periods from 1 hour through to 1 year. We will then compare those counts against a threshold to report the "triggering" of a fraud rule. An example set of rules is the number of times a person, identified by "first name", "last name" and "date of birth", has created accounts in a particular industry over the last hour, day, week, month, quarter and year. Then if the count is greater than, say, 2, 5, 7, 10, then we will return a potential fraud flag. Druid experts, do you think Druid is a suitable technology for this use case? If so, do you have any advice that would help us? Thank you! Michael

Yep, we have several other folks doing fraud detection as well. The only thing to keep in mind is Druid doesn't have a "trigger" mechanism, so that will have to be orchestrated externally. Other than that, querying real-time data and running high frequency low latencies are no problem. https://imply.io/podcast/fraud-fighters-apache-druid-imply-help-ibotta-combat-fraud-faster-anomaly-detection-jaylyn-stoesz/

Hi Benjamin - thanks for the quick response. I'll have a read of that case study. Cheers - M