For a bank in particular, data is the most important asset. However, this data has become very large over the years. Within the last year we have seen a lot: over 8,000 row datasets, Map of Maps, Array of Structs. We've been able to do many things with KSQL that we would normally have used kStreams for. We’ve learned a lot and tried many different things: From invocation functions to custom types. We built a complex streaming architecture to create a highly integrated data landscpae. This talk is meant to summarise the high potential and challenges of ksql that we encountered. At the end, we also want to give a brief outlook on why we are looking forward to Apache Flink.