Difference between Zipkin and Elastic Stack(ELK)?

Spring Cloud Sleuth is used for creating traceIds (Unique to request across services) and spanId (Same for one unit for work). My idea is that Zipkin server is used to get collective visualization of these logs across service. But I know and have used ELK stack which does necessarily the same function. I mean we can group requests with the same traceId for visualising, using ELK stack. But I do see people trying to implement distributed tracing with Sleuth, ELK along with Zipkin, as in these examples (Link1,Link2). But why do we need Zipkin if there is already ELK for log collection and visualising? Where I am missing?