Ask HN: Why both distributed tracing and logging (for backend services)?

I know most observability companies (Datadog, New Relic, Elastic, grafana, ...) will want us to believe that we need *both*. But why?

Isn't tracing a superset of logs? Instead of logging won't it be better to add some attributes to traces?

I used ddog apm, and GCP stackdriver too. But I believe if I got the query capabilities of stackdriver in ddog, I would have saved few $k's in monthly infra expense.

Or am I missing something?