diff --git a/README.md b/README.md
index 303edb1e0f2779d6a2cf1f147bcb0ad08cdcee4a..a7a869b3a66c6a6bf6c8092fd66822bef4aac218 100644
--- a/README.md
+++ b/README.md
@@ -3,6 +3,13 @@ The idea of Performance Model eXtractor (PMX) is described
 [here](http://se.informatik.uni-wuerzburg.de/tools/pmx/). 
 Please honor developers work by citing publications listed in the bibliography.
 
+# How to get monitoring traces
+* How to setup Kieker monitoring is described best at http://kieker-monitoring.net/
+* Note that accurate performance models can only be derived if you do not have side effects within your measurements. It may be worthwhile to inspect traces, e.g., using https://se3.informatik.uni-wuerzburg.de/kieker4dql/KiekerDQLAdapter
+* Two options
+** either monitor resource utilization in addition to response times 
+** or perform a calbration run that has no/low contention in it (which could be even without resource utilization logging)
+
 # Examples for builders
 For concrete builder implementations see
 * [PMX-DML](https://se3.informatik.uni-wuerzburg.de/descartes/pmx-dml)