diff --git a/reduce/report/report.pdf b/reduce/report/report.pdf
index bf51283..025ea31 100755
Binary files a/reduce/report/report.pdf and b/reduce/report/report.pdf differ
diff --git a/reduce/report/report.tex b/reduce/report/report.tex
index 86740cc..1b99efd 100755
--- a/reduce/report/report.tex
+++ b/reduce/report/report.tex
@@ -375,6 +375,15 @@ The number of repetitions is the same as for the last test at 30.
 \section{Analysis}
 \label{sec:analysis}
 
+Our first result seen in \prettyref{fig:nodeplot} suggests that our implementation using the binomial tree got the best results by a big margin.
+This result was very surprising to us because we expected that the MPI\_Reduce function of the library would outperform our rather simple implementations.
+However this seems not to be the case and such tree algorithms are apparently really good for the dataset tested there.
+Although the result of the MPI\_Reduce function seems to very unstable and it varies a lot during the test.
+This might be due to a too low number of repetitions, the very short execution time or some other factors.
+That the binary tree performed better than the Fibonacci tree was also quite surprising, since the communication pattern of the Fibonacci tree is almost round optimal in contrast to the binary tree.
+
+\newpage
+
 \section{Appendix}
 \label{sec:appendix}