diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index b1b15bacd..12f7f21a6 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -1341,8 +1341,8 @@ Here are some useful optional tools:
 
 Eg:
 
-    stack install ghcid shelltestrunner quickbench hasktags
-    brew install entr
+    stack install ghcid shelltestrunner hasktags
+    git clone https://github.com/simonmichael/quickbench; cd quickbench; stack install  # must run in source dir
 
 ### Get the code
 
diff --git a/Makefile b/Makefile
index 1e6fd91ee..a57fe85c7 100644
--- a/Makefile
+++ b/Makefile
@@ -648,7 +648,7 @@ bench: quickbench
 
 quickbench: samplejournals bench.sh $(call def-help,quickbench, benchmark commands in bench.sh with quickbench and $(BENCHEXES))
 	@echo; echo "run quick performance benchmarks in bench.sh (approximate, can be skewed):"
-	@which -s quickbench && quickbench -w $(BENCHEXES) || echo "quickbench not installed, skipping"
+	@which -s quickbench && quickbench -w $(BENCHEXES) || echo "quickbench not installed (see bench.sh), skipping"
 
 # bench: samplejournals tests/bench.tests tools/simplebench \
 # 	$(call def-help,bench,\
diff --git a/bench.sh b/bench.sh
index e2895fb07..d6296cb46 100755
--- a/bench.sh
+++ b/bench.sh
@@ -4,15 +4,23 @@
 # can be affected by system activity. Usually this isn't a problem. The last
 # (criterion) is more robust.
 # 
-# Generate the test journals: make samplejournals 
-# Get quickbench: stack install quickbench
+# Generate the test journals: 
+# make samplejournals 
+#
+# Get quickbench: 
+# git clone https://github.com/simonmichael/quickbench
+# cd quickbench
+# stack install  # must be run in source dir
+#
 # Measure performance:
-# time ./bench.sh      # show if these work, what they do, total time
+# time ./bench.sh      # show if these work, what they do, total time (GNU time also shows max memory)
 # quickbench [OPTS]    # time each command, one or more times
 # make bench           # time each command with several hledger versions (BENCHEXES in Makefile) 
 # stack bench hledger  # time a different set of benchmarks (bench/bench.hs) 
 # stack bench hledger --ba --criterion  # time more carefully, using criterion 
 
+# commands to benchmark:
+
 # hledger -f examples/100x100x10.journal print
 # hledger -f examples/1000x1000x10.journal print
 hledger -f examples/10000x1000x10.journal print