diff --git a/cmd/benchstat/main_test.go b/cmd/benchstat/main_test.go
index 331b29f..8bdd58b 100644
--- a/cmd/benchstat/main_test.go
+++ b/cmd/benchstat/main_test.go
@@ -39,6 +39,7 @@
 func check(t *testing.T, name string, files ...string) {
 	t.Run(name, func(t *testing.T) {
 		os.Args = append([]string{"benchstat"}, files...)
+		t.Logf("running %v", os.Args)
 		r, w, err := os.Pipe()
 		if err != nil {
 			t.Fatal(err)
diff --git a/cmd/benchstat/testdata/examplehtml.golden b/cmd/benchstat/testdata/examplehtml.golden
index 602f5fb..193ae06 100644
--- a/cmd/benchstat/testdata/examplehtml.golden
+++ b/cmd/benchstat/testdata/examplehtml.golden
@@ -8,19 +8,19 @@
 .benchstat .worse td.delta { font-weight: bold; color: #c00; }
 </style>
 
-
 <table class='benchstat oldnew'>
+<tr class='configs'><th><th>exampleold.txt<th>examplenew.txt
 
 
 <tbody>
-<tr><th><th>old time/op<th>new time/op<th>delta<th>
+<tr><th><th colspan='2' class='metric'>time/op<th>delta
 <tr class='better'><td>GobEncode<td>13.6ms ± 1%<td>11.8ms ± 1%<td class='delta'>−13.31%<td class='note'>(p=0.016 n=4&#43;5)
 <tr class='unchanged'><td>JSONEncode<td>32.1ms ± 1%<td>31.8ms ± 1%<td class='nodelta'>~<td class='note'>(p=0.286 n=4&#43;5)
 <tr><td>&nbsp;
 </tbody>
 
 <tbody>
-<tr><th><th>old speed<th>new speed<th>delta<th>
+<tr><th><th colspan='2' class='metric'>speed<th>delta
 <tr class='better'><td>GobEncode<td>56.4MB/s ± 1%<td>65.1MB/s ± 1%<td class='delta'>&#43;15.36%<td class='note'>(p=0.016 n=4&#43;5)
 <tr class='unchanged'><td>JSONEncode<td>60.4MB/s ± 1%<td>61.1MB/s ± 2%<td class='nodelta'>~<td class='note'>(p=0.286 n=4&#43;5)
 <tr><td>&nbsp;
diff --git a/cmd/benchstat/testdata/exampleoldhtml.golden b/cmd/benchstat/testdata/exampleoldhtml.golden
index 8dcb5c9..d7c651c 100644
--- a/cmd/benchstat/testdata/exampleoldhtml.golden
+++ b/cmd/benchstat/testdata/exampleoldhtml.golden
@@ -8,7 +8,6 @@
 .benchstat .worse td.delta { font-weight: bold; color: #c00; }
 </style>
 
-
 <table class='benchstat '>
 
 
diff --git a/cmd/benchstat/testdata/oldnew4html.golden b/cmd/benchstat/testdata/oldnew4html.golden
index 43453b2..2f47768 100644
--- a/cmd/benchstat/testdata/oldnew4html.golden
+++ b/cmd/benchstat/testdata/oldnew4html.golden
@@ -8,7 +8,6 @@
 .benchstat .worse td.delta { font-weight: bold; color: #c00; }
 </style>
 
-
 <table class='benchstat '>
 <tr class='configs'><th><th>old.txt<th>new.txt<th>slashslash4.txt
 
diff --git a/cmd/benchstat/testdata/oldnewhtml.golden b/cmd/benchstat/testdata/oldnewhtml.golden
index be0b0e0..5eb7a2a 100644
--- a/cmd/benchstat/testdata/oldnewhtml.golden
+++ b/cmd/benchstat/testdata/oldnewhtml.golden
@@ -8,12 +8,12 @@
 .benchstat .worse td.delta { font-weight: bold; color: #c00; }
 </style>
 
-
 <table class='benchstat oldnew'>
+<tr class='configs'><th><th>old.txt<th>new.txt
 
 
 <tbody>
-<tr><th><th>old time/op<th>new time/op<th>delta<th>
+<tr><th><th colspan='2' class='metric'>time/op<th>delta
 <tr class='better'><td>CRC32/poly=IEEE/size=15/align=0-8<td>46.9ns ± 8%<td>44.5ns ± 3%<td class='delta'>−5.01%<td class='note'>(p=0.008 n=10&#43;10)
 <tr class='unchanged'><td>CRC32/poly=IEEE/size=15/align=1-8<td>44.7ns ± 5%<td>44.5ns ± 4%<td class='nodelta'>~<td class='note'>(p=0.539 n=10&#43;10)
 <tr class='worse'><td>CRC32/poly=IEEE/size=40/align=0-8<td>41.0ns ± 1%<td>42.5ns ± 6%<td class='delta'>&#43;3.56%<td class='note'>(p=0.000 n=8&#43;10)
@@ -54,7 +54,7 @@
 </tbody>
 
 <tbody>
-<tr><th><th>old speed<th>new speed<th>delta<th>
+<tr><th><th colspan='2' class='metric'>speed<th>delta
 <tr class='better'><td>CRC32/poly=IEEE/size=15/align=0-8<td>321MB/s ± 8%<td>337MB/s ± 3%<td class='delta'>&#43;5.06%<td class='note'>(p=0.009 n=10&#43;10)
 <tr class='unchanged'><td>CRC32/poly=IEEE/size=15/align=1-8<td>336MB/s ± 4%<td>337MB/s ± 4%<td class='nodelta'>~<td class='note'>(p=0.579 n=10&#43;10)
 <tr class='worse'><td>CRC32/poly=IEEE/size=40/align=0-8<td>975MB/s ± 1%<td>942MB/s ± 5%<td class='delta'>−3.37%<td class='note'>(p=0.001 n=8&#43;10)