diff --git a/src/benchmark/mod.rs b/src/benchmark/mod.rs
index 0699f7d..658494d 100644
--- a/src/benchmark/mod.rs
+++ b/src/benchmark/mod.rs
@@ -110,6 +110,14 @@ impl<'a> Benchmark<'a> {
         self.run_intermediate_command(command, error_output)
     }
 
+    /// Run the command specified by `--conclude`.
+    fn run_conclusion_command(&self, command: &Command<'_>) -> Result<TimingResult> {
+        let error_output = "The conclusion command terminated with a non-zero exit code. \
+                            Append ' || true' to the command if you are sure that this can be ignored.";
+
+        self.run_intermediate_command(command, error_output)
+    }
+
     /// Run the benchmark for a single command
     pub fn run(&self) -> Result<BenchmarkResult> {
         if self.options.output_style != OutputStyleOption::Disabled {
@@ -146,6 +154,25 @@ impl<'a> Benchmark<'a> {
                 .transpose()
         };
 
+        let conclusion_command = self.options.conclusion_command.as_ref().map(|values| {
+            let conclusion_command = if values.len() == 1 {
+                &values[0]
+            } else {
+                &values[self.number]
+            };
+            Command::new_parametrized(
+                None,
+                conclusion_command,
+                self.command.get_parameters().iter().cloned(),
+            )
+        });
+        let run_conclusion_command = || {
+            conclusion_command
+                .as_ref()
+                .map(|cmd| self.run_conclusion_command(cmd))
+                .transpose()
+        };
+
         self.run_setup_command(self.command.get_parameters().iter().cloned())?;
 
         // Warmup phase
@@ -163,6 +190,7 @@ impl<'a> Benchmark<'a> {
             for _ in 0..self.options.warmup_count {
                 let _ = run_preparation_command()?;
                 let _ = self.executor.run_command_and_measure(self.command, None)?;
+                let _ = run_conclusion_command()?;
                 if let Some(bar) = progress_bar.as_ref() {
                     bar.inc(1)
                 }
@@ -191,10 +219,16 @@ impl<'a> Benchmark<'a> {
         let (res, status) = self.executor.run_command_and_measure(self.command, None)?;
         let success = status.success();
 
+        let conclusion_result = run_conclusion_command()?;
+        let conclusion_overhead =
+            conclusion_result.map_or(0.0, |res| res.time_real + self.executor.time_overhead());
+
         // Determine number of benchmark runs
         let runs_in_min_time = (self.options.min_benchmarking_time
-            / (res.time_real + self.executor.time_overhead() + preparation_overhead))
-            as u64;
+            / (res.time_real
+                + self.executor.time_overhead()
+                + preparation_overhead
+                + conclusion_overhead)) as u64;
 
         let count = {
             let min = cmp::max(runs_in_min_time, self.options.run_bounds.min);
@@ -251,6 +285,8 @@ impl<'a> Benchmark<'a> {
             if let Some(bar) = progress_bar.as_ref() {
                 bar.inc(1)
             }
+
+            run_conclusion_command()?;
         }
 
         if let Some(bar) = progress_bar.as_ref() {
diff --git a/src/cli.rs b/src/cli.rs
index a3c0c7e..cc924cb 100644
--- a/src/cli.rs
+++ b/src/cli.rs
@@ -102,6 +102,23 @@ fn build_command() -> Command {
                      be run prior to the corresponding benchmark command.",
                 ),
         )
+        .arg(
+            Arg::new("conclude")
+                .long("conclude")
+                .short('C')
+                .action(ArgAction::Append)
+                .num_args(1)
+                .value_name("CMD")
+                .value_hint(ValueHint::CommandString)
+                .help(
+                    "Execute CMD after each timing run. This is useful for killing \
+                     long-running processes started (e.g. a web server started in --prepare), \
+                     for example.\nThe --conclude option can be specified once for all \
+                     commands or multiple times, once for each command. In the latter case, \
+                     each conclude command will be run after the corresponding benchmark \
+                     command.",
+                ),
+        )
         .arg(
             Arg::new("cleanup")
                 .long("cleanup")
diff --git a/src/options.rs b/src/options.rs
index 2f83889..43bb82b 100644
--- a/src/options.rs
+++ b/src/options.rs
@@ -207,6 +207,9 @@ pub struct Options {
     /// Command(s) to run before each timing run
     pub preparation_command: Option<Vec<String>>,
 
+    /// Command(s) to run after each timing run
+    pub conclusion_command: Option<Vec<String>>,
+
     /// Command to run before each *batch* of timing runs, i.e. before each individual benchmark
     pub setup_command: Option<String>,
 
@@ -243,6 +246,7 @@ impl Default for Options {
             min_benchmarking_time: 3.0,
             command_failure_action: CmdFailureAction::RaiseError,
             preparation_command: None,
+            conclusion_command: None,
             setup_command: None,
             cleanup_command: None,
             output_style: OutputStyleOption::Full,
@@ -304,6 +308,10 @@ impl Options {
             .get_many::<String>("prepare")
             .map(|values| values.map(String::from).collect::<Vec<String>>());
 
+        options.conclusion_command = matches
+            .get_many::<String>("conclude")
+            .map(|values| values.map(String::from).collect::<Vec<String>>());
+
         options.cleanup_command = matches.get_one::<String>("cleanup").map(String::from);
 
         options.command_output_policy = if matches.get_flag("show-output") {
@@ -432,6 +440,15 @@ impl Options {
             );
         }
 
+        if let Some(conclusion_command) = &self.conclusion_command {
+            ensure!(
+                conclusion_command.len() <= 1
+                    || commands.num_commands() == conclusion_command.len(),
+                "The '--conclude' option has to be provided just once or N times, where N is the \
+             number of benchmark commands."
+            );
+        }
+
         Ok(())
     }
 }
diff --git a/tests/execution_order_tests.rs b/tests/execution_order_tests.rs
index fbbfe4e..5404157 100644
--- a/tests/execution_order_tests.rs
+++ b/tests/execution_order_tests.rs
@@ -54,6 +54,11 @@ impl ExecutionOrderTest {
         self.command(output)
     }
 
+    fn conclude(&mut self, output: &str) -> &mut Self {
+        self.arg("--conclude");
+        self.command(output)
+    }
+
     fn cleanup(&mut self, output: &str) -> &mut Self {
         self.arg("--cleanup");
         self.command(output)
@@ -160,6 +165,24 @@ fn prepare_commands_are_executed_before_each_timing_run() {
         .run();
 }
 
+#[test]
+fn conclude_commands_are_executed_after_each_timing_run() {
+    ExecutionOrderTest::new()
+        .arg("--runs=2")
+        .conclude("conclude")
+        .command("command 1")
+        .command("command 2")
+        .expect_output("command 1")
+        .expect_output("conclude")
+        .expect_output("command 1")
+        .expect_output("conclude")
+        .expect_output("command 2")
+        .expect_output("conclude")
+        .expect_output("command 2")
+        .expect_output("conclude")
+        .run();
+}
+
 #[test]
 fn prepare_commands_are_executed_before_each_warmup() {
     ExecutionOrderTest::new()
@@ -187,6 +210,33 @@ fn prepare_commands_are_executed_before_each_warmup() {
         .run();
 }
 
+#[test]
+fn conclude_commands_are_executed_after_each_warmup() {
+    ExecutionOrderTest::new()
+        .arg("--warmup=2")
+        .arg("--runs=1")
+        .conclude("conclude")
+        .command("command 1")
+        .command("command 2")
+        // warmup 1
+        .expect_output("command 1")
+        .expect_output("conclude")
+        .expect_output("command 1")
+        .expect_output("conclude")
+        // benchmark 1
+        .expect_output("command 1")
+        .expect_output("conclude")
+        // warmup 2
+        .expect_output("command 2")
+        .expect_output("conclude")
+        .expect_output("command 2")
+        .expect_output("conclude")
+        // benchmark 2
+        .expect_output("command 2")
+        .expect_output("conclude")
+        .run();
+}
+
 #[test]
 fn cleanup_commands_are_executed_once_after_each_benchmark() {
     ExecutionOrderTest::new()
@@ -234,6 +284,44 @@ fn setup_prepare_cleanup_combined() {
         .run();
 }
 
+#[test]
+fn setup_prepare_conclude_cleanup_combined() {
+    ExecutionOrderTest::new()
+        .arg("--warmup=1")
+        .arg("--runs=2")
+        .setup("setup")
+        .prepare("prepare")
+        .command("command1")
+        .command("command2")
+        .conclude("conclude")
+        .cleanup("cleanup")
+        // 1
+        .expect_output("setup")
+        .expect_output("prepare")
+        .expect_output("command1")
+        .expect_output("conclude")
+        .expect_output("prepare")
+        .expect_output("command1")
+        .expect_output("conclude")
+        .expect_output("prepare")
+        .expect_output("command1")
+        .expect_output("conclude")
+        .expect_output("cleanup")
+        // 2
+        .expect_output("setup")
+        .expect_output("prepare")
+        .expect_output("command2")
+        .expect_output("conclude")
+        .expect_output("prepare")
+        .expect_output("command2")
+        .expect_output("conclude")
+        .expect_output("prepare")
+        .expect_output("command2")
+        .expect_output("conclude")
+        .expect_output("cleanup")
+        .run();
+}
+
 #[test]
 fn single_parameter_value() {
     ExecutionOrderTest::new()
diff --git a/tests/integration_tests.rs b/tests/integration_tests.rs
index 0c8dbf0..2bbde41 100644
--- a/tests/integration_tests.rs
+++ b/tests/integration_tests.rs
@@ -75,6 +75,31 @@ fn fails_with_wrong_number_of_prepare_options() {
         ));
 }
 
+#[test]
+fn fails_with_wrong_number_of_conclude_options() {
+    hyperfine()
+        .arg("--runs=1")
+        .arg("--conclude=echo a")
+        .arg("--conclude=echo b")
+        .arg("echo a")
+        .arg("echo b")
+        .assert()
+        .success();
+
+    hyperfine()
+        .arg("--runs=1")
+        .arg("--conclude=echo a")
+        .arg("--conclude=echo b")
+        .arg("echo a")
+        .arg("echo b")
+        .arg("echo c")
+        .assert()
+        .failure()
+        .stderr(predicate::str::contains(
+            "The '--conclude' option has to be provided",
+        ));
+}
+
 #[test]
 fn fails_with_duplicate_parameter_names() {
     hyperfine()
@@ -167,6 +192,18 @@ fn fails_for_unknown_prepare_command() {
         ));
 }
 
+#[test]
+fn fails_for_unknown_conclude_command() {
+    hyperfine()
+        .arg("--conclude=some-nonexisting-program-b5d9574198b7e4b12a71fa4747c0a577")
+        .arg("echo test")
+        .assert()
+        .failure()
+        .stderr(predicate::str::contains(
+            "The conclusion command terminated with a non-zero exit code.",
+        ));
+}
+
 #[cfg(unix)]
 #[test]
 fn can_run_failing_commands_with_ignore_failure_option() {
@@ -320,6 +357,48 @@ fn takes_preparation_command_into_account_for_computing_number_of_runs() {
         .stdout(predicate::str::contains("30 runs"));
 }
 
+#[test]
+fn takes_conclusion_command_into_account_for_computing_number_of_runs() {
+    hyperfine_debug()
+        .arg("--conclude=sleep 0.02")
+        .arg("sleep 0.01")
+        .assert()
+        .success()
+        .stdout(predicate::str::contains("100 runs"));
+
+    // Shell overhead needs to be added to both the conclude command and the actual command,
+    // leading to a total benchmark time of (cmd + shell + conclude + shell = 0.1 s)
+    hyperfine_debug()
+        .arg("--shell=sleep 0.01")
+        .arg("--conclude=sleep 0.03")
+        .arg("sleep 0.05")
+        .assert()
+        .success()
+        .stdout(predicate::str::contains("30 runs"));
+}
+
+#[test]
+fn takes_both_preparation_and_conclusion_command_into_account_for_computing_number_of_runs() {
+    hyperfine_debug()
+        .arg("--prepare=sleep 0.01")
+        .arg("--conclude=sleep 0.01")
+        .arg("sleep 0.01")
+        .assert()
+        .success()
+        .stdout(predicate::str::contains("100 runs"));
+
+    // Shell overhead needs to be added to both the prepare, conclude and the actual command,
+    // leading to a total benchmark time of (prepare + shell + cmd + shell + conclude + shell = 0.1 s)
+    hyperfine_debug()
+        .arg("--shell=sleep 0.01")
+        .arg("--prepare=sleep 0.01")
+        .arg("--conclude=sleep 0.01")
+        .arg("sleep 0.05")
+        .assert()
+        .success()
+        .stdout(predicate::str::contains("30 runs"));
+}
+
 #[test]
 fn shows_benchmark_comparison_with_relative_times() {
     hyperfine_debug()