diff --git a/crates/connect/src/dataframe.rs b/crates/connect/src/dataframe.rs
index a69f655..e67e2f6 100644
--- a/crates/connect/src/dataframe.rs
+++ b/crates/connect/src/dataframe.rs
@@ -730,7 +730,27 @@ impl DataFrame {
         DataFrameNaFunctions::new(self)
     }
 
-    // !TODO observe
+    /// Define (named) metrics to observe on the DataFrame.
+    ///
+    /// This method attaches metric expressions to the query plan. The metrics
+    /// are computed during query execution and can be retrieved from the
+    /// execution metrics.
+    ///
+    /// # Arguments
+    /// * `name` - Name of the observation
+    /// * `exprs` - Metric expressions (e.g., `count(lit(1))`, `avg(col("value"))`)
+    pub fn observe<I>(self, name: &str, exprs: I) -> DataFrame
+    where
+        I: IntoIterator,
+        I::Item: Into<Column>,
+    {
+        let plan = self.plan.observe(name, exprs);
+
+        DataFrame {
+            spark_session: self.spark_session,
+            plan,
+        }
+    }
 
     /// Returns a new [DataFrame] by skiping the first n rows
     pub fn offset(self, num: i32) -> DataFrame {
diff --git a/crates/connect/src/plan.rs b/crates/connect/src/plan.rs
index 30fabb5..ea36696 100644
--- a/crates/connect/src/plan.rs
+++ b/crates/connect/src/plan.rs
@@ -516,6 +516,23 @@ impl LogicalPlanBuilder {
         LogicalPlanBuilder::from(hint_rel)
     }
 
+    pub fn observe<I>(self, name: &str, exprs: I) -> LogicalPlanBuilder
+    where
+        I: IntoIterator,
+        I::Item: Into<Column>,
+    {
+        let metrics: Vec<spark::Expression> =
+            exprs.into_iter().map(|e| e.into().expression).collect();
+
+        let collect_metrics = RelType::CollectMetrics(Box::new(spark::CollectMetrics {
+            input: self.relation_input(),
+            name: name.to_string(),
+            metrics,
+        }));
+
+        LogicalPlanBuilder::from(collect_metrics)
+    }
+
     pub fn join<'a, T, I>(
         self,
         right: LogicalPlanBuilder,