diff --git a/asap-planner-rs/src/optimizer/aqe_extractor.rs b/asap-planner-rs/src/optimizer/aqe_extractor.rs
index 3e4a2d48..15ed2064 100644
--- a/asap-planner-rs/src/optimizer/aqe_extractor.rs
+++ b/asap-planner-rs/src/optimizer/aqe_extractor.rs
@@ -2,6 +2,8 @@ use std::collections::HashMap;
 
 use asap_types::query_requirements::{build_query_requirements_promql, QueryRequirements};
 use asap_types::PromQLSchema;
+use promql_parser::parser::token::{self, TokenType};
+use promql_parser::parser::{AggregateExpr, Call, Expr, Function};
 use promql_utilities::data_model::KeyByLabelNames;
 use promql_utilities::query_logics::enums::Statistic;
 
@@ -138,10 +140,11 @@ pub(super) fn gcd(a: u64, b: u64) -> u64 {
 /// into their arms. Scalar arms (e.g. the `100` in `rate(x[5m]) * 100`) are
 /// dropped — they contribute no AQE. Only arithmetic operators are split;
 /// comparison and set operators are left as-is (treated as opaque leaves).
+/// An `avg` leaf becomes its sum and count leaves (see `rewrite_avg`).
 fn decompose_to_leaves(query: &str) -> Vec<String> {
     let (lhs, rhs) = match parse_binary_arms(query) {
         Some(arms) => arms,
-        None => return vec![query.to_string()],
+        None => return rewrite_avg(query),
     };
 
     let mut leaves = Vec::new();
@@ -153,6 +156,42 @@ fn decompose_to_leaves(query: &str) -> Vec<String> {
     leaves
 }
 
+/// Rewrite an `avg` leaf into the sum and count leaves it is computed from:
+/// `avg by (l) (x)` → `sum by (l) (x)`, `count by (l) (x)`, and
+/// `avg_over_time(x[5m])` → `sum_over_time(x[5m])`, `count_over_time(x[5m])`.
+/// Any other query is returned unchanged.
+fn rewrite_avg(query: &str) -> Vec<String> {
+    match promql_parser::parser::parse(query) {
+        Ok(Expr::Aggregate(agg)) if agg.op.id() == token::T_AVG => [token::T_SUM, token::T_COUNT]
+            .into_iter()
+            .map(|op| {
+                Expr::Aggregate(AggregateExpr {
+                    op: TokenType::new(op),
+                    ..agg.clone()
+                })
+                .to_string()
+            })
+            .collect(),
+        // sum_over_time and count_over_time share avg_over_time's signature.
+        Ok(Expr::Call(call)) if call.func.name == "avg_over_time" => {
+            ["sum_over_time", "count_over_time"]
+                .into_iter()
+                .map(|name| {
+                    Expr::Call(Call {
+                        func: Function {
+                            name,
+                            ..call.func.clone()
+                        },
+                        args: call.args.clone(),
+                    })
+                    .to_string()
+                })
+                .collect()
+        }
+        _ => vec![query.to_string()],
+    }
+}
+
 /// Try to extract `QueryRequirements` from a single leaf PromQL query string.
 /// Returns `None` if the query cannot be parsed or does not match any pattern.
 fn extract_requirements(
@@ -301,6 +340,74 @@ mod tests {
         ));
     }
 
+    fn sorted_statistics_and_queries(
+        items: &[OptimizerItem],
+    ) -> Vec<(Vec<Statistic>, Vec<String>)> {
+        let mut out: Vec<_> = items
+            .iter()
+            .map(|i| (i.requirements.statistics.clone(), i.query_strings.clone()))
+            .collect();
+        out.sort_by_key(|(_, q)| q.clone());
+        out
+    }
+
+    // A multi-statistic [Sum, Count] item has no single-sketch candidate, so
+    // avg must reach the optimizer as separate sum and count items.
+    #[test]
+    fn spatial_avg_becomes_sum_and_count_items() {
+        let rqes = vec![rqe("avg by (job) (metric)", 60_000)];
+        let items = extract_aqes(&rqes, &empty_schema(), 15_000).unwrap();
+        assert_eq!(
+            sorted_statistics_and_queries(&items),
+            vec![
+                (
+                    vec![Statistic::Count],
+                    vec!["count by (job) (metric)".into()]
+                ),
+                (vec![Statistic::Sum], vec!["sum by (job) (metric)".into()]),
+            ]
+        );
+        for item in &items {
+            assert_eq!(
+                item.requirements.grouping_labels,
+                KeyByLabelNames::new(vec!["job".into()])
+            );
+        }
+    }
+
+    #[test]
+    fn avg_over_time_becomes_sum_and_count_over_time_items() {
+        let rqes = vec![rqe("avg_over_time(metric[5m])", 60_000)];
+        let items = extract_aqes(&rqes, &empty_schema(), 15_000).unwrap();
+        assert_eq!(
+            sorted_statistics_and_queries(&items),
+            vec![
+                (
+                    vec![Statistic::Count],
+                    vec!["count_over_time(metric[5m])".into()]
+                ),
+                (
+                    vec![Statistic::Sum],
+                    vec!["sum_over_time(metric[5m])".into()]
+                ),
+            ]
+        );
+        for item in &items {
+            assert_eq!(item.requirements.data_range_ms, 300_000);
+        }
+    }
+
+    #[test]
+    fn avg_arm_of_binary_query_is_rewritten() {
+        let rqes = vec![rqe(
+            "avg_over_time(metric_a[5m]) / sum_over_time(metric_b[5m])",
+            60_000,
+        )];
+        let items = extract_aqes(&rqes, &empty_schema(), 15_000).unwrap();
+        assert_eq!(items.len(), 3);
+        assert!(items.iter().all(|i| i.requirements.statistics.len() == 1));
+    }
+
     #[test]
     fn spatial_only_query_sets_range_to_scrape_interval() {
         let rqes = vec![rqe("sum(metric)", 60_000)];