diff --git a/atlasbot/engine/answerer.py b/atlasbot/engine/answerer.py
index 78974c5..b30862f 100644
--- a/atlasbot/engine/answerer.py
+++ b/atlasbot/engine/answerer.py
@@ -215,6 +215,7 @@ class AnswerEngine:
             classify.setdefault("focus_entity", "unknown")
             classify.setdefault("focus_metric", "unknown")
             _debug_log("route_parsed", {"classify": classify, "normalized": normalized})
+            force_metric = bool(re.search(r"\bhow many\b|\bcount\b|\btotal\b", normalized.lower()))
             cluster_terms = (
                 "atlas",
                 "cluster",
@@ -305,7 +306,7 @@ class AnswerEngine:
                     max_lines=max(4, plan.max_subquestions * 2),
                 )
                 metric_facts: list[str] = []
-                if classify.get("question_type") in {"metric", "diagnostic"}:
+                if classify.get("question_type") in {"metric", "diagnostic"} or force_metric:
                     if observer:
                         observer("retrieve", "extracting fact types")
                     fact_types = await _extract_fact_types(
@@ -541,7 +542,7 @@ class AnswerEngine:
                     model=plan.model,
                     tag="focus_fix",
                 )
-            if classify.get("question_type") in {"metric", "diagnostic"} and metric_facts:
+            if (classify.get("question_type") in {"metric", "diagnostic"} or force_metric) and metric_facts:
                 best_line = None
                 lowered_keywords = [kw.lower() for kw in keyword_tokens if kw]
                 for line in metric_facts: