diff --git a/atlasbot/engine/answerer.py b/atlasbot/engine/answerer.py
index 187caa6..835981d 100644
--- a/atlasbot/engine/answerer.py
+++ b/atlasbot/engine/answerer.py
@@ -78,6 +78,7 @@ class ModePlan:
     chunk_group: int
     parallelism: int
     score_retries: int
+    use_deep_retrieval: bool
     use_tool: bool
     use_critic: bool
     use_gap: bool
@@ -360,57 +361,59 @@ class AnswerEngine:
                 )
                 metric_facts: list[str] = []
                 if classify.get("question_type") in {"metric", "diagnostic"} or force_metric:
-                    if observer:
-                        observer("retrieve", "extracting fact types")
-                    fact_types = await _extract_fact_types(
-                        call_llm,
-                        normalized,
-                        keyword_tokens,
-                        plan,
-                    )
-                    if observer:
-                        observer("retrieve", "deriving signals")
-                    signals = await _derive_signals(
-                        call_llm,
-                        normalized,
-                        fact_types,
-                        plan,
-                    )
-                    if isinstance(signals, list):
-                        signal_tokens = [str(item) for item in signals if item]
                     all_tokens = _merge_tokens(signal_tokens, keyword_tokens, question_tokens)
-                    if observer:
-                        observer("retrieve", "scanning chunks")
-                    candidate_lines: list[str] = []
-                    if signals:
-                        for chunk in selected:
-                            chunk_lines = chunk["text"].splitlines()
-                            if not chunk_lines:
-                                continue
-                            hits = await _scan_chunk_for_signals(
-                                call_llm,
-                                normalized,
-                                signals,
-                                chunk_lines,
-                                plan,
-                            )
-                            if hits:
-                                candidate_lines.extend(hits)
-                    candidate_lines = list(dict.fromkeys(candidate_lines))
-                    if candidate_lines:
+                    if plan.use_deep_retrieval:
                         if observer:
-                            observer("retrieve", "pruning candidates")
-                        metric_facts = await _prune_metric_candidates(
+                            observer("retrieve", "extracting fact types")
+                        fact_types = await _extract_fact_types(
                             call_llm,
                             normalized,
-                            candidate_lines,
+                            keyword_tokens,
                             plan,
-                            plan.metric_retries,
                         )
-                        if metric_facts:
-                            key_facts = _merge_fact_lines(metric_facts, key_facts)
-                            if self._settings.debug_pipeline:
-                                _debug_log("metric_facts_selected", {"facts": metric_facts})
+                        if observer:
+                            observer("retrieve", "deriving signals")
+                        signals = await _derive_signals(
+                            call_llm,
+                            normalized,
+                            fact_types,
+                            plan,
+                        )
+                        if isinstance(signals, list):
+                            signal_tokens = [str(item) for item in signals if item]
+                            all_tokens = _merge_tokens(signal_tokens, keyword_tokens, question_tokens)
+                        if observer:
+                            observer("retrieve", "scanning chunks")
+                        candidate_lines: list[str] = []
+                        if signals:
+                            for chunk in selected:
+                                chunk_lines = chunk["text"].splitlines()
+                                if not chunk_lines:
+                                    continue
+                                hits = await _scan_chunk_for_signals(
+                                    call_llm,
+                                    normalized,
+                                    signals,
+                                    chunk_lines,
+                                    plan,
+                                )
+                                if hits:
+                                    candidate_lines.extend(hits)
+                        candidate_lines = list(dict.fromkeys(candidate_lines))
+                        if candidate_lines:
+                            if observer:
+                                observer("retrieve", "pruning candidates")
+                            metric_facts = await _prune_metric_candidates(
+                                call_llm,
+                                normalized,
+                                candidate_lines,
+                                plan,
+                                plan.metric_retries,
+                            )
+                            if metric_facts:
+                                key_facts = _merge_fact_lines(metric_facts, key_facts)
+                                if self._settings.debug_pipeline:
+                                    _debug_log("metric_facts_selected", {"facts": metric_facts})
                     if not metric_facts:
                         if observer:
                             observer("retrieve", "fallback metric selection")
@@ -1074,6 +1077,7 @@ def _mode_plan(settings: Settings, mode: str) -> ModePlan:
             chunk_group=4,
             parallelism=4,
             score_retries=3,
+            use_deep_retrieval=True,
             use_tool=True,
             use_critic=True,
             use_gap=True,
@@ -1092,6 +1096,7 @@ def _mode_plan(settings: Settings, mode: str) -> ModePlan:
             chunk_group=4,
             parallelism=2,
             score_retries=2,
+            use_deep_retrieval=True,
             use_tool=True,
             use_critic=True,
             use_gap=True,
@@ -1109,6 +1114,7 @@ def _mode_plan(settings: Settings, mode: str) -> ModePlan:
         chunk_group=5,
         parallelism=1,
         score_retries=1,
+        use_deep_retrieval=False,
         use_tool=False,
         use_critic=False,
         use_gap=False,