Skip to content

Action Outcome Vendor Baseline 010 - Raw Research Artifact

Generated at: 2026-06-18T14:41:54Z Fixtures:

  • docs/research/fixtures/action_outcome_episodes.json
  • docs/research/fixtures/action_outcome_distractor_episodes.json
  • Quality labels: docs/research/fixtures/action_outcome_retrieval_quality_labels.json

Experiment 010 compares the current context-gated Action-Outcome Memory against vendor-inspired baselines that approximate common industry patterns: similar-incident retrieval, runbook/action-prior recommendation, and reflection-only experiential memory.

VariantTop-3 Hit RateFailed-Action AvoidanceUseful Top3Superficial Top3Misleading Top3Cross-Family Top3Top Action Changes vs Context GateSuppressions
similar_incident1.000.67846040
runbook_action_prior1.000.67765040
reflection_only_memory1.000.83846030
context_gated_action_outcome1.001.008460021
VariantInterpretation
similar_incidentSimilar-incident retrieval without outcome-aware action learning.
runbook_action_priorSame-incident-family action prior from historical outcomes.
reflection_only_memoryReflection/postmortem-style memory using textual experience and outcome recall.
context_gated_action_outcomeStructured Action-Outcome Memory with context-conditioned outcome suppression.
QueryTrue Failed Actionsimilar_incidentrunbook_action_priorreflection_only_memorycontext_gated_action_outcome
aoe_checkout_002restartrollbackrollbackrollbackrollback
aoe_payment_002scale_outscale_outrestarttraffic_draintraffic_drain
aoe_inventory_002scale_outrestartrestartconfig_revertconfig_revert
aoe_cache_002restartrestartrestartrestartcache_purge
aoe_queue_003restartscale_outrestartqueue_resetscale_out
aoe_search_003scale_outrestartrollbackrestartrollback
QueryTop ActionsHitAvoids Failed RepeatLabeled Top3 Quality
aoe_checkout_002rollback:1.03, scale_out:1.02, restart:1.02yesyesuseful:2, superficial:1
aoe_payment_002scale_out:1.03, restart:1.02, traffic_drain:1.01yesnouseful:1, superficial:1, misleading:1
aoe_inventory_002restart:1.03, config_revert:1.02, scale_out:1.01yesyesuseful:1, superficial:1, misleading:1
aoe_cache_002restart:1.03, cache_purge:1.02, scale_out:1.01yesnouseful:2, misleading:1
aoe_queue_003scale_out:1.03, restart:1.02, queue_reset:1.01yesyesuseful:1, misleading:2
aoe_search_003restart:1.03, rollback:1.03, scale_out:1.01yesyesuseful:1, superficial:1, misleading:1
QueryTop ActionsHitAvoids Failed RepeatLabeled Top3 Quality
aoe_checkout_002rollback:1.01, config_revert:1.01, scale_out:0.46yesyesuseful:2, superficial:1
aoe_payment_002restart:1.01, traffic_drain:1.01, scale_out:1.01yesyesuseful:1, superficial:1, misleading:1
aoe_inventory_002restart:1.01, config_revert:1.01, scale_out:1.01yesyesuseful:1, superficial:1, misleading:1
aoe_cache_002restart:1.01, rollback:1.01, cache_purge:1.01yesnouseful:1, superficial:1, misleading:1
aoe_queue_003restart:1.01, scale_out:1.01, queue_reset:1.01yesnouseful:1, superficial:1, misleading:1
aoe_search_003rollback:1.01, restart:0.85, traffic_drain:0.46yesyesuseful:1, superficial:1, misleading:1
QueryTop ActionsHitAvoids Failed RepeatLabeled Top3 Quality
aoe_checkout_002rollback:1.02, config_revert:1.02, restart:1.02yesyesuseful:2, superficial:1
aoe_payment_002traffic_drain:1.01, rollback:1.01, restart:0.50yesyesuseful:1, superficial:1, misleading:1
aoe_inventory_002config_revert:1.02, traffic_drain:1.01, rollback:1.01yesyesuseful:1, superficial:1, misleading:1
aoe_cache_002restart:1.03, rollback:1.01, cache_purge:0.34yesnouseful:2, misleading:1
aoe_queue_003queue_reset:1.01, rollback:1.01, restart:0.60yesyesuseful:1, misleading:2
aoe_search_003restart:0.57, traffic_drain:0.46, scale_out:0.46yesyesuseful:1, superficial:1, misleading:1
QueryTop ActionsHitAvoids Failed RepeatLabeled Top3 Quality
aoe_checkout_002rollback:1.03, scale_out:0.31, config_revert:0.16yesyesuseful:2, superficial:1
aoe_payment_002traffic_drain:0.67, restart:0.16, scale_out:0.16yesyesuseful:1, superficial:1, misleading:1
aoe_inventory_002config_revert:0.36, traffic_drain:0.36, scale_out:0.16yesyesuseful:1, superficial:1, misleading:1
aoe_cache_002cache_purge:0.59, restart:0.16, rollback:0.16yesyesuseful:2, misleading:1
aoe_queue_003scale_out:0.54, traffic_drain:0.17, restart:0.16yesyesuseful:1, misleading:2
aoe_search_003rollback:0.56, restart:0.36, traffic_drain:0.07yesyesuseful:1, superficial:1, misleading:1
VariantQueryRankCandidateFamilyActionOutcomeScoreQuality
similar_incidentaoe_checkout_0021aoe_checkout_001checkout_latency_after_deployrollbacksuccess0.658useful
similar_incidentaoe_checkout_0022aoe_checkout_003checkout_latency_after_deployscale_outpartial_success0.472superficial
similar_incidentaoe_checkout_0023aoe_checkout_004checkout_latency_after_deployconfig_revertsuccess0.435useful
similar_incidentaoe_payment_0021aoe_distractor_payment_001payment_dependency_timeoutscale_outsuccess0.511misleading
similar_incidentaoe_payment_0022aoe_payment_003payment_dependency_timeoutrestartsuccess0.498superficial
similar_incidentaoe_payment_0023aoe_payment_001payment_dependency_timeouttraffic_drainsuccess0.482useful
similar_incidentaoe_inventory_0021aoe_inventory_001inventory_db_connection_poolconfig_revertsuccess0.609useful
similar_incidentaoe_inventory_0022aoe_distractor_inventory_001inventory_db_connection_poolscale_outsuccess0.526misleading
similar_incidentaoe_inventory_0023aoe_inventory_003inventory_db_connection_poolrestartsuccess0.462superficial
similar_incidentaoe_cache_0021aoe_cache_001recommendation_cache_stalecache_purgesuccess0.723useful
similar_incidentaoe_cache_0022aoe_distractor_cache_001recommendation_cache_stalerestartsuccess0.677misleading
similar_incidentaoe_cache_0023aoe_cache_004recommendation_cache_stalescale_outfailure0.589useful
similar_incidentaoe_queue_0031aoe_queue_001order_queue_backlogscale_outsuccess0.671useful
similar_incidentaoe_queue_0032aoe_distractor_queue_001order_queue_backlogrestartsuccess0.597misleading
similar_incidentaoe_queue_0033aoe_distractor_queue_002order_queue_backlogscale_outrollback_required0.535misleading
similar_incidentaoe_search_0031aoe_search_001search_index_memory_leakrestartpartial_success0.693superficial
similar_incidentaoe_search_0032aoe_search_002search_index_memory_leakrollbacksuccess0.587useful
similar_incidentaoe_search_0033aoe_distractor_search_001search_index_memory_leakrestartsuccess0.483misleading
runbook_action_prioraoe_checkout_0021aoe_checkout_001checkout_latency_after_deployrollbacksuccess1.034useful
runbook_action_prioraoe_checkout_0022aoe_checkout_004checkout_latency_after_deployconfig_revertsuccess1.028useful
runbook_action_prioraoe_checkout_0023aoe_checkout_003checkout_latency_after_deployscale_outpartial_success0.485superficial
runbook_action_prioraoe_payment_0021aoe_payment_003payment_dependency_timeoutrestartsuccess1.074superficial
runbook_action_prioraoe_payment_0022aoe_payment_001payment_dependency_timeouttraffic_drainsuccess0.974useful
runbook_action_prioraoe_payment_0023aoe_distractor_payment_001payment_dependency_timeoutscale_outsuccess0.973misleading
runbook_action_prioraoe_inventory_0021aoe_inventory_003inventory_db_connection_poolrestartsuccess1.075superficial
runbook_action_prioraoe_inventory_0022aoe_inventory_001inventory_db_connection_poolconfig_revertsuccess1.034useful
runbook_action_prioraoe_inventory_0023aoe_distractor_inventory_001inventory_db_connection_poolscale_outsuccess0.977misleading
runbook_action_prioraoe_cache_0021aoe_distractor_cache_001recommendation_cache_stalerestartsuccess1.036misleading
runbook_action_prioraoe_cache_0022aoe_cache_003recommendation_cache_stalerollbacksuccess1.036superficial
runbook_action_prioraoe_cache_0023aoe_cache_001recommendation_cache_stalecache_purgesuccess1.034useful
runbook_action_prioraoe_queue_0031aoe_distractor_queue_001order_queue_backlogrestartsuccess1.045misleading
runbook_action_prioraoe_queue_0032aoe_queue_001order_queue_backlogscale_outsuccess1.029useful
runbook_action_prioraoe_queue_0033aoe_queue_002order_queue_backlogqueue_resetsuccess0.975superficial
runbook_action_prioraoe_search_0031aoe_distractor_search_001search_index_memory_leakrestartsuccess1.012misleading
runbook_action_prioraoe_search_0032aoe_search_002search_index_memory_leakrollbacksuccess0.974useful
runbook_action_prioraoe_search_0033aoe_search_001search_index_memory_leakrestartpartial_success0.479superficial
reflection_only_memoryaoe_checkout_0021aoe_checkout_001checkout_latency_after_deployrollbacksuccess0.490useful
reflection_only_memoryaoe_checkout_0022aoe_checkout_004checkout_latency_after_deployconfig_revertsuccess0.339useful
reflection_only_memoryaoe_checkout_0023aoe_checkout_003checkout_latency_after_deployscale_outpartial_success0.303superficial
reflection_only_memoryaoe_payment_0021aoe_payment_001payment_dependency_timeouttraffic_drainsuccess0.328useful
reflection_only_memoryaoe_payment_0022aoe_payment_003payment_dependency_timeoutrestartsuccess0.318superficial
reflection_only_memoryaoe_payment_0023aoe_distractor_payment_001payment_dependency_timeoutscale_outsuccess0.312misleading
reflection_only_memoryaoe_inventory_0021aoe_inventory_001inventory_db_connection_poolconfig_revertsuccess0.508useful
reflection_only_memoryaoe_inventory_0022aoe_distractor_inventory_001inventory_db_connection_poolscale_outsuccess0.341misleading
reflection_only_memoryaoe_inventory_0023aoe_inventory_003inventory_db_connection_poolrestartsuccess0.339superficial
reflection_only_memoryaoe_cache_0021aoe_cache_001recommendation_cache_stalecache_purgesuccess0.509useful
reflection_only_memoryaoe_cache_0022aoe_distractor_cache_001recommendation_cache_stalerestartsuccess0.467misleading
reflection_only_memoryaoe_cache_0023aoe_cache_004recommendation_cache_stalescale_outfailure0.425useful
reflection_only_memoryaoe_queue_0031aoe_queue_001order_queue_backlogscale_outsuccess0.440useful
reflection_only_memoryaoe_queue_0032aoe_distractor_queue_001order_queue_backlogrestartsuccess0.412misleading
reflection_only_memoryaoe_queue_0033aoe_distractor_queue_002order_queue_backlogscale_outrollback_required0.368misleading
reflection_only_memoryaoe_search_0031aoe_search_001search_index_memory_leakrestartpartial_success0.495superficial
reflection_only_memoryaoe_search_0032aoe_search_002search_index_memory_leakrollbacksuccess0.395useful
reflection_only_memoryaoe_search_0033aoe_distractor_search_001search_index_memory_leakrestartsuccess0.385misleading
context_gated_action_outcomeaoe_checkout_0021aoe_checkout_001checkout_latency_after_deployrollbacksuccess0.698useful
context_gated_action_outcomeaoe_checkout_0022aoe_checkout_003checkout_latency_after_deployscale_outpartial_success0.494superficial
context_gated_action_outcomeaoe_checkout_0023aoe_checkout_004checkout_latency_after_deployconfig_revertsuccess0.432useful
context_gated_action_outcomeaoe_payment_0021aoe_payment_001payment_dependency_timeouttraffic_drainsuccess0.514useful
context_gated_action_outcomeaoe_payment_0022aoe_payment_003payment_dependency_timeoutrestartsuccess0.508superficial
context_gated_action_outcomeaoe_payment_0023aoe_distractor_payment_001payment_dependency_timeoutscale_outsuccess0.484misleading
context_gated_action_outcomeaoe_inventory_0021aoe_inventory_001inventory_db_connection_poolconfig_revertsuccess0.589useful
context_gated_action_outcomeaoe_inventory_0022aoe_distractor_inventory_001inventory_db_connection_poolscale_outsuccess0.496misleading
context_gated_action_outcomeaoe_inventory_0023aoe_inventory_003inventory_db_connection_poolrestartsuccess0.481superficial
context_gated_action_outcomeaoe_cache_0021aoe_cache_001recommendation_cache_stalecache_purgesuccess0.754useful
context_gated_action_outcomeaoe_cache_0022aoe_distractor_cache_001recommendation_cache_stalerestartsuccess0.658misleading
context_gated_action_outcomeaoe_cache_0023aoe_cache_004recommendation_cache_stalescale_outfailure0.532useful
context_gated_action_outcomeaoe_queue_0031aoe_queue_001order_queue_backlogscale_outsuccess0.700useful
context_gated_action_outcomeaoe_queue_0032aoe_distractor_queue_001order_queue_backlogrestartsuccess0.585misleading
context_gated_action_outcomeaoe_queue_0033aoe_distractor_queue_002order_queue_backlogscale_outrollback_required0.486misleading
context_gated_action_outcomeaoe_search_0031aoe_search_001search_index_memory_leakrestartpartial_success0.683superficial
context_gated_action_outcomeaoe_search_0032aoe_search_002search_index_memory_leakrollbacksuccess0.628useful
context_gated_action_outcomeaoe_search_0033aoe_distractor_search_001search_index_memory_leakrestartsuccess0.470misleading
QueryCandidateActionOutcomeRaw ValueGated ValueMultiplierContext ScoreReasons
aoe_checkout_002aoe_checkout_003scale_outpartial_success0.450.290.650.507none
aoe_checkout_002aoe_checkout_004config_revertsuccess1.000.150.150.243change_type_mismatch:deploy->config
aoe_payment_002aoe_payment_001traffic_drainsuccess1.000.650.650.571none
aoe_payment_002aoe_payment_003restartsuccess1.000.150.150.083single_entity_vs_service_wide
aoe_payment_002aoe_distractor_payment_001scale_outsuccess1.000.150.150.196candidate_cpu_saturated_query_not
aoe_payment_002aoe_payment_004rollbacksuccess1.000.150.150.210change_type_mismatch:none->deploy
aoe_inventory_002aoe_inventory_001config_revertsuccess1.000.350.350.421change_type_mismatch:none->config
aoe_inventory_002aoe_distractor_inventory_001scale_outsuccess1.000.150.150.217query_db_saturated_candidate_not
aoe_inventory_002aoe_inventory_003restartsuccess1.000.150.150.048single_entity_vs_service_wide
aoe_inventory_002aoe_inventory_004traffic_drainsuccess1.000.350.350.417none
aoe_cache_002aoe_distractor_cache_001restartsuccess1.000.150.150.309single_entity_vs_service_wide
aoe_cache_002aoe_cache_004scale_outfailure-0.70-0.450.650.635none
aoe_cache_002aoe_distractor_cache_002cache_purgefailure-0.70-0.100.150.060cache_symptom_from_deploy_not_flag, change_type_mismatch:feature_flag->deploy
aoe_cache_002aoe_cache_003rollbacksuccess1.000.150.150.000cache_symptom_from_deploy_not_flag, change_type_mismatch:feature_flag->deploy
aoe_queue_003aoe_distractor_queue_001restartsuccess1.000.150.150.200single_entity_vs_service_wide
aoe_queue_003aoe_distractor_queue_002scale_outrollback_required-0.90-0.140.150.008candidate_consumer_errors_query_healthy, change_type_mismatch:traffic_spike->schema_change
aoe_queue_003aoe_queue_002queue_resetsuccess1.000.150.150.000candidate_consumer_errors_query_healthy, change_type_mismatch:traffic_spike->schema_change
aoe_queue_003aoe_queue_004traffic_drainpartial_success0.450.160.350.383none
aoe_search_003aoe_distractor_search_001restartsuccess1.000.150.150.000single_entity_vs_service_wide, change_type_mismatch:deploy->none
aoe_search_003aoe_distractor_search_002rollbackfailure-0.70-0.100.150.350change_type_mismatch:deploy->index_refresh
aoe_search_003aoe_search_004traffic_drainpartial_success0.450.070.150.194change_type_mismatch:deploy->index_refresh

The key comparison is not whether every approach finds at least one historically successful action. On this fixture, that metric is too easy. The stronger commercial signal is failed-action avoidance under noisy same-family distractors.

context_gated_action_outcome is the only variant that both preserves top-3 useful-action coverage and reaches perfect failed-action avoidance on the noisy fixture. The suppressions show why: it refuses to reuse positive or negative outcomes from mismatched topology, change, and metric contexts.

Replay Experiment 010 through SQL tables after alias-aware hash JOIN WHERE predicates are implemented, so the comparison can be audited with native ZeptoDB JOIN/window queries instead of only Python fixtures.