Kontakt- und Warteschlangenmetriken Leistungskennzahlen für Agenten Chat-Metriken Metriken zur Konversationsanalyse Kennzahlen für KI-Agenten Flow-Metriken Bewertungsmetriken Metriken für ausgehende Kampagnen Fallmetriken Bot-Metriken Allgemeine Abfragemuster Einhaltung des Zeitplans für Agenten (Aktivitätsebene)Bewährte Methoden

Referenzabfragen für den Connect Customer Data Lake

Dieses Thema enthält Athena-SQL-Abfragen (Trino Engine v3) zur Berechnung gängiger Connect Customer-Metriken aus Data-Lake-Tabellen. Alle Abfragen verwenden Bezeichner in doppelten Anführungszeichen und gehen von einem Datenbanknamen aus. connect_datalake Passen Sie den Datenbanknamen an Ihre Glue-Katalogkonfiguration an.

Ersetzen Sie <YOUR_INSTANCE_ID> in jeder Abfrage durch Ihre Connect Customer-Instanz-ID.

Inhalt

Kontakt- und Warteschlangenmetriken
Leistungskennzahlen für Agenten
Chat-Metriken
Metriken zur Konversationsanalyse
Kennzahlen für KI-Agenten
Flow-Metriken
Bewertungsmetriken
Metriken für ausgehende Kampagnen
Fallmetriken
Bot-Metriken
Allgemeine Abfragemuster
Einhaltung des Zeitplans für Agenten (Aktivitätsebene)
Bewährte Methoden

Kontakt- und Warteschlangenmetriken

Abbruchrate

Definition: Prozentsatz der Kontakte, die vom Kunden während der Warteschleife unterbrochen wurden. Rückrufe ausgeschlossen.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    CAST(SUM("is_abandoned") AS DOUBLE) 
        / NULLIF(SUM("is_queued"), 0) * 100.0 AS "abandonment_rate_pct"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id"
ORDER BY "abandonment_rate_pct" DESC;

Contacts abandoned (Abgebrochene Kontakte)

Definition: Anzahl der Kontakte, die vom Kunden unterbrochen wurden, während er in der Warteschlange wartete.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    SUM("is_abandoned") AS "contacts_abandoned"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Contacts abandoned in X seconds (Kontakte, die innerhalb von X Sekunden abgebrochen wurden)

Definition: Anzahl der Kontakte, die innerhalb von X Sekunden, nachdem sie in die Warteschlange gestellt wurden, verlassen wurden.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    SUM(
        CASE WHEN "is_abandoned" = 1 
             AND "queue_time_ms" <= 30000 
             THEN 1 ELSE 0 END
    ) AS "contacts_abandoned_in_30s"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Average queue abandon time (Durchschnittliche Warteschlangenabbruchzeit)

Definition: Durchschnittliche Wartezeit von Kontakten in der Warteschlange, bevor sie den Vorgang abbrechen.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    AVG("abandon_time_ms") / 1000.0 AS "avg_queue_abandon_time_sec"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "is_abandoned" = 1
  AND "abandon_time_ms" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Durchschnittliche Warteschlangenantwortzeit

Definition: Durchschnittliche Zeit, in der Kontakte in der Warteschlange warteten, bis sie von einem Agenten beantwortet wurden.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    AVG("queue_answer_time_ms") / 1000.0 AS "avg_queue_answer_time_sec"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "is_handled" = 1
  AND "queue_answer_time_ms" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Serviceniveau

Definition: Anzahl und Prozentsatz der Kontakte, die innerhalb von X Sekunden beantwortet wurden.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    SUM(CASE WHEN "is_handled" = 1 AND "queue_answer_time_ms" <= 20000 
             THEN 1 ELSE 0 END) AS "contacts_answered_in_20s",
    SUM("is_queued") AS "contacts_queued",
    CAST(SUM(CASE WHEN "is_handled" = 1 AND "queue_answer_time_ms" <= 20000 
                  THEN 1 ELSE 0 END) AS DOUBLE)
        / NULLIF(SUM("is_queued"), 0) * 100.0 AS "service_level_20s_pct"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Contacts queued (Kontakte in Warteschlange)

Definition: Anzahl der Kontakte, die in eine Warteschlange aufgenommen wurden.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    SUM("is_queued") AS "contacts_queued"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Contacts handled (Bearbeitete Kontakte)

Definition: Anzahl der Kontakte, die mit einem Agenten verbunden sind.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    SUM("is_handled") AS "contacts_handled"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Contacts transferred in (Weitergeleitete Kontakte ein)

Definition: Kontakte, die in eine Warteschlange übertragen wurden.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    SUM("is_transferred_in") AS "contacts_transferred_in"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Contacts transferred out (Weitergeleitete Kontakte aus)

Definition: Kontakte, die aus einer Warteschlange übertragen wurden.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    SUM("is_transferred_out") AS "contacts_transferred_out",
    SUM("is_transferred_out_internal") AS "transferred_out_internal",
    SUM("is_transferred_out_external") AS "transferred_out_external"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Maximum queued time (Maximale Zeit in der Warteschlange)

Definition: Längste Zeit, die ein Kontakt damit verbracht hat, in der Warteschlange zu warten.

Quelltabelle: contact_record


SELECT
    "queue_id",
    MAX("queue_duration_ms") / 1000.0 AS "max_queued_time_sec"
FROM "connect_datalake"."contact_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "queue_duration_ms" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Durchschnittliche Kontaktdauer

Definition: Durchschnittliche Zeit von der Kontaktaufnahme bis zur Unterbrechung der Verbindung.

Quelltabelle: contact_record


SELECT
    "queue_id",
    AVG(
        date_diff('millisecond', "initiation_timestamp", "disconnect_timestamp")
    ) / 1000.0 AS "avg_contact_duration_sec"
FROM "connect_datalake"."contact_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "initiation_timestamp" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Leistungskennzahlen für Agenten

Average handle time (Durchschnittliche Bearbeitungszeit)

Definition: Durchschnittliche Zeit von der Kontaktverbindung bis zum Abschluss von ACW.

Quelltabelle: contact_statistic_record


SELECT
    "agent_id",
    AVG("handle_time_ms") / 1000.0 AS "avg_handle_time_sec"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "is_handled" = 1
  AND "handle_time_ms" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "agent_id";

After contact work time (Kontaktnachbearbeitungszeit)

Definition: Gesamtzeit, die Agenten im Status ACW verbracht haben.

Quelltabelle: contact_statistic_record


SELECT
    "agent_id",
    SUM("after_contact_work_time_ms") / 1000.0 AS "total_acw_time_sec"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "after_contact_work_time_ms" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "agent_id";

Customer hold time (Kundenhaltezeit)

Definition: Gesamtzeit, die Kunden in der Warteschleife verbracht haben, nachdem sie eine Verbindung zum Agenten hergestellt haben.

Quelltabelle: contact_statistic_record


SELECT
    "agent_id",
    SUM("customer_hold_time_ms") / 1000.0 AS "total_hold_time_sec"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "customer_hold_time_ms" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "agent_id";

Agent idle time (Leerlaufzeit von Kundendienstmitarbeitern)

Definition: Zeit, die der Agent im Status „Verfügbar“ verbracht hat, ohne Kontakte zu bearbeiten.

Quelltabelle: agent_statistic_record


SELECT
    "user_id" AS "agent_id",
    SUM("agent_idle_time") / 1000.0 AS "total_idle_time_sec"
FROM "connect_datalake"."agent_statistic_record"
WHERE "published_date" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "published_date" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "user_id";

Occupancy (Auslastung)

Definition: Prozentualer Anteil der Zeit, in der Agenten bei Kontakten aktiv waren, im Vergleich zu „verfügbar“ und „aktiv“.

Quelltabelle: agent_statistic_record


SELECT
    "user_id" AS "agent_id",
    CAST(SUM("agent_on_contact_time") AS DOUBLE)
        / NULLIF(SUM("agent_on_contact_time") + SUM("agent_idle_time"), 0) 
        * 100.0 AS "occupancy_pct"
FROM "connect_datalake"."agent_statistic_record"
WHERE "published_date" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "published_date" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "user_id";

Kundendienstmitarbeiter non-response (Nichtbeantwortung des Kundendienstmitarbeiters)

Definition: Anzahl der Kontakte, die an den Agenten weitergeleitet, aber nicht beantwortet wurden.

Quelltabelle: agent_queue_statistic_record


SELECT
    "user_id" AS "agent_id",
    "queue_id",
    SUM("agent_non_response") AS "agent_non_response_count"
FROM "connect_datalake"."agent_queue_statistic_record"
WHERE "published_date" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "published_date" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "user_id", "queue_id";

Agent answer rate (Antwortrate des Kundendienstmitarbeiters)

Definition: Prozentsatz der weitergeleiteten Kontakte, die vom Agenten beantwortet wurden.

Quelltabelle: agent_queue_statistic_record


SELECT
    "user_id" AS "agent_id",
    CAST(SUM("contacts_handled") AS DOUBLE) 
        / NULLIF(SUM("contacts_offered"), 0) * 100.0 AS "agent_answer_rate_pct"
FROM "connect_datalake"."agent_queue_statistic_record"
WHERE "published_date" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "published_date" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "user_id";

Online time (Online-Zeit)

Definition: Gesamtzeit, in der Agent CCP auf einen anderen Status als Offline gesetzt wurde.

Quelltabelle: agent_statistic_record


SELECT
    "user_id" AS "agent_id",
    SUM("online_time") / 1000.0 AS "total_online_time_sec"
FROM "connect_datalake"."agent_statistic_record"
WHERE "published_date" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "published_date" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "user_id";

Chat-Metriken

Durchschnittl. Zeit erste Reaktion Kundendienstmitarbeiter

Definition: Durchschnittliche Zeit, die ein Agent benötigt, um die erste Nachricht zu senden, nachdem er einen Chat-Kontakt erhalten hat.

Quelltabelle: contact_record


SELECT
    "queue_id",
    AVG("chat_contact_metrics_agent_first_response_time_ms") / 1000.0 
        AS "avg_agent_first_response_sec"
FROM "connect_datalake"."contact_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "channel" = 'CHAT'
  AND "chat_contact_metrics_agent_first_response_time_ms" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Durchschnittl. Zeit Reaktion Kundendienstmitarbeiter

Definition: Durchschnittliche Zeit, die Agenten benötigen, um auf Kundennachrichten zu antworten.

Quelltabelle: contact_record


SELECT
    "queue_id",
    CAST(SUM("chat_agent_metrics_total_response_time_ms") AS DOUBLE)
        / NULLIF(SUM("chat_agent_metrics_num_responses"), 0) / 1000.0
        AS "avg_agent_response_time_sec"
FROM "connect_datalake"."contact_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "channel" = 'CHAT'
  AND "chat_agent_metrics_total_response_time_ms" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Durchschnittliche Gesamtzahl der Nachrichten

Definition: Durchschnittliche Gesamtzahl der Nachrichten pro Chat-Kontakt.

Quelltabelle: contact_record


SELECT
    "queue_id",
    AVG(CAST("chat_contact_metrics_total_messages" AS DOUBLE)) AS "avg_total_messages"
FROM "connect_datalake"."contact_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "channel" = 'CHAT'
  AND "chat_contact_metrics_total_messages" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Abgebrochene Konversationen

Definition: Kontakte, bei denen der Chat von einem Agenten oder Kunden abgebrochen wurde.

Quelltabelle: contact_record


SELECT
    "queue_id",
    COUNT(*) AS "conversations_abandoned"
FROM "connect_datalake"."contact_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "channel" = 'CHAT'
  AND ("chat_agent_metrics_conversation_abandon" = true 
       OR "chat_customer_metrics_conversation_abandon" = true)
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id";

Metriken zur Konversationsanalyse

Durchschnittliche Gesprächszeit

Definition: Durchschnittliche kombinierte Gesprächszeit zwischen Agenten und Kunde pro Sprachkontakt.

Quelltabelle: contact_lens_conversational_analytics


SELECT
    AVG("talk_time_total_ms") / 1000.0 AS "avg_talk_time_sec"
FROM "connect_datalake"."contact_lens_conversational_analytics"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "channel" = 'VOICE'
  AND "instance_id" = '<YOUR_INSTANCE_ID>';

Durchschnittliche Nicht-Gesprächszeit

Definition: Durchschnittliche Haltezeit plus Ruhezeit pro Sprachkontakt.

Quelltabelle: contact_lens_conversational_analytics


SELECT
    AVG("non_talk_time_total_ms") / 1000.0 AS "avg_non_talk_time_sec"
FROM "connect_datalake"."contact_lens_conversational_analytics"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "channel" = 'VOICE'
  AND "instance_id" = '<YOUR_INSTANCE_ID>';

Stimmungsbewertung

Definition: Allgemeine Stimmungswerte für Agenten und Kunden.

Quelltabelle: contact_lens_conversational_analytics


SELECT
    AVG("sentiment_overall_score_agent") AS "avg_agent_sentiment",
    AVG("sentiment_overall_score_customer") AS "avg_customer_sentiment",
    AVG("sentiment_end_score_agent") AS "avg_agent_end_sentiment",
    AVG("sentiment_end_score_customer") AS "avg_customer_end_sentiment"
FROM "connect_datalake"."contact_lens_conversational_analytics"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>';

Durchschnittliche Kundendienstmitarbeiterunterbrechungen

Definition: Durchschnittliche Anzahl von Agentenunterbrechungen pro Kontakt.

Quelltabelle: contact_lens_conversational_analytics


SELECT
    AVG(CAST("interruptions_agent_count" AS DOUBLE)) AS "avg_agent_interruptions"
FROM "connect_datalake"."contact_lens_conversational_analytics"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "channel" = 'VOICE'
  AND "instance_id" = '<YOUR_INSTANCE_ID>';

Kennzahlen für KI-Agenten

Erfolgsquote beim Aufrufen von KI-Agenten

Definition: Rate erfolgreicher Aufrufe von KI-Agenten.

Quelltabelle: ai_agent


SELECT
    "ai_agent_name",
    SUM(CASE WHEN "invocation_success" = true THEN 1 ELSE 0 END) AS "success_count",
    COUNT(*) AS "total_invocations",
    CAST(SUM(CASE WHEN "invocation_success" = true THEN 1 ELSE 0 END) AS DOUBLE)
        / NULLIF(COUNT(*), 0) * 100.0 AS "success_rate_pct"
FROM "connect_datalake"."ai_agent"
WHERE "creation_timestamp" >= CAST('2026-06-09' AS TIMESTAMP) * 1000
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
  AND "ai_agent_id" IS NOT NULL
GROUP BY "ai_agent_name";

AI-Übergaberate

Definition: Rate der KI-Sitzungen, die an menschliche Agenten eskalierten.

Quelltabelle: ai_session


SELECT
    SUM(CASE WHEN "is_handed_off" = true THEN 1 ELSE 0 END) AS "ai_handoffs",
    COUNT(*) AS "ai_involved_contacts",
    CAST(SUM(CASE WHEN "is_handed_off" = true THEN 1 ELSE 0 END) AS DOUBLE)
        / NULLIF(COUNT(*), 0) * 100.0 AS "handoff_rate_pct"
FROM "connect_datalake"."ai_session"
WHERE "creation_timestamp" >= CAST('2026-06-09' AS TIMESTAMP) * 1000
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
  AND "ai_session_id" IS NOT NULL;

KI-Qualitätswerte

Definition: Durchschnittliche Werte für Torerfolg, Treue und Vollständigkeit.

Quelltabelle: ai_session


SELECT
    AVG("goal_success_rate") AS "avg_goal_success_rate",
    AVG("faithfulness_score") AS "avg_faithfulness_score",
    AVG("completeness_score") AS "avg_completeness_score"
FROM "connect_datalake"."ai_session"
WHERE "creation_timestamp" >= CAST('2026-06-09' AS TIMESTAMP) * 1000
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
  AND "goal_success_rate" IS NOT NULL;

Genauigkeit des KI-Tools

Definition: Genauigkeitswerte für die Verwendung, Auswahl und Nutzung von KI-Tool-Parametern.

Quelltabelle: ai_tool


SELECT
    "ai_tool_name",
    AVG("ai_tool_parameter_accuracy") AS "avg_parameter_accuracy",
    AVG("ai_tool_selection_accuracy") AS "avg_selection_accuracy",
    AVG("ai_tool_utilization_accuracy") AS "avg_use_accuracy"
FROM "connect_datalake"."ai_tool"
WHERE "creation_timestamp" >= CAST('2026-06-09' AS TIMESTAMP) * 1000
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
  AND "ai_tool_id" IS NOT NULL
GROUP BY "ai_tool_name";

Flow-Metriken

Gestartete Flows

Definition: Anzahl der Flows, deren Ausführung begonnen hat.

Quelltabelle: contact_flow_events


SELECT
    "flow_resource_id",
    "flow_type",
    COUNT(*) AS "flows_started"
FROM "connect_datalake"."contact_flow_events"
WHERE "start_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "start_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "flow_resource_id", "flow_type";

Prozentsatz des Ablaufergebnisses

Definition: Prozentsatz jedes Flow-Ergebnistyps.

Quelltabelle: contact_flow_events


WITH flow_counts AS (
    SELECT
        "flow_resource_id",
        "flow_outcome",
        COUNT(*) AS "outcome_count",
        SUM(COUNT(*)) OVER (PARTITION BY "flow_resource_id") AS "total_completed"
    FROM "connect_datalake"."contact_flow_events"
    WHERE "start_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
      AND "start_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
      AND "end_timestamp" IS NOT NULL
      AND "instance_id" = '<YOUR_INSTANCE_ID>'
    GROUP BY "flow_resource_id", "flow_outcome"
)
SELECT
    "flow_resource_id",
    "flow_outcome",
    "outcome_count",
    CAST("outcome_count" AS DOUBLE) / "total_completed" * 100.0 AS "outcome_pct"
FROM flow_counts
ORDER BY "flow_resource_id", "outcome_pct" DESC;

Durchschnittliche Flow-Zeit

Definition: Durchschnittliche Dauer der Flow-Ausführungen.

Quelltabelle: contact_flow_events


SELECT
    "flow_resource_id",
    AVG(
        date_diff('millisecond', "start_timestamp", "end_timestamp")
    ) / 1000.0 AS "avg_flow_time_sec"
FROM "connect_datalake"."contact_flow_events"
WHERE "start_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "start_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "end_timestamp" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "flow_resource_id";

Bewertungsmetriken

Durchgeführte Evaluationen

Definition: Anzahl der eingereichten Bewertungen.

Quelltabelle: contact_evaluation_record


SELECT
    COUNT(DISTINCT "evaluation_id") AS "evaluations_performed"
FROM "connect_datalake"."contact_evaluation_record"
WHERE "evaluation_submitted_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "evaluation_submitted_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "item_type" = 'Form'
  AND "to_delete" = false
  AND ("evaluation_type" IS NULL OR "evaluation_type" != 'calibration')
  AND "instance_id" = '<YOUR_INSTANCE_ID>';

Durchschnittliches Bewertungsergebnis

Definition: Durchschnittliche Bewertungspunktzahl aller eingereichten Bewertungen.

Quelltabelle: contact_evaluation_record


SELECT
    AVG("score") AS "avg_evaluation_score_pct"
FROM "connect_datalake"."contact_evaluation_record"
WHERE "evaluation_submitted_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "evaluation_submitted_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "item_type" = 'Form'
  AND "to_delete" = false
  AND ("evaluation_type" IS NULL OR "evaluation_type" != 'calibration')
  AND "instance_id" = '<YOUR_INSTANCE_ID>';

Automatische Fehlschläge in Prozent

Definition: Prozentsatz der Bewertungen, die zu einem automatischen Fehlschlagen geführt haben.

Quelltabelle: contact_evaluation_record


SELECT
    CAST(
        COUNT(DISTINCT CASE WHEN "automatic_fail" = true THEN "evaluation_id" END) AS DOUBLE
    ) / NULLIF(COUNT(DISTINCT "evaluation_id"), 0) * 100.0 
        AS "automatic_fail_pct"
FROM "connect_datalake"."contact_evaluation_record"
WHERE "evaluation_submitted_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "evaluation_submitted_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "item_type" = 'Form'
  AND "to_delete" = false
  AND ("evaluation_type" IS NULL OR "evaluation_type" != 'calibration')
  AND "instance_id" = '<YOUR_INSTANCE_ID>';

Metriken für ausgehende Kampagnen

Kontakte zur Kampagne

Definition: Anzahl der ausgehenden Kampagnenkontakte.

Quelltabelle: contact_record


SELECT
    "campaign_id",
    COUNT(*) AS "campaign_contacts"
FROM "connect_datalake"."contact_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "campaign_id" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "campaign_id";

Mensch hat geantwortet

Definition: Ausgehende Kampagnenanrufe, die mit einem Live-Kunden verbunden sind.

Quelltabelle: contact_record


SELECT
    "campaign_id",
    COUNT(*) AS "human_answered"
FROM "connect_datalake"."contact_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "campaign_id" IS NOT NULL
  AND "answering_machine_detection_status" = 'HUMAN_ANSWERED'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "campaign_id";

Fallmetriken

Erstellte Fälle

Definition: Gesamtzahl der in einem Zeitraum erstellten Fälle.

Quelltabelle: case_events


SELECT
    COUNT(DISTINCT "case_id") AS "cases_created"
FROM "connect_datalake"."case_events"
WHERE "event_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "event_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "event_type" = 'CASE.CREATED'
  AND "instance_id" = '<YOUR_INSTANCE_ID>';

Durchschnittliche Falllösungszeit

Definition: Durchschnittliche Zeit von der Erstellung des Falls bis zum Abschluss.

Quelltabelle: case_events


SELECT
    AVG(
        date_diff('hour', "created_timestamp", "last_closed_timestamp")
    ) AS "avg_resolution_time_hours"
FROM "connect_datalake"."case_events"
WHERE "last_closed_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "last_closed_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "created_timestamp" IS NOT NULL
  AND "instance_id" = '<YOUR_INSTANCE_ID>';

Bot-Metriken

Ergebnisse der Bot-Konversation

Definition: Prozentuale Aufschlüsselung der Ergebnisse von Bot-Konversationen.

Quelltabelle: bot_conversations


WITH bot_outcomes AS (
    SELECT
        "bot_id",
        "bot_conversation_outcome",
        COUNT(*) AS "cnt",
        SUM(COUNT(*)) OVER (PARTITION BY "bot_id") AS "total"
    FROM "connect_datalake"."bot_conversations"
    WHERE "bot_conversation_start_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
      AND "bot_conversation_start_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
      AND "instance_id" = '<YOUR_INSTANCE_ID>'
    GROUP BY "bot_id", "bot_conversation_outcome"
)
SELECT
    "bot_id",
    "bot_conversation_outcome",
    "cnt",
    CAST("cnt" AS DOUBLE) / "total" * 100.0 AS "outcome_pct"
FROM bot_outcomes;

Allgemeine Abfragemuster

Die folgenden Muster zeigen, wie Sie mehrere Data-Lake-Tabellen kombinieren können, um umfassende Dashboards und Berichte zu erhalten.

Tägliches Übersichts-Dashboard

Definition: Umfassende tägliche Kennzahlen zur Warteschlange, einschließlich des Servicelevels.

Quelltabelle: contact_statistic_record


SELECT
    "queue_id",
    SUM("is_queued") AS "contacts_queued",
    SUM("is_handled") AS "contacts_handled",
    SUM("is_abandoned") AS "contacts_abandoned",
    AVG(CASE WHEN "is_handled" = 1 THEN "queue_answer_time_ms" END) / 1000.0 
        AS "avg_answer_time_sec",
    AVG(CASE WHEN "is_handled" = 1 THEN "handle_time_ms" END) / 1000.0 
        AS "avg_handle_time_sec",
    CAST(SUM(CASE WHEN "is_handled" = 1 AND "queue_answer_time_ms" <= 20000 
                  THEN 1 ELSE 0 END) AS DOUBLE)
        / NULLIF(SUM("is_queued"), 0) * 100.0 AS "sl_20s_pct"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY "queue_id"
ORDER BY "contacts_queued" DESC;

Stündliche Trendanalyse

Definition: Stündliches Kontaktvolumen und Trends beim Servicelevel.

Quelltabelle: contact_statistic_record


SELECT
    date_trunc('hour', "disconnect_timestamp") AS "hour",
    "queue_id",
    SUM("is_queued") AS "contacts_queued",
    SUM("is_handled") AS "contacts_handled",
    SUM("is_abandoned") AS "contacts_abandoned",
    CAST(SUM("is_abandoned") AS DOUBLE) 
        / NULLIF(SUM("is_queued"), 0) * 100.0 AS "abandon_rate_pct",
    AVG(CASE WHEN "is_handled" = 1 THEN "handle_time_ms" END) / 1000.0 AS "aht_sec"
FROM "connect_datalake"."contact_statistic_record"
WHERE "disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND "disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND "instance_id" = '<YOUR_INSTANCE_ID>'
GROUP BY date_trunc('hour', "disconnect_timestamp"), "queue_id"
ORDER BY "hour";

Mit Kontaktlinsen angereicherte Kontaktlinsen

Definition: Bereichern Sie Kontaktdatensätze mit Kontaktlinsenanalysen.

Quelltabelle: contact_record verknüpft mit contact_lens_conversational_analytics


SELECT
    cr."contact_id",
    cr."queue_id",
    cr."agent_id",
    cr."agent_interaction_duration_ms" / 1000.0 AS "interaction_sec",
    cl."talk_time_agent_ms" / 1000.0 AS "agent_talk_sec",
    cl."talk_time_customer_ms" / 1000.0 AS "customer_talk_sec",
    cl."sentiment_overall_score_agent",
    cl."sentiment_overall_score_customer"
FROM "connect_datalake"."contact_record" cr
JOIN "connect_datalake"."contact_lens_conversational_analytics" cl
    ON cr."contact_id" = cl."contact_id"
    AND cr."instance_id" = cl."instance_id"
WHERE cr."disconnect_timestamp" >= TIMESTAMP '2026-06-09 00:00:00'
  AND cr."disconnect_timestamp" <  TIMESTAMP '2026-06-10 00:00:00'
  AND cr."instance_id" = '<YOUR_INSTANCE_ID>'
  AND cr."channel" = 'VOICE';

Einhaltung des Zeitplans für Agenten (Aktivitätsebene)

Definition: Vergleicht den tatsächlichen Aktivitätsstatus eines Agenten (vonagent_statistic_record) mit seinen geplanten Schichtaktivitäten (aus Planungstabellen) für jedes Zeitintervall eines Tages. Ermittelt pro Intervall die Einhaltung: IN (der Agent hat getan, was für ihn geplant war) oder OUT (er hat nicht getan).

Ausgabespalten: Agent, Datum, Beginn, Ende, geplante Aktivität, Aktuelle Aktivität, Status der Einhaltung, Dauer

Quelltabellen:

staff_shifts— Der Agent wechselt für den Tag (letzte, nicht gelöschte Version)
staff_shift_activities— Geplante Aktivitätsblöcke innerhalb jeder Schicht
shift_activities— Suche nach Aktivitätsnamen (ordnet ARN einem menschenlesbaren Namen zu)
agent_statistic_record— Aktueller Agentenstatus pro Intervall
users— Agentenname und ARN-Auflösung

Adhärenzlogik (vereinfacht):

Geplantes „Öffnen“ — der Agent ist AKTIV, wenn der Status „Verfügbar“, „In Kontakt“ oder „ACW“ lautet
Geplante „Pause“ — Der Agent ist ZUGESCHALTET, wenn der Status Pause oder Mittagessen lautet
Geplantes „Meeting“ — Der Agent ist ANGEMELDET, wenn der Status „Schulung“ oder „Besprechung“ lautet
Andernfalls — AUS


WITH latest_shift_versions AS (
    -- Get the latest (non-deleted) shift version per shift_id
    SELECT
        shift_id,
        MAX(shift_version) AS max_version
    FROM "connect_datalake"."staff_shifts"
    WHERE is_deleted = false
      AND CAST(shift_start_timestamp AS DATE) = DATE '2026-06-10'  -- SET REPORT DATE
    GROUP BY shift_id
),

latest_shifts AS (
    SELECT
        ss.shift_id,
        ss.agent_arn,
        ss.shift_start_timestamp,
        ss.shift_end_timestamp
    FROM "connect_datalake"."staff_shifts" ss
    INNER JOIN latest_shift_versions lsv
        ON ss.shift_id = lsv.shift_id
        AND ss.shift_version = lsv.max_version
    WHERE ss.is_deleted = false
),

-- Get scheduled activity blocks with human-readable activity names
scheduled_blocks AS (
    SELECT
        ls.agent_arn,
        ssa.activity_start_timestamp,
        ssa.activity_end_timestamp,
        sa.shift_activity_name,
        CASE
            WHEN sa.shift_activity_name IN ('Work', 'Overtime') THEN 'Open'
            WHEN sa.shift_activity_name IN ('Break', 'Lunch') THEN 'Break'
            WHEN sa.shift_activity_name = 'Training' THEN 'Meeting'
            WHEN sa.shift_activity_name = 'PTO' THEN 'PTO'
            ELSE sa.shift_activity_name
        END AS scheduled_activity_label
    FROM "connect_datalake"."staff_shift_activities" ssa
    INNER JOIN latest_shifts ls
        ON ssa.shift_id = ls.shift_id
    INNER JOIN latest_shift_versions lsv
        ON ssa.shift_id = lsv.shift_id
        AND ssa.shift_version = lsv.max_version
    INNER JOIN "connect_datalake"."shift_activities" sa
        ON ssa.shift_activity_arn = sa.shift_activity_arn
    WHERE ssa.is_deleted = false
),

-- Get actual agent state intervals for the day
actual_states AS (
    SELECT
        u.user_arn AS agent_arn,
        u.first_name,
        u.last_name,
        asr.interval_start_time,
        asr.interval_end_time,
        asr.agent_status_name,
        asr.online_time,
        asr.agent_idle_time,
        asr.agent_on_contact_time,
        asr.non_productive_time,
        CASE
            WHEN asr.agent_on_contact_time IS NOT NULL AND asr.agent_on_contact_time > 0
                THEN 'On Inbound Call'
            WHEN asr.agent_idle_time IS NOT NULL AND asr.agent_idle_time > 0
                THEN 'Available'
            WHEN asr.non_productive_time IS NOT NULL AND asr.non_productive_time > 0
                THEN COALESCE(asr.agent_status_name, 'Non-Productive')
            WHEN asr.online_time IS NOT NULL AND asr.online_time > 0
                THEN 'Available'
            ELSE COALESCE(asr.agent_status_name, 'Offline')
        END AS actual_activity_label
    FROM "connect_datalake"."agent_statistic_record" asr
    INNER JOIN "connect_datalake"."users" u
        ON asr.user_id = u.user_id
    WHERE asr.interval_start_time >= TIMESTAMP '2026-06-10 00:00:00' -- SET REPORT DATE (UTC)
      AND asr.interval_start_time < TIMESTAMP '2026-06-11 00:00:00'
),

-- Join actual states with scheduled blocks
activity_timeline AS (
    SELECT
        act.first_name || ' ' || act.last_name AS agent_name,
        act.interval_start_time,
        act.interval_end_time,
        act.actual_activity_label,
        act.agent_status_name,
        COALESCE(sb.scheduled_activity_label, 'Open') AS scheduled_activity
    FROM actual_states act
    LEFT JOIN scheduled_blocks sb
        ON act.agent_arn = sb.agent_arn
        AND act.interval_start_time < sb.activity_end_timestamp
        AND act.interval_end_time > sb.activity_start_timestamp
)

SELECT
    agent_name AS "AGENT",
    CAST(interval_start_time AS DATE) AS "DATE",
    DATE_FORMAT(interval_start_time, '%H:%i:%s') AS "BEGIN",
    DATE_FORMAT(interval_end_time, '%H:%i:%s') AS "END",
    scheduled_activity AS "SCHEDULED ACTIVITY",
    actual_activity_label AS "ACTUAL ACTIVITY",
    CASE
        WHEN scheduled_activity = 'Open'
            AND actual_activity_label IN ('Available', 'On Inbound Call', 'On Outbound Call',
                                          'Call Ringing', 'Aftercall (ACW)')
            THEN 'IN'
        WHEN scheduled_activity = 'Break'
            AND agent_status_name IN ('Break', 'Lunch')
            THEN 'IN'
        WHEN scheduled_activity = 'Meeting'
            AND agent_status_name IN ('Training', 'Meeting')
            THEN 'IN'
        ELSE 'OUT'
    END AS "ADHERENCE STATE",
    CAST(DATE_DIFF('second', interval_start_time, interval_end_time) / 3600 AS VARCHAR)
        || ':' ||
        LPAD(CAST((DATE_DIFF('second', interval_start_time, interval_end_time) % 3600) / 60 AS VARCHAR), 2, '0')
        || ':' ||
        LPAD(CAST(DATE_DIFF('second', interval_start_time, interval_end_time) % 60 AS VARCHAR), 2, '0')
    AS "DURATION"
FROM activity_timeline
ORDER BY interval_start_time ASC;

Bewährte Methoden

Löschen von Partitionen — Verwenden Sie immer Partitionsfilter (disconnect_timestamppublished_date, odercreation_timestamp), um die Scankosten zu minimieren.
Deduplizierung — Connect Customer liefert Datensätze mindestens einmal. Wird für Primärschlüssel verwendetDISTINCT, wenn genaue Zählungen erforderlich sind.
Zeitzonen — Alle Zeitstempel sind in UTC. AT TIME ZONEBeantragen Sie lokale Berichterstattung.
Millisekunden — Die meisten Felder für die Dauer werden in Millisekunden gespeichert. Teilen Sie für Sekunden durch 1000,0.
Instanz-ID-Filter — In Umgebungen mit mehreren Instanzen wird immer nach instance_id gefiltert.
Real-time Metriken — Verwenden Sie die GetCurrentMetricData API, um echte Echtzeit-Metriken zu erhalten. Der Data Lake stellt nur historische Daten zur Verfügung.

Warnung JavaScript ist in Ihrem Browser nicht verfügbar oder deaktiviert.

Zur Nutzung der AWS-Dokumentation muss JavaScript aktiviert sein. Weitere Informationen finden auf den Hilfe-Seiten Ihres Browsers.

Dokumentkonventionen

Planungsdaten

Datenaufbewahrung