Gong model

Rep interaction stats by week

Gong's Interaction stats per rep per week: talk ratio, longest monologue, longest customer story, interactivity, patience and question rate.

Grain
ISO week (starting Monday) x gong_account x rep (primary user)
Built from
  1. 6
    Raw tables
  2. 6
    Staging
  3. 2
    Core
  4. 1
    Analytics
Reads from Gong
callcall_context_integrationcall_interactioncall_participantcall_speakeruser
Ships as
Weld and dbt SQL, open source on GitHub
core.gong.rep_interaction_stats_by_week
1-- core.gong.rep_interaction_stats_by_week
2-- Gong's Interaction stats per rep per week: talk ratio, longest monologue,
3-- longest customer story, interactivity, patience and question rate.
4--
5-- Grain: ISO week (starting Monday) x gong_account x rep (primary user).
6-- Depends on: core.gong.calls
7--
8-- EXTERNAL CALLS ONLY. Interaction stats are a coaching measure for conversations
9-- with customers and prospects; an internal stand-up would drag every rep's talk
10-- ratio toward 50%. Remove the scope filter to include them.
11--
12-- AVERAGE OF PER-CALL VALUES, ON PURPOSE. Gong's help centre defines a rep's
13-- average talk ratio as the sum of their per-call talk ratios divided by the number
14-- of calls, so a 10-minute call weighs the same as a 60-minute one. The avg_gong_*
15-- columns follow that definition so they can be checked against the Gong UI.
16-- talk_time_weighted_rep_talk_ratio is the other reading (total rep seconds over
17-- total seconds), which long calls dominate. They disagree whenever call lengths
18-- vary; neither is a bug.
19--
20-- To re-aggregate to month or team, sum the *_sum and *_calls columns and divide.
21-- Averaging the avg_ columns gives a mean of means.
22--
23-- No spine: a week with no external calls has no stats to average, and a zero
24-- would be read as a talk ratio of 0%.
25
26SELECT
27    DATE_TRUNC(call_date, ISOWEEK)                                AS week_start,
28    gong_account,
29    primary_user_id                                               AS user_id,
30    ANY_VALUE(rep_name)                                           AS user_name,
31    ANY_VALUE(rep_manager_name)                                   AS manager_name,
32
33    COUNT(*)                                                      AS external_calls,
34    SUM(duration_minutes)                                         AS external_call_minutes,
35
36    -- Gong's own statistics, averaged per call.
37    COUNT(gong_talk_ratio)                                        AS gong_talk_ratio_calls,
38    SUM(gong_talk_ratio)                                          AS gong_talk_ratio_sum,
39    AVG(gong_talk_ratio)                                          AS avg_gong_talk_ratio,
40    AVG(gong_longest_monologue)                                   AS avg_gong_longest_monologue,

This model and the 8 it depends on. Deploy them together; refs like {{staging.…}} resolve to the matching folders in Weld, and the dbt view uses source() and ref().

Notes

How it counts. And where it stops.

EXTERNAL CALLS ONLY. Interaction stats are a coaching measure for conversations with customers and prospects; an internal stand-up would drag every rep's talk ratio toward 50%. Remove the scope filter to include them. AVERAGE OF PER-CALL VALUES, ON PURPOSE. Gong's help centre defines a rep's average talk ratio as the sum of their per-call talk ratios divided by the number of calls, so a 10-minute call weighs the same as a 60-minute one. The avg_gong_* columns follow that definition so they can be checked against the Gong UI. talk_time_weighted_rep_talk_ratio is the other reading (total rep seconds over total seconds), which long calls dominate. They disagree whenever call lengths vary; neither is a bug. To re-aggregate to month or team, sum the *_sum and *_calls columns and divide. Averaging the avg_ columns gives a mean of means. No spine: a week with no external calls has no stats to average, and a zero would be read as a talk ratio of 0%.

Raw tables it reads

  • raw.gong.call
  • raw.gong.call_context_integration
  • raw.gong.call_interaction
  • raw.gong.call_participant
  • raw.gong.call_speaker
  • raw.gong.user

Put Gong in front
of your agents.

Connect Gong, deploy the models, and ask questions in Claude, Cursor or ChatGPT. Free for 14 days, no credit card.