1. Dataset Name

survey_test_response_rate

2. Description

(rephrased) A response-rate dataset that combines two event sources to calculate the share of candidates who submitted a survey out of those who were served one. Each row represents a single event — either a survey submission (from survey_analytics) or a survey delivery (from www_server_log). Combining these sources lets response rate be calculated in one pass, broken down by device and country.

3. Source Tables

• customer_views.survey_analytics (submit_survey events — numerator of response rate, filtered to event = 'submit_survey' and t_create >= '2021-03-14')
• customer_views.www_server_log (get_survey events — denominator of response rate, filtered to event = 'get_survey')

4. Grain

One row per vs_cookie per event type per session. The UNION ALL produces two distinct event types in a single table: 'submit_survey' rows (from survey_analytics) and 'get_survey' rows (from www_server_log). Response rate aggregations use COUNT DISTINCT on vs_cookie, partitioned by event value, to isolate numerator and denominator in a single pass.

5. Dashboards Supported

• Survey Analytics Dashboard

6. Metrics Supported

7. Known Exclusions / Filters Applied

• The survey_analytics branch is filtered to event = 'submit_survey' and t_create >= '2021-03-14' at dataset build time — survey views, partial submissions, and pre-2021 submissions are excluded from the numerator.
• The www_server_log branch is filtered to event = 'get_survey' at dataset build time — only delivery events contribute to the denominator.
• embedding_id is present in the survey_analytics branch but NULLed out in the www_server_log branch (www_server_log does not carry embedding context). Drilldowns by embedding_id will therefore only reflect submission-side data.
• A single vs_cookie that submitted multiple surveys is counted only once in the numerator; a vs_cookie that was served multiple surveys is counted only once in the denominator. NULLIF prevents division by zero when no get_survey events exist for a given segment.
• embedding_id, interface, and candidate_country are available as additional outer select filters for dashboard-level cross-filtering.
• Monthly trend widgets use DATE_TRUNC('month', t_create) and are ordered ascending by month.

8. Schema