love
02278ddac7
Add targeted tests for rejection paths, settlement overrides, and key analytics flows, and align localization/research docs with the shipped contract.
39 lines
831 B
Markdown
39 lines
831 B
Markdown
# Research Export Path
|
|
|
|
## Goal
|
|
|
|
Export study data without adding JSONB or ad hoc blobs.
|
|
|
|
## Source Tables
|
|
|
|
- `sessions`
|
|
- `experiment_assignments`
|
|
- `bet_intents`
|
|
- `settlements`
|
|
- `analytics_events`
|
|
- `analytics_event_attributes`
|
|
|
|
## Recommended Export Shape
|
|
|
|
- `session_id`
|
|
- `user_id`
|
|
- `experiment_variant`
|
|
- `locale_code`
|
|
- `event_name`
|
|
- `occurred_at`
|
|
- `attribute_key`
|
|
- `attribute_value`
|
|
|
|
## Export Flow
|
|
|
|
1. Join relational tables into a flat study dataset.
|
|
2. Filter by session, date range, or experiment variant.
|
|
3. Write CSV for quick review or Parquet for downstream analysis.
|
|
4. Keep the export job read-only and repeatable.
|
|
|
|
## Notes
|
|
|
|
- Use the server clock for ordering and latency calculations.
|
|
- Prefer one row per analytics event with child attribute expansion.
|
|
- Keep locale and variant fields in every export.
|