Note that much of this information is specifically geared toward edX.
This page supplements the wealth of knowledge in /wiki/spaces/AN/overviewand Mobile Analytics.
Analytics Events in edx-platform
Designing new events
When designing new events, see the /wiki/spaces/AN/pages/13205895.
NAMING CONVENTION
Determine if your new event is a research event for partners that goes into the publicly documented data package, or whether it will only be used for internal BI purposes.
- For internal-only events, use the special prefix 'edx.bi.' for your event name.
- For research events, make sure you tag someone on the Docs team, to make sure your new event is documented in the edX Research Guide.
Choose between server-side and client-side implementation of the event, and consider whether or not Mobile Analytics events need to be implemented separately for a native client-side feature.
Implementing in edx-platform
Implementing events in edx-platform is documented in the edX Developer's Guide and /wiki/spaces/AN/pages/30967007. This section provides some additional clarity and examples.
Server-side events (includes celery)
- Rule of thumb: Use server-side events where possible. More context is available to server-side events. Server-side events are less likely to have false duplicate events.
- Call tracker.emit
- Emits to Tracking Log and downstream tools (e.g. Vertica/Tableau, edX Research Package, etc.)
- Event must be whitelisted to also be emitted to Segment.
- In turn, Segment emits to other tools: Google Analytics (GA), Google Tag Manager, Optimizely, Qualaroo, Crazy Egg, Marketing tools.
- Here are sample commits for updating the whitelist for edge.edx.org and the whitelists for edx.org.
- At this time, it doesn't look like our code supports sending special labels to GA from server-side events. See /wiki/spaces/AN/pages/30967007 for possibly enhancing this.
- To send events directly to Segment:
- This is appropriate when server-side event does not need to appear in tracking logs.
- Call segment.track(), after "from common.djangoapps.track import segment".
- If appropriate, also call segment.identify(). Currently the LMS only does this at login/registration.
Client-side events
- Rule of thumb: Where you have an option of client-side or server-side event, choose server-side (see above).
- Logger.log
- Important: This should be renamed to Tracker.emit, but that requires a deprecation strategy including telling course teams.
- Emits to Tracking Log and downstream tools (e.g. Vertica/Tableau, edX Research Package, etc.)
- Event must be whitelisted to emit to Segment (see above).
- analytics.track
- Use this if you want Segment and its downstream tools only. If you want the Tracking Log or both, use the other call (with or without the whitelist).
- Emits to Segment and its downstream tools.
Ecommerce Events
For any ecommerce related activity, including any messages encouraging a purchase, you can use Segment's Ecommerce Events.
Any message encouraging a user to purchase could be considered a Segment Ecommerce "Promotion" Event. These, in turn, can be tracked in Google Analytices: Conversions > Ecommerce > Marketing > Internal Promotions. Here is an example of implementing a "promotion" event in edx-platform for a verified upgrade reminder.
Analytics Tool Usage by Team
| Team | Sub-Team | Details | Tools |
|---|---|---|---|
| Learner | Mercury Team | How we work | Google Analytics (GA), Tableau, and Optimizely. |
Tools
Analytics Tools:
- Google Analytics (GA)
- Google Tag Manager
- Firebase Analytics
- Fabric (Mobile)
- Segment (a.k.a. Segment.io)
- Tracking Log
- Splunk4Tracking
- Vertica (Data Warehouse)
- Tableau (Visualization using Data Warehouse)
- /wiki/spaces/PT/pages/159531726
- edX Research Packages
- Optimizely (A/B Testing)
- Qualaroo (survey tool)
- Crazy Egg (screen recording/heat mapping, may or may not be used)
- Marketing
- Sailthru (email marketing)
- Facebook ad reporting
Diagnostic Tools (includes Performance):
- /wiki/spaces/PT/pages/96698687 (query course content usage)
- Splunk (logs)
- NewRelic (errors, usage, health)
- Datadog (deprecated)
Consider the following data storage questions:
- Limits (e.g. Splunk Logs)
- Privacy (who has access?)
- How long does data live?