Unified Telemetry/Status reports/Android/February 4 2016
From MozillaWiki
Contents
UT for Android status report January 16, 2015
Overall Project Health
Last week: Green
This week: Green
Project summary
- Project in Green as no known blockers exist
Exec Summary
- Q1 deliverable for unified telemetry in android, powering executive dashboards
- iOS is using Adjust for some KPI's, hence the focus is on android as since we turned V3 off and are currently in need of more data
- Ratio (MAU/DAU) is currently planned on being calculated by use of Adjust
Accomplished for Last Period
- Regular meetings and roles defined
- First iteration (v0) of Android core pings has landed - This is enough to handle active users and retention and includes experiments data.
Planned for Next Period
Validation:
- checking the outgoing ping format and ingestion into pipeline
Client:
- client work on the next iterations (ping persistence, session information, ...)
Pipeline:
- 1242258 - make "core" pings available in Spark by docType
- 1243723 - extract more dimensions from "core" pings (for use with get_pings() etc)
EPM:
- Data collection review
Risks and items to retire
Dashboards and consumers:
- Dashboard form core pings (need min requirements)
- PM input
- Significance of abnormal session and search counts for first rev.
Validation:
- automated tests
- manual QA
- mobile team has in-house and SoftVision QA
- scope of v0 ping version is not be appropriate for heavy test effort
- manual QA will happen from v3 on
- Android Nightly has small population, so:
- Sanity-check on Nightly
- use Beta for validation
- update latency to beta ~2 weeks to 90% of population
- server-side analysis
- quality checks for ordering of pings, etc. and verifying the data looks sane
- sanity checks at volume (look for large gaps etc)
- validate against adjust data?
- we should get to an ongoing "health of data" report from there
- QA & analysis work needs to be defined and scoped/packed
- schema validation
- pipeline-side schema validation support is on the way
- pipeline team may be interested to start this with the "core" ping