OpenTV Analytics
OpenTV Video Platform supports business intelligence and analytics to enable the operator to monitor and analyse user behaviour. It allows them to:
Gather consumer usage data from all devices
Visualize data in included Tableau dashboards
Combine with other data to define their own metrics and monitor their KPI’s on operator’s own dashboards
Access to the underlying data warehouse for deep-dive analysis
Act on the data with wider OpenTV products to promote content, resolve delivery problems and optimize commercial offers
Export data for further analysis
Quality of Service
The Quality of Service (QoS) module allows operators to capture quality of service data from client applications, measure it, score it against ‘good’ performance levels and then identify problems with their service.
The data analysed covers:
Video Sessions
Session quality score
Buffering
Bitrate
Bitrate downgrades/upgrades
Frame drops
Errors
Time to start
Active Users
Device
Content Type
Network
Content Quality
Application
Accounts
Quality of service data is prioritized in the data pipeline to be available for reporting within 10 minutes from occurring on a device (including device buffering).
Error data can be used as a metric and view in detailed with ability to drill down to detailed individual errors.
TV Analytics
The TV Analytics module allows operators to measure the performance of their service with consumers.
The data analysed covers:
Content
Time Viewed
Live / Video On Demand / Catch up/ DVR
Series / Episode
Genre
Channel
Video format/quality
Completion
Consumers
Sessions
Devices
Daily/Hourly usage
Max concurrent
Devices
Form Factor (Mobile, Tablet, Big Screen, Desktop)
Device model
Technical Specifications
Application
Platform
Connection Type
TV Analytics data is de-prioritized in the data pipeline (compared to QoS) to be available for reporting within 1 hour from occurring on device.
Video Platform Operational
The Video Platform Operational module allows operators to montitor the operational behaviour of the OpenTV Video Platform.
The data analysed covers:
Content
Metadata
Workflow Activity
Accounts
Activity
Devices
Products
Entitlements
Video Platform Operational data is de-prioritized in the data pipeline (compared to QoS) to be available for reporting within 1 hour from occurring.
Data Collection
Data is collected for OpenTV Analytics from
Client Applications via the OpenTV Analytics Agent, integrated by the application vendor, which supports:
iOS
Android
HTML/JavaScript
Via the OpenTV Analytics Collector API
Via the OpenTV Video Platform Activity API
From OpenTV Video Platform (e.g. Metadata, Workflow)
Redshfit Datawarehouse
Collected data is processed into the OpenTV Analytics Redshift data warehouse.
The warehouse contains the data supplied by the data collection assembled into a data model. This allows event level reporting and a series of aggregated views to allow rapid aggregated reporting.
The warehouse can be accessed by Tableau or SQL access with other reporting/analytics tooling.
Tableau
The dashboards are delivered as a series of Tableau Cloud (https://www.tableau.com/products/cloud-bi ) dashboards which provide pre-defined visualization for the data.
Tableau is an off-the-shelf product which provides a wide range of dashboarding/visualization features including:
Dashboarding
Alerting
Metrics
Fair Use Policy
The DWH SQL access is governed by your NAGRAVISION SLA which includes terms for using the services as intended.
The SQL access is designed for aggregated reporting for the TV service, or detailed reporting on specific consumers/devices/content. It’s not designed to support bulk replication of the data. For bulk data access either the raw data can be accessed via S3 or additional redshift nodes can be provisioned (additional cost) to provide bulk access.
Things that are 'fair':
Aggregated queries on aggregation tables across the full time period
Aggregated queries on raw data with 'fair' time periods/groupings
Running the Tableau Dashboards or creating similar dashboards in other tools
Fetching data for single account
Examples of things we know break fair use include:
Dumping whole DWH tables via SQL
Non aggregated queries across large datasets (e.g., select * from activity)
Aggregated queries across 'raw' data and long time frames (e.g. months/years)
Writing in-efficient SQL that has large cost for the result and can be easily optimized
To help manage the fair usage policy the Redshift database uses a ‘Work Load Manager’ to prevent some queries being executed. The rules this applies aim to implement the above guidence and include a max-returned row count limit of two million rows and max execution times per query of 6 minutes.