AWS Partner Datazoom integrates with AWS for real-time video analytics

This blog was co-authored with Diane Strutner from Datazoom.

The media and entertainment industry is arriving at an important phase of streaming video deployment—observability and optimization. In order to achieve a fully optimized end-to-end video stack, it must be monitored and adjusted for trade-offs between performance and efficiency or profitability. The first goal is to measure the end-user Quality of Experience (QoE) in order to understand if and when users are negatively impacted by system performance.

This blog post outlines the benefits of understanding video QoE, and offers a suggested framework for building a QoE analytics solution using the combination of AWS Partner Datazoom’s Video Data Platform and Amazon Web Services (AWS).

Building an end-to-end video streaming workflow is a complex process and it is difficult to understand how well it is actually working for your viewers. Degraded or error-prone streaming experiences can cause users to stop watching, which lowers consumption of content and creates viewer churn.

In order to understand if experience issues lead to negative viewing outcomes, data can be used to calculate, analyze, and filter viewing consumption patterns. There are various ways to examine viewing behavior, and having granular data and flexible tools to analyze and evolve monitoring over time is key.

The first step in understanding the end-user QoE is collecting data from the video player where streams are consumed. However, challenges in collecting this dataset is threefold:

• There are more than one hundred data points that need to be analyzed
• Data is not standardized across video players
• Data needs to be re-ordered and cleaned before it can be used

Datazoom solves these challenges by offering a software platform that continually gathers data from endpoints, including video players, CDNs, Origin Services, and SSAI services, and sends out normalized event stream that adheres to standardized data definitions, and routes data to an ecosystem of Connectors.

Datazoom has Connectors for multiple services on Amazon Web Services. In the following diagram, you’ll find Datazoom’s suggested workflow to stand up video analytics using Amazon Web Services. Using this setup, data is ultimately visualized and analyzed in Amazon Managed Grafana, a service for scalable and secure data visualization of operational metrics, logs, and traces.

If you’re using Datazoom’s platform, there’s an easily modifiable template for Grafana available that works out-of-the-box. There are several key metrics to monitor in understanding the viewer experience and what may be impacting it.

The first step is to establish a current consumption baseline for each service. Here are some suggested metrics:

• Content Minutes (Count) – The minutes consumed in a given time period
• Content Requests (Count) – The number of requests for content in a given time period
• Content Starts (Count) – The number of successful requests for content that lead to video playback
• Content Sessions (Count) – The number of individual content starts
• App Sessions (Count) – The number of total sessions, which may contain one or more content sessions

Next, examine how the viewing experience impacts the consumption baseline. Here are some suggested QoE metrics that, when correlated against consumption metrics, can expose negative outcomes:

• Exit Before Video Start (Percentage) – The percentage of viewers who exit a video playback experience before the first frame is visible
• Time To First Frame (Average) – The average time it takes for a video to start playback
• Stall Duration (Average) – The average time spent with the video stalled during playback for the session
• Bitrate (Average) – The average bitrate streamed for the session

When QoE metrics negatively impact the viewing experience, the next goal is to understand how and why. Filtering can help identify commonalities amongst poor experiences to help identify where to focus and prioritize problem-solving efforts. Here are some common filters to consider using when examining QoE metrics:

• Title
• App / Website / Service Name
• Geography
• Device type
• ISP
• ASN
• CDN
• Time of day
• Connection type

Using consumption baselines, QoE metrics, and filters allows teams to dig into data and take actions to identify the root cause of behaviors. An example CFA (Contributing Factor Analysis) might look something like:

When our service went from .25% to 1% stall ratio, minutes consumed fell by 10%. Our team determined there’s an overall 10% stall ratio for viewers in Seattle, Washington, which accounts for this increase. The team is looking into additional data from the services (Application monitors, CDNs, Origins, etc.) involved in delivering content to these viewers during that time. 

By using the integrated solutions with Datazoom and Amazon Web Services, companies in the media and entertainment industry can build real-time video analytics, and use the insights gleaned to improve the viewing experience.

AWS Partner Spotlight

Datazoom is a Video Data Platform offering software and services to capture, standardize, enrich, sample, and route data in real-time from players, CDNs, ad servers, and other third party systems used in the end-to-end workflow of video. Its data pipeline works in sub-second real-time to power observability, analytics, and reporting solutions that support product, advertising, engineering, and operations teams.

Contact Datazoom | Partner Overview