Leadership
Streaming services can leverage useful analytics to lower costs, increase service quality, and guide critical decisions. You can also use them to guide programming decisions and deliver relevant content. Conviva’s platform makes it easy to connect insight to action, but it helps to know what’s going on under the hood.
This glossary provides a strong background for understanding any discussion about streaming analytics. We’ll cover the background terms, video quality metrics, business performance metrics, and key technical vocabulary it helps to know when putting these analytics to work for your streaming business.
Key takeaways:
-
- You don’t need to be a data scientist to get the most out of Conviva.
- Holistic KPIs simplify things by bringing multiple technical metrics into the equation.
- Contemporary data analytics provide powerful insights quickly.
Foundational terms
Conviva offers a next-generation streaming analytics platform. However, how our clients use our tools is just as important as what the tools do. Here are a few of the general terms relevant to one critical Conviva application: video streaming analysis.
Over-the-top (OTT) services
“OTT” generally refers to streaming video. The name comes from the fact that it goes “over the top” of broadband internet service rather than on a direct infrastructure line from the broadcaster to the consumer.
Video embed: (behind sub wall) https://www.conviva.com/on-demand-webinar-the-road-to-ott-balancing-innovation-with-profitability/
Contrast this with traditional, linear cable TV, in which the publisher usually has more control over (and insight into) the performance of their infrastructure. Conviva’s next-gen analytics allows OTT quality control that rivals or beats linear.
Video on demand (VoD)
When you stream any non-live broadcast, then you are probably using a VoD service. These services typically rely on complex network architectures into which Conviva provides critical insights.
Content delivery network / media delivery network (CDN)
A CDN is an internet infrastructure service provider that helps deliver content across the internet and essentially extends the reach of a service, such as a media publisher. For example, it could let the publisher place content closer to viewers to make delivering streaming video more efficient. Many contemporary streaming enterprises contract with multiple CDNs, and the performance of these vendors is a major factor in audience experience.
Video player
A video player is the specific application that plays video. Traditional video analytics solutions limited themselves to events that happened in this small window, but Conviva measures the entire viewer experience. That includes ads, apps, and your infrastructure.
Critical quality of experience (QoE) metrics
One of the reasons our clients use Conviva to measure QoE for customers is that our platform takes a holistic approach to audience experience. We can empower publishers because the platform collects massive amounts of data and processes it efficiently. A breakdown of those details will follow, but for now let’s focus on the critical QoE determiners.
Streaming Performance Index (SPI)
SPI is the core of Conviva’s utility proposal for video publishers. It provides a single key QoE performance indicator that globally distributed organizations can operationalize across distributed teams and technically disparate efforts to unite in pursuing quality goals.
In other words, it is a North Star metric. The Conviva platform measures the metrics discussed below for every stream, processes them against variables in the technical environment (such as screen resolution, for example), and returns the SPI as a single, holistic representation of QoE for the entire population.
Due to the granularity of the data, you can drill down into specific audience subsegments, such as device type or service tier, with pinpoint accuracy.
Video start failures (VSF)
VSFs are when a video does not start. This can be caused by techical error, but isn’t always a quality issue. Sometimes a VSF can be the result of deliberate business logic, such as when a user at a geo-blocked address tries to stream a video.
This is a prime example of a metric that you need to place in context. Without context, your engineers might be chasing down start failures that aren’t actually errors. At the very least, including deliberate VSFs in your quality measurements will give you a deceptively low score.
Video playback failures (VPBF/VPF)
A VPF occurs when a video stops playing for a reason other than the user exiting the player or manually terminating the stream. Similar to a VSF, a VPF might be due to deliberate business logic, such as a viewer account limitation.
Exits before video start (EBVS/EBS)
The definition of an EBVS is relatively specific: an attempt must stop before the video starts without a fatal error. A classic cause is a viewer who gets tired of waiting for a video to load and leaves.
This is a difficult and complex issue for engineering teams, especially if they do not have proper, data-driven insight into the entire viewer experience. Conviva provides that end-to-end information, empowering your teams to quickly isolate the causes of EBVS, whether they stem from your delivery network, video player, advertisements, or any other measured aspect of your service.
Buffering (rebuffering) times
Buffering occurs when a video needs time to load before continuing. This is a classic frustration for streaming video viewers, and there are a host of possible causes.
However, not all buffering is equal. For example, the cause matters. Events that occur after a user navigates within the video player do not impact QoE significantly. On the other hand, those that arise from network congestion might drive viewers to leave your service.
Similarly, the type of stream is important. Live sports QoE is drastically impacted by buffering events — in a tense game, every second counts. In contrast, users binging classic TV shows on a VoD service are more tolerant. Conviva’s industry-leading benchmarking capability gives you insight into exactly how buffering impacts each segment’s QoE.
Other key QoE and performance metrics
Unlike endless buffering and playback failures, many other technical issues might not be deal-breakers for audiences. However, some still may affect QoE significantly, such as these key examples.
Video startup times (VST) and restart times (VRT)
Your VST is the time it takes between pressing play and the video start, not counting advertising. Buffering during ads also does not impact VST. VRT, on the other hand, is the amount of time the video takes to start after a viewer navigates in the player.
VRT and VST are important QoE indicators, particularly when you’re optimizing your service. Faster startup times and restart times will increase QoE.
After all, it’s called “video on demand”, not “video after demand”. Audiences expect responsive, immediate service — and Conviva empowers you to drill down and eliminate the root causes of any delays.
Bitrate
Bitrate is the amount of data being delivered in a video stream. High bitrates generally mean higher quality and more detailed images. Bitrates can vary by device or region depending on service reliability.Bitrate discrepancies in different audience segments can guide you to optimization opportunities.
Frame rate in frames per second (FPS)
Videos play multiple still images, or frames, in succession to create the illusion of smooth movement. Up to a certain point, higher frame rates will increase QoE. Low frame rates might indicate a CDN or encoding issue.
Rendering quality ratio
Rendering ratio is the number of frames per second that a video player shows divided by the number of frames that your video file could show. In other words, it is the rendered FPS over the encoded FPS.
If your service is not optimized, video players often show (render) fewer frames than they receive from the server. This results in a rendering quality ratio that is less than 100%. This is undesirable because it means you are delivering lower-quality video — and paying to store more video than you can deliver.
Application and ad analytics
QoE is holistic. It depends on a user’s experience both inside and outside of your video player. That’s why Conviva offers advertising and application insights.
For applications, Conviva’s platform allows you to see user behavior in context and focus on key groups’ behavior within your application. For example, you can call out rage clicking on specific devices, or hone in on audience segments that completed purchases after watching an ad.
In advertisements, you can get similarly powerful insights. You can see how your ad video quality impacts other critical QoE metrics, for example. Alternatively, you could measure overall QoE against advertising policy changes to take a data-driven approach toward optimizing ad saturation/revenue.
Performance metrics
Apart from direct QoE information, your viewers’ behavior tells you a lot. There are audience metrics you can combine with QoE to track and optimize the capacity versus the performance of your network.
Streaming plays, active streaming plays, and concurrent streaming plays
Your “streaming plays” metric is a representation of the number of successful playing events that your service had over a specific period of time. Failures and exits don’t count, regardless of the cause.
Active plays are successful streaming plays that last a certain amount of time. For example, you might consider 45 minutes of a 4-hour-long live-streaming event to be an active play.
Concurrent streaming plays is the maximum number of streaming sessions over a given interval. For example, you might have peak concurrency in the millions when live-streaming a major sports event. This is an absolutely critical metric when you’re navigating high network volume.
Total minutes, total bandwidth, and other totals
Measuring totals across your service gives you an essential, high-level, strategic perspective of the scale and performance of your operations. For example, total video minutes watched gives you an idea of the revenue potential of an advertising implementation.
Drilling down into these numbers can also empower teams across your organization. Conviva’s configurable dashboards provide instant executive summaries that lead to immediate, granular insights in just a few clicks.
Unique devices
This metric shows you the devices customers on your service are using. Tracking device types helps you define your end-user landscape, narrowing in on optimization opportunities, guiding development choices, and providing superior support.
Technical analytics terms
Accurate, actionable metrics require powerful technology. Check out common streaming data analytics terminology.
Data warehousing
Data warehouses and data lakes are traditional ways of storing analytics data. While they can still be useful, they are not efficient when you need to calculate complex video-streaming metrics. Running analysis with Conviva is often at least an order of magnitude less costly than doing the same thing with structured query language (SQL) from warehoused data, for example.
Census-based measurement vs. survey
A census-based system takes measurements from an entire population, similarly to how Conviva measures every second of every stream for every viewer on your service. A survey, on the other hand, selects representative samples. Census measurement is superior for many strategic and tactical applications in a video-streaming enterprise.
Real-time analytics
Real-time analytics are not technically instantaneous, but they are much faster than traditional methods. Even the most complex metrics are available with sub-60-second latencies. This is in contrast with older data models, which typically require a much higher computing cost in both time and processor load.
Time-state model
The time-state model is at the core of Conviva’s analytics engine. It creates a multi-factor, time-dependent story of changes of state across the network for each viewer.
This is an important contrast to older, tabular data models. Those older models require significant reprocessing to be useful for contemporary video-streaming analytics, whereas the time-state model is already in a usable form.
Time-state data makes it possible for Conviva to provide complex, census-based calculations in real time. This type of data abstraction is also what makes the cost of insight much lower with Conviva than it would be with a traditional model.
Benchmarking
Benchmarks are essentially quantified technical standards. A good QoE benchmark should be specific to the goal you are attempting to reach, your industry, and the expectations of your audience.
For example, SPI is backed by Conviva’s industry-leading benchmarking. There are different, configurable target levels that help you track and accelerate progress in your optimization efforts.
A more complete discussion
For a more detailed, technical look at Conviva, check out our massive API library in the developer portal. Better yet, schedule a demo today to see Conviva in action.