Publishing video over the Internet is a complex process in which many different components take part. Every byte of video distributed across the network carries its own expense.
The great advantage of Internet is the possibility to collect all kinds of data pertaining to every single video playback that allows to track anything all along the entire workflow, from its production to its consumption.
The potential of this data is huge: it allows you to find out everything in the context of each streaming session, from the content of the video itself, to the viewer behavior, to the performance of each component involved (CDN, network, player, device, …). Furthermore, you could potentially figure out which factors affect, positively or negatively, your audience’s engagement with your content. It sounds promising. However, are you harnessing this opportunity to help out your business?
The world is now filled with data. And video is no exception. Companies are collecting a great deal of analytics, but just few of those data points are actually being used. It is not only a matter of getting data, but also to effectively understand it and action data in the field. For this reason, it is not enough to just collect data. There has to be a process to analyze it, to grant data integrity along the process and to obtain insights. That is the bare minimum. But to embrace the full power of data, companies must learn how to predict future performance based on past observations and even prescribe actions to obtain the desired performance. In other words, it is all about actioning the data. This process must be governed by professionals, but the tools used escape from human capabilities, and here is where artificial intelligence plays a crucial role.
In its “2017 Planning Guide for Data and Analytics”, Gartner sharply represents the Analytics Continuum, where each capability builds upon the previous one: Description → Diagnostic → Prediction → Prescription.
Making sense of video analytics has two main problems. First one is to determine the most relevant metrics for each particular case (each playback may include hundreds of them). Second one is how you take automated and immediate actions according to these data, which will allow you to drive your business proactively and prescriptively.
In the case of Content Aware Encoding, also known as Smart Encoding, for example, optimization is performed based on the nature of the content, namely, which type of footage (action, talking heads, sports, movement, dynamic range, etc.) comprises a video asset. Thinking more holistically, we could feed that analysis using additional data about the delivery and consumption process. This approach allows to set up, prepare and deliver each content in the optimal way, truly adapted to each playback, each user and all circumstances involved, such as location, time, device, network status/availability and some others. This is the individual context or, as we want to call it, the unique playback fingerprint.
On top of that, another set of variables that can be included when considering the most adequate reproduction parameters are the business drivers for each publisher. Some companies’ video strategies are driven by quality, some others by cost, most of them by viewer engagement. There are business models based on subscription, some others based on ads. Just to add more variability, sometimes videos are a complement to some other sellable assets, as it is the case for fashion companies, car manufacturers or even professional sport leagues. Finally, factoring all those variables –the unique playback fingerprint and the publishers’ KPIs– allows to action over them and improve the viewer quality of experience, directly benefiting revenue or cost bottom line.