By Melissa Yurash, Sr. Customer Success Engineer, Conviva
Why Real-Time Matters
Have you ever tuned in to your favorite TV show at the end of a long day, only to experience endless rebuffering? How long did you wait or did you try turning off your TV, restarting your device, clear cache, or just switch services? Did you blame the device, ISP, CDN, CMS, or the app? Viewers are becoming less and less patient when it comes to issues in their viewing experience. And with an abundance of services to choose from, any quality issues can cause a viewer to quickly switch to a competitor.
That’s why Ops teams need the ability to react faster than their users. We call this Real-Time Actionability. Real-Time Actionability refers to the ability of streaming companies to diagnose and address issues with viewer experience while the viewer is still in session. Real-Time responsiveness is critical because viewers are increasingly less tolerant of service interruptions and can switch more easily than ever to a competing streaming service. Failure to respond to issues before users decide to look elsewhere can cause massive, irreversible losses in viewer retention, and ultimately impact overall revenue.
Continuously monitoring, identifying, and addressing quality errors across a global audience is a massive and complex undertaking, but the world doesn’t wait. And as the scale of devices, networks, geographies, and content types continues to expand, time in which customers expect their issues to resolve is only getting shorter. To accurately address specific challenges under these conditions will require increasingly granular analysis.
For example, when you have viewers who can’t get their video to play they just want you to fix it – or even better, to detect and stop an issue before it affects them. To accomplish both of those things, you need a way to rapidly diagnose the root cause of that increased Technical Video Start Failure. It isn’t enough to know it’s happening you need to know, right now, that it’s happening on Android devices, in New York, on one of 10 CDNs, at 7:15pm AM last Tuesday, on one of your most profitable channels. And to even be in a position to compute that kind of in-depth, contextual root-cause analysis, you need to collect, cleanse and process millions of data streams and hundreds of combinations of attributes at scale in real time.
To stay competitive in an increasingly saturated streaming market, speed has becoming a critical economic differentiator. Keeping the 10 – 20% of viewers you would otherwise have lost to the competition translates into higher ARR. And improving that experience also means they will watch longer – in some cases a lot longer (3 vs. 87 minutes) which means you have greater opportunity to monetize those additional moments with premium offers or ad revenue.
Taking Action on Your Streaming Data
How can you implement real-time responsiveness into your streaming operations? The key is a blend of flexible, multi-dimensional analysis, AI-powered alerting, and effective resolution process.
Visualization of Data in Context
While real-time insights are crucial, access to real-time insights in context makes a significant difference in reducing time to resolution for errors and outages. As mentioned above, detecting an increase in video start failure across all devices is valuable, but to be able to identify which devices, on which CDNs, and in which region provides even richer context to be able to solve that issue.
Visualizing that rich, contextual data is also critical to real-time responsiveness. Leveraging a single dashboard on which you can monitor users’ service interruptions through trusted and measured KPIs in real-time makes it easier to detect even slight decreases in overall performance and how an experience issue like an ISP outage, a bad Device OS Version update, DRM key error, or encoding issue can manifest differently for several different metrics.
A viewer isn’t concerned about the root cause of a technical disruption. They only know that they are experiencing multiple buffering interrupts, repeat errors on a specific show, or terrible picture quality on the basketball game. Ops teams need to be able to immediately filter and segment from aggregate to the impacted cohort, so they can immediately begin to diagnose where and why an error may be occurring, making root-cause analysis more efficient.
Conviva’s Metric Filtering allows teams to filter QoE data on granular metric buckets to more quickly uncover performance outliers that might be impacting multiple key performance metrics. In just a few clicks, you can go from the comprehensive dashboard of all viewer experience data in Trends to a highly filtered snapshot of a specific subset of viewers experiencing fluctuations in multiple key performance metrics.
It’s the difference between fixing an issue in 8 minutes versus 182 minutes. That increases your viewer retention and keeps users watching more content for longer.
Automation is a critical component of real-time actionability. Strategic technical operations teams use machine driven alerts to highlight notable fluctuations in performance that a human might miss. Conviva offers powerful AI Alerts that work to reduce MTTR by alerting teams to slight degradations in quality as they’re happening. Leveraging automation is crucial to increasing the speed of response and uncovering issues that are just starting to negatively impact the viewer.
Effective automation can also be an indicator of the accuracy of your KPIs and efficacy of your processes. When AI Alerts are properly instrumented, you can verify the accuracy of the algorithm by how fast the alert is triggered, avoiding false positives and ensuring slight variations in performance are addressed.
Conviva has enhanced our AI Alerts to include connection induced rebuffering, automatically surfacing those viewers who are facing significant rebuffering that is not caused by viewer behavior. With the new AI Alert for CIRR, rather than RR, you can automatically uncover technical issues with video playback and address them in real-time, ensuring that no viewer exits your service because of excessive rebuffering that was in your control to prevent.
Focus on Long Term Outcomes
One out of every three customers will leave a brand they love based on one bad experience. So what constitutes a bad experience? How long does an issue have to persist for a user to give up and go to another service?
With hundreds of combinations of devices, locations, ISPs, CDNs, assets, and app versions, it’s impossible to manually monitor all of those unique combinations at once. That’s why you need a comprehensive set of algorithms to determine which multi-combinatorial segments of users are experiencing an issue right now that significantly deviates from that same cohort’s baseline of performance.
And what about a slow degradation over time that is not a sudden spike, but otherwise goes undetected? You need to be able to identify, measure, and react to those changes in metrics too, even when the deviation is not significant enough to prove statistically in 3 minutes. Perhaps a slow propagation of an app version, an OS Version, or the popularity of a piece of content does not make the issue measurable and significant until hours, days, or even months later. But real-time actionability is still meaningful in this scenario because it means that teams can be notified and react as soon as the deviation occurs to quickly resolve the issue.
Notifying Analyst, Operations, NOC, and Automations teams of any deviation on any cohort of viewers as quickly as possible means that your teams have the fastest possible MTTR and your service has a high up-time . This keeps subscribers watching and enables higher ad impression goals to be met.
Real-Time Responsiveness in Action
Imagine identifying, being notified, and having enough context for an issue that you can easily fix it – in just 3 minutes. That’s the power of AI Alerts.
Recently a Conviva customer with an ad supported platform set up AI Alerts to detect and alert the team to any anomalous decreases in concurrent plays. One day, a decrease in PCP on a specific subset of devices triggered an alert, this was quickly followed by additional alerts as concurrency continued to drop. The team immediately got to work diagnosing and addressing the cause of the drop in concurrent viewership. The math behind the urgency is simple – fewer viewers mean fewer viewing minutes… and ad breaks, and advertising revenue.
Within 8 minutes of the initial alert, the issue was resolved, and viewers started returning to the service. However, an estimated 7% of those impacted never returned.
In this case, the issue was resolved so quickly that less than 3% of total viewers were impacted at all, which raises the question of how many more viewers and how much more ad revenue would have been lost if it had taken 2-3x longer to address this issue? Consider if this provider did not have access to real-time viewer experience data and AI-powered alerting, how long could an issue of this magnitude go undetected?
Actionability: The Key to High-Quality Viewer Experience
So, what are the critical elements you need to take action on your streaming data?
- Multi-dimensional Analysis: Flexibility to drill-down into streaming data across infinite dimensions in real-time to surface anomalous events and behaviors across tens of thousands of cohorts for easy diagnosis and decisioning on the right actions to take.
- Time-State Metrics: Context is key. You need fine-grained and actionable metrics that consider the state of the viewer at every moment, and understand the difference between viewer-induced, device-induced, connection-induced, business logic behaviors, and more.
- AI-Powered Alerting: AI-driven automatic detection of anomalous shifts in performance and uncover hidden patterns to drive effective decision making and error resolution.
- Industry Benchmarking: Critical industry-level insights to determine if performance is aligned with competition and to prioritize performance improvement opportunities that can be strategic differentiators.
As you can see, real-time actionability is very different from real-time monitoring. It goes beyond observation and identification of potential issues, enabling you to proactively detect problems with automated alerts, rapidly drill down into why they are happening, and model the real world so you can eliminate the noise, prioritize and immediately take action that’s meaningful for your customers and your company’s bottom line.
To stay competitive, streaming companies need to unlock the power of real-time actionability to improve viewer experience and ensure their operations are built to drive profitable subscriber and advertising growth.