In our recent interview with Marty Roberts, Wicket Labs’ CEO, we explored a broad range of topics. From how to use viewership data in streaming, to machine learning and the future of the industry.
Q: What do you guys do?
A: We focus on telling a story with data so you can really understand what’s going on with your business and what actions to take, based on where your business stands today. Essentially, we help to answer questions such as: When should I be talking to my audience? How do I better convert them so they become subscribers? How do I save a particular audience member if they’re at risk of leaving the service? So you really think through that whole customer lifecycle.
We use the data from our analysis and export it into marketing systems so that our customers can take action on their OTT analytics. With this, they can create better acquisition and engagement campaigns to get more people onboard.
The value of data really comes into its own when you’re actually taking action. Completing the circle essentially. Plugging it back in, knowing when to talk to your audience members, how to talk to them, and what are the most effective messages to talk to them with. So that’s what we’re doing with our customers today.
Q: It’s really cool what you guys are doing. How does your customer use your platform?
A: Our customers have access to a scorecard with all kinds of charts and graphs, that we call wickets. They help tell a story about which marketing channels are most effective, what are some of the biggest causes of churn that’s going on right now in the business, or how people are engaging with the service. We are big enough now where we can actually compare that against benchmarks so you can see if you’re above or below.
Let me explain a simple use case that we have found happens all the time, for example. You have a big show and the marketing team keeps promoting heavily. It ends up receiving a lot of video views, because people watch more than thirty seconds of the video. Well, we have what we call the attention index, where we pick everybody that watches more than 75% of a video and we subtract those who watch less than 10% of the video. Kind of like a net promoter score, where we find that on those very popular videos, sometimes everybody’s falling out. They’ve got an attention index of 10 or even as low as -20. Nobody is making it into that 10% mark and so you have this false positive.
You have to go a little bit deeper into the data to really understand what’s going on. Understand what’s the entertainment value of each show or each movie, if I should be promoting it as heavily as I am, or if it is just killing it. And in that particular case, to everybody who loves that show, I should be promoting it more.
Q: Going back to the beginning then, what was the problem that brought the company into existence?
A: We’ve been in the digital video space for a very long time. My co-founder founded one of the original online video platforms or video management systems. A company called thePlatform, that was acquired by Comcast in 2006. I joined right about the time of the acquisition and we worked together for another eight years running the subsidiary of Comcast. Then I took over for him as CEO for two years.
When I left, about 65% of all the US broadcast and cable networks were using us to get videos online and to their customers’ Roku device, iPhone, and other devices. It’s just very much on the operational side of things.
At the time we were aggregating a lot of metadata for Content Discovery. We were also gathering a lot of entitlement data, which is essentially who can watch what video, when, and where. Our customers would come to us and tell ask, “can you also aggregate our analytics data?” That was way outside of our core competency and our road map and we never thought that was part of that company’s mission.
After leaving the company we came back together, about three years ago, and started to look at this particular problem. If you think about it, traditional media companies have never had a direct relationship with their viewers before. Their customers were advertisers and cable companies. You would tell them how many people with certain demographics watch their show. Pretty scarce and obtuse data.
Now it’s a strategic imperative for media companies to establish a direct relationship with their audience and data is kind of the conduit to make that happen. But obviously, it’s a harder problem to solve. Making it a really interesting opportunity for our company to work and grow inside of this environment. Really helping these media companies better connect with their audience.
Q: Building off of that, there is so much data that can be collected. What should be the approach to handle all of it?
A: One of the problems that exist is an information overload. A company we talked to told us, “I’m looking at our dashboards and we have 218 of them.” And he does not know what’s important and what’s not important.
That is why we spend a lot of time curating the data and the presentation to make sure that it tells a very actionable story, so as you look at that data, you not only understand where the business is, but you understand how some of the decisions that you make, can affect the business over time. That’s our aspiration. That, as you look at a chart or graph it’s pretty clear where the business has been but also what routes it can take, based on this data.
Some great examples of that would be a free trial. Every video service has people that come in and watch one show or one movie and then they stop to engage. Well, typically, you only have seven days to identify those users and do some outreach to improve the conversion percentage, as they work through the cycle.
The first thing we have to do is identify what percent of your trial is stalled out at the moment and show exactly what their engagement level is comparing those that are stalled vs those that aren’t. Once they are identified you then can, for example, kick-off an early email campaign. So we really work to make that whole cycle, with carefully curated data, easy to understand.
Q: How are you using AI/ML?
A: Our application is a risk analysis on who’s going to leave the service. We’ve got around 65 different data features. We look at things like content, the activity of the service, and experience. This data gets piled into the model, so we can accurately predict which customers are actually at risk of leaving the service. This is our customer happiness index and it provides a CHI score for each audience member.
Machine learning is very interesting and a very valuable feature for our customers, as they can proactively identify a group of customers at risk. We started going over why some users were and found their activity was above normal. Which seems good at first but when you dig into those users, they fall into a binge-watching category. They’re just cranking through your content library.
You then have to take on more aggressive actions, in terms of communicating the content that’s available. As soon as they complete one show, they then become at risk of leaving the service. Normally, you’d look at these users thinking it’s great that they’re using the service, but not in this particular case. We can prove a causal relationship with their use of your service and churn. Really interesting things that come out of our machine learning model.
Q: What are the challenges you see around it?
A: The first challenge is that a lot of companies think that if they have two or three people in the data department, they can do everything themselves. You used to be able to set up data in a SQL database and if you had a business analyst with SQL skills, you could probably answer most of the questions on there.
Today the data sets are getting so large and the queries are getting so complex that SQL doesn’t keep up anymore. You have to employ more of a data engineering perspective and bring the same discipline used in software engineering to data engineering.
The second thing we look at is how clean the data set is. No one has a perfect data set to begin with. You have to handle data anomalies and keep track of changes over time. If an API changes over time, it can break reports downstream. We have to catch those variations and quickly address them. This requires a sophisticated data pipeline process.
It starts out as a black box. You feed data to a model and it gives answers on the other side but never tells you why, so you have to go the extra mile.
Q: In this time of strong competition, with different technologies on the rise, how do you see the streaming landscape evolving in the next years and what is the role you’ll play in it?
Obviously there’s the big players that are trying to dominate the space. They measure their success in the tens of millions of users. That’s great and if you have a Content Library that has that broad applicability, absolutely go that direction.
On the other hand, the power of the internet is that you can take a niche that people are interested in and because it’s a worldwide audience, actually build up a healthy subscriber base.
We have customers that are in anime, news, sports, and all kinds of different passion areas that people are really into and they’re building great businesses. They’re quickly getting passed a thousand subscribers and then setting their sights on a million subscribers. A million subscribers who spend 5 bucks a month is a pretty good business
There’s room for lots and lots of different services that consumers are going to jump into. Competition will always be there but I believe in the power of building what you think is a niche or a small pasion area. Then when aggregated across a global audience, it can turn into a healthy and successful business.
About the author
José is part of the Teltoo team, whose software-only decentralized video delivery technology helps live-streaming providers to improve quality and optimize delivery costs.