Dr. Daniel Wegman - July 30th 2021
Let’s take a not-so-imaginary scenario: You buy a new smart TV, plug it in, turn it on and discover you have more than 100 channels to browse arriving directly to the TV, thanks to the magic of the internet. After a quick look around, you find out that your favorite channel is available. From that point forward, you watch your favorite news channel in the morning and after work, you use your favorite video-on-demand (VOD) platform to enjoy the show you are binge watching. Do you think you’ll ever seek out new channels? Or will you keep to your routine?
That’s what I’ll be focusing on in this month’s A Wurl of Data blog: are people watching new channels?
First off, let’s set the rules. What exactly will we be calculating? We want to know the time it takes for someone to begin watching a new channel. For that, the first thing we do is look at the timestamp of when a viewer first watched TV on the platform, then we look at the first time the same viewer watched a specific channel. A lot of “first time” channel viewing happens on the first day a person joins the platform – any channel watched on that day would be considered new. With that in mind, we calculate how many days it takes for a specific viewer to watch a channel they haven’t watched before. As expected, almost half (~46%) of all new channel views happen on the same day the person joins the platform. After that, the numbers start to drop very fast, ~2.7% of new views happen on day two and 1.5% on day three. In Figure 1, we see just that. It shows the cumulative percentage of people converting to a new channel beginning at the time they join the platform.
At what point do we consider that the viewer is still exploring what they have in their new smart TV, and at what point do we consider the person has established a routine, and their viewing is officially a new habit? These are important questions because they will tell us when we can consider that a viewer has converted to a new channel. If we set the parameters of a conversion to 28 days after the viewer joins the platform, then 63% of all new channel views happen within that time. Meaning, about two thirds of all new channel views occur in the first four weeks after the person joins the platform, and one third after. We call the latter conversions (they have seen the light!). But don’t get confused, this is not religion, our conversions don’t mean that they stop watching other channels they already watch.
So, why do we care about these conversions? Well, simply put, more channels means the viewers are watching more TV. Actually, the correlation of channels viewed vs. total hours viewed is so high it seems inaccurate (but it’s not, we checked this multiple times). We can see this in Figure 2., where we plotted the average hours of viewing (HOV) per day vs how many channels the viewer has watched in total.
The next thing to calculate is how many people are converting per month. While we can’t show you the actual data, we can show the rates (or percentages). We take all the people that converted to a new channel and divide the number by all viewers for a platform during that month. What we see is that over the last year things have changed, but in general the number of people converting is about 7-8%, with most people trying new channels in January. Why January? Well, your guess is as good as mine. Maybe the whole “trying new things with the new year” is true? Or maybe people are gifted new TVs over the holidays.
What about the people who actually convert and try a new channel? How long do they keep watching for? Well, if we again take those who converted, we see that ~60% try the new channel for one day (in other words, if we eliminate one day converted viewers, we will only keep 40% of conversions). We see that 15% of converted viewers watch the channel for at least 50 days, and 10% for at least 100 days.
Finally, let’s say I’m a content company and I’m very interested in converted people. More viewers = more HOV = more advertising views = more money (simple math, right?). The plot we showed above for conversion rate per month shows a conversion of around 7-8%, but that conversion is for all channels. So, what is the percentage of people who convert to any new channel when we actually look at the conversion rate of each individual channel? That is shown on the boxplot in Figure 5. For those who don’t understand boxplots, here is a quick explanation: the box represents the 25-75% of all the data, the line in the middle of the box is the median, the legs cover (almost) all data – the left leg being 0 to 25% and the right one being 75 to 100%. The leftover diamonds represent outliers (points that are so far from the median that they are considered extreme values). The plot indicates that most channels have a conversion rate of less than 0.4% (with a median of less than 0.1%). The best values were a little higher than 1%.
So what does this tell us? If I’m a content company, I can’t just depend on people randomly deciding to watch my channel to increase my viewage. There needs to be a better way… and in fact, there is. Stay tuned for future Wurl of Data blogs to learn more.