Does a bad relationship with your ad lead to abandonment issues? (a FAST perspective)

Are people watching ads?

We’ve all been guilty of changing the channel when an ad break starts. But that presents a potentially expensive lost opportunity.

A 30-second ad during the 2024 Super Bowl costs 7 million dollars, a 75% increase from 10 years ago. Combined linear and CTV ad spending is set to hit $90 billion in 2024. Such a large investment highlights the significant value companies place on advertising. All this spending leads us to our question of the day:

Are people actually watching ads? Or do they switch channels when an ad break occurs?

In this article, we hope to answer those questions, along with two further $90 billion questions: 1. How many ads do people actually watch, and 2. If they change the channel when an ad break starts, do they return to the original channel afterward?

A quick glossary

Before we tackle our questions from the previous section, here are some quick definitions:

• Session: A contiguous time interval a viewer spends watching a channel. A session ends every time a person turns their TV off or changes channels (in which case, a new session starts on the new channel).
• Ad break: A specific period during a streaming video where advertisements are displayed to the viewer. An ad break usually contains multiple ads.
• Episode End: This is not a general definition, but for the rest of this blog and in the following plots, we will define the end of an episode as a one-minute interval following the end of a TV episode or movie being played.
• Content: Again, this is not a general definition but just for this blog. Content is all the time the episode or movie is playing outside of ad breaks or the one minute we just defined as the episode end.

Let’s talk about ad breaks

To answer the questions we posed at the start of the blog, let’s examine the image below. It contains four panels.

On the left side, we see data for channel 1 in platform 1, and to the right, we have channel 2 in platform 2. We will compare two different channels side by side like this to illustrate that the findings are universal, and independent of channel or platform.

On the top row, we have plotted the count of sessions ending, and below, we have the active viewers. All of the plots are for a 4-hour time period shared for each of the channels.
Channel 1 is an episodic channel, while channel 2 is a movie channel.

Now that we know what the images are about, let’s discuss what we see in them. The red areas are the ad break (usually around 2 minutes), and the white lines are when the movie/episode ends.

If you look closely at the top images, which plot sessions ending, we see clearly that viewers change channels (or turn the TV off) right at the start of an ad break; on the top plots, there is ALWAYS a peak on the left side of the red areas, showing that many sessions end right at the start of the ad break.

We can see a similar effect in the bottom plots, where the number of active sessions reduces throughout the ad break. Basically, many viewers end sessions right at the moment an ad starts, and as the break progresses, more people drop off.

However, the bottom plots show that sessions increase right after the break ends.

A quick summary here would be the following: many people change channels during an ad break.

Next, let’s look at whether viewers come back.

Are viewers boomerangs? (Do they come back?)

For the rest of the blog, we wanted to check viewers’ behavior across multiple channels. For simplicity, we took data from 28 channels on two different platforms for seven days. This will ensure that the patterns and conduct we will discuss are generic.

First, let’s have a quick explanation of box plots. Basically, they show the data distribution. It looks like a rectangle (the box) with lines (whiskers) stretching out on each side.

The box represents the middle 50% of the data. The whiskers extend to the smallest and largest values, excluding any outliers, which might be shown as separate dots. This plot helps quickly see where most of the data lies and how spread out it is.

Viewers lost per minute

The next step in our analysis is to see how many people leave a channel, depending on the content type (ad, episode end, or normal content). For this, we look at the percentage of sessions that end. Here is the methodology used to calculate the percentage of users lost per minute.

• Users lost: Sum of session ends during an ad break or content before changing to a new type of content
• Session count: Max active sessions during break or piece of content before changing to a new type of content
• Time: Minutes the content type lasts before changing to a new type

The plot below shows how many sessions are lost per minute. For example, if we see that the median per channel is ~4% of viewers lost per minute of an ad break, it means that, since ad breaks are two minutes, there is usually a loss of 8% of viewers during said break.

This plot shows something that is quite simple to digest: the number of people leaving during the content part of the show is about half the rate during ad breaks and one-third the rate at the end of the episode.

Percentage of content type and how many viewers leave per type

For the next image, we will see how many people end sessions during a break, the end of an episode or just during the episode/movie. This will allow us to understand if the people who end sessions during an ad break are unusual.

The two different color markers show two different measures: the red bars show what percentage of an hour the relevant part takes up — that is, how much of an hour is spent watching ad breaks, content, or the end of an episode. The blue marker shows what percentage of people leave during ad breaks, content, or episode ends.

The top red bar shows that for our 28 channels, the ad break takes up between ~10 and ~20% of viewing time (depending on the channel). To put that into plain numbers, there are between 6 and 12 minutes of ad breaks every hour.

However, around ~20% to ~40% of the viewers leave during this time (top blue bar). If we keep analyzing the plot, we see a mismatch between behavior during the content portion of an episode and the time when most people leave.

If the blue median is to the right of the red median for a particular section, that means people disproportionately leave during that section. If the blue median is to the left of the red median, that means people are less likely to leave during that section.

This confirms what we thought: people leave disproportionately more during ad breaks than programmed content.

Percentage of viewers who leave and then return within 30 mins:

• 26% of users that leave during ad breaks return within 30 min.
• 12.8% of users that leave during content return within 30 min.
• 12.3% of users who leave at the end of an episode return within 30 min.

The data above highlights that viewers who leave during ad breaks return at a higher rate than viewers who leave during content or at the end of the episode, with around 50% of viewers not returning in the 30 minutes after they leave.

However, averages can be deceiving since they don’t include distribution information. When we break down how many viewers return (or don’t!) on a per-channel basis, it shows the distribution of user return rates — and whether channels can do anything to influence.

This chart shows that the variation in returning viewers ranges from 10 to 50%, with most data points falling on the lower end. We can see that of the 28 channels, 8 of them fall in the 10-15% or 15-20% bins. These channels seem to be good at getting viewers to return. On the other hand, 4 channels fall into the higher bins of 40-45%, 45-50%, and 50-55%, showing a significant number of viewers who do not return within 30 minutes.

All-in-all, this data shows that the return rate is content-dependent to some extent. There are clearly programming decisions that influence users to return at a higher rate than other channels. Resonance between the ad and the content is also an influential factor.

Conclusion

Despite the huge spending on both linear and CTV, the data above proves that ads can disrupt many viewers’ viewing experiences. For publishers, this is an important problem to solve to maintain viewership — and monetize your channels efficiently.

There must be a better way to create less distributive ad breaks.

In the final part of this two-part series, we will examine Wurl’s latest product, BrandDiscovery, and determine whether GenAI can provide users with relevant ads they enjoy watching.