Yuval Fisher - December 1st 2021
Victim blaming is really not nice.
Don’t ever do that.
Except in this case.
Because, it can happen that ad holes – unfilled ad break time – are a consequence of something completely under the control of the content owner.
On Wurl’s platform, three fifths of ads are 30 seconds long and two fifths are 15 seconds long, almost exactly. In fact, ads with other durations almost never occur. This means that scheduling an ad break that’s 20 seconds long will result, at best, in a 15-second ad and a 5-second ad hole. How ad breaks are scheduled is important; poor scheduling can lead to impossible-to-fill ad breaks, dreaded ad holes.
“Surely this doesn’t happen,” you exclaim in shocked disbelief. “Oh, the humanity!”
Well, let’s explore. We can measure the “left-over” time that is not fillable by 15- and 30- second ads. If we add up all that residual time over a day and divide by the total scheduled ad break time, we can get a sense of the ad-break-duration-inefficiency of a channel: what percentage of their scheduled ad breaks will not be fillable (by 15- and 30- second ads).
It turns out that it’s not that bad. Most channels have less than 1% inefficiency. But a few have significant inefficiencies (as seen below for a cross section of Wurl’s channels).
For a 15-second inefficiency, the bulk of channels are below 2%, but surprisingly, for a 30-second inefficiency one channel has a 20% inefficiency rate, while a handful are above 5%. Oy.
Note, however, that this is not the real inefficiency; it’s the scheduled inefficiency. If badly-sized ad breaks are scheduled at prime time, they’ll contribute disproportionately to the inefficiency, making it much worse. Vice versa, break durations scheduled at off viewing hours will have less impact.
Are there other ways we can blame our woes on content schedulers?
Of course! It’s plausible that long ad breaks might be harder to fill. For one thing, demand sources aren’t set up to respond with very many ads, and secondly, as more ads are placed in a break, the processing takes longer and the chance of a time-out increases.
If you schedule ad breaks for a living, put down your phone (your Instagram friends can find out what you had for lunch later) and pay attention.
The plots below show data for two Wurl-delivered platforms at different times. Looking at data for January 2021, we see that the services had worse fill when ad breaks were longer. It’s important to understand that the data below is not representative of all channels and all channel sources. Different inventory portions are managed differently, with advertising teams that have different levels of sophistication and different levels of performance. These channels and fill rates were chosen to remove that variability. The goal is to see whether a uniform process for filling ads has a dependency on break duration. And it does, at least for these services.
Early in the year, budgets and campaigns are being sorted out, and fill is not that great. Looking at the same plot for October 2021, we see a different picture: fill is much better all around, and for one service, the fill rate is, on average, independent of the average break duration. Even long breaks are well filled. (One differentiator for the best performing service is that it utilizes Wurl AdPool Plus, a service which brings additional fill to the available ad break time.)
So should schedules be set up with the shortest ad breaks possible? Not necessarily. If we look at just the ad breaks of duration for approximately two minutes or less, we see that when long breaks are avoided, break duration doesn’t really affect fill. This is actually a big deal. Scheduling shorter breaks means they must be placed more often, which may have side effects.
So what are the take-aways here?
Is careful scheduling of ad breaks important for other reasons?
Let’s consider some obvious follow-up questions. For example: how does ad break duration relate to user churn or channel engagement? While these questions are not about ad holes, we’ll take a peek at the answers while we’re in the data neighborhood.
First, because churn is a delicate topic, we’ve normalized the churn graphed below: the worst churn is scaled to have value 1, and the other churns are scaled proportionately. No one wants to hang out their dirty churn.
More importantly, what is this churn of which we speak? In this case, we looked at the proportion of the users watching a channel in September 2021 who did not watch the channel in October 2021. They churned away.
The plot below shows that average ad break duration doesn’t really contribute to churn. This is really important! How ads are scheduled (within reason) will not drive users away from your channel!
In fact, the ad load – the total number of minutes per hour – also doesn’t contribute to churn. People won’t leave a channel if there are more ads. Or rather, this doesn’t seem to be a strong contributing factor to churn. Content is king!
What about sessions? Stay tuned and find out. The results are surprising!