The last blog entry introduced the spot TuneAway metric and showed how it varies by month, along with some prior research on the topic.  Now, we can use the metric and other variables to answer the question of optimal spot position within a pod.  Marketers seek to maximize the return on investment of their promotional campaigns, and since viewership varies by spot position, the specific placement of an ad within a pod can have a meaningful impact on ROI.

A “pod” is essentially a commercial break, a break in the middle of the scheduled programming.  Pods contain “spots,” or advertisements, which can vary in length but tend to be 30 seconds long.  Pods vary in the number of spots that they contain.  A commercial break during a 20-second timeout of a close basketball game will probably contain only one spot, whereas a pod between two movies on a movie channel could have 10 or more spots.  As a point of reference, we can look at the distribution of pod lengths in the original dataset, which looks at major broadcast channels for the last four months of 2009:

In this dataset, the average pod had 7.5 spots, while 85% of pods had between 4 and 11 spots.  Spot length is not considered in this graph; an interesting point for future study is whether average spot length changes depending on the number of spots in a pod.

Next we can investigate the best position in a pod.  Conventional wisdom says that the first position is the best, since viewers will tend to watch at least some of it before changing the channel.  Does the data tell the same story?  Here is the graph for TuneAway by pod position, using the same dataset as before:

The first spot in a pod has by far the highest TuneAway.  Conceptually, this finding makes sense.  Active viewers who are most likely to change the channel will do so during the first spot in a pod, which raises the TuneAway value for the first spot.  These active viewers are no longer on the channel for subsequent spots in the pod, so those spots have lower TuneAway values.

Does the high first-spot TuneAway value mean that the conventional wisdom is wrong?  Are later spots actually more valuable than the first one?  TuneAway is an interesting metric, but what marketers really care about is how many people actually watched the advertisement, or how much of the advertisement was actually watched.  In fact, what we find is that regardless of pod size, the first spot tends to have the highest TuneAway but also the highest viewership, as demonstrated here, with pod size held constant at eight:

The first spot gets watched more than any other spot, even though it also has the highest TuneAway.  This conclusion may seem counterintuitive, but it makes sense when we remember that the TuneAway metric does not have a concept of volume.  In other words, there tend to be more people watching the first spot in a pod than later spots in the pod, so even though the first spot has the highest TuneAway, it can still have the highest viewership, as the graphs show.  Assuming constant cost across spots, marketers seeking to maximize ROI should try to place their ads first within their respective pods.

An upcoming entry will look at TuneAway from other angles, including how the metric varies by day and time.

Data from Kantar Media’s DirecTView product.

Not all methods of bypassing ads can be easily measured.  We wouldn’t be able to tell when someone is sending a text message rather than watching a commercial without using extremely invasive measurement techniques.  Thanks to set-top box (STB) data, however, we can easily measure channel-changing during commercials, behavior that hurts both campaigns and the carrier channels, since there is no guarantee that viewers will return once the commercial break ends.

Kantar Media’s InfoSys data product contains a “TuneAway” metric that measures loss of audience during advertisements.  Specifically, TuneAway is the percentage of seconds not watched of an advertisement, compared to the number of seconds that could have been watched based on viewership one second before the ad aired.  For example, if exactly one STB is watching a network, and that STB stays tuned in for the full duration of a given commercial, TuneAway for that commercial would be 0%.  If a network airs a 30-second commercial and the STB tunes away after 27 seconds, TuneAway would be 10%.

The TuneAway metric has no concept of quantity or total viewership, so we cannot look at raw TuneAway numbers and make conclusions about the number of people who remained to watch the ad.  Nevertheless, TuneAway is a useful metric for comparing one advertisement to another, or for aggregating to look at larger trends.

Various interesting studies have looked at audience retention during advertisements to try to understand viewing habits and ultimately improve advertising yield.  One study looked at viewing behavior during targeted advertisements versus non-targeted advertisements, and found that viewers were 32% less likely to change the channel during targeted ads than during other ads.  Both channels and marketers benefit from this finding.  Google TV has also done some relevant work in this area, including an analysis that looks at historical reactions to advertisements and tries to use that information to predict future audience behavior.

TuneAway data can clearly be examined in a number of different ways.  For now, we can start by looking at TuneAway by month.  The graph below is based on a dataset containing all television advertisements (both promos and commercials) on five leading broadcast channels from September through December 2009:

On average, TuneAway tends to be a little under 2%.  Of the last four months, October had the highest TuneAway at 1.95%, while December had the lowest with 1.75%.  Individual advertisements can have TuneAway rates of 10% or higher, but on average the values tend to hover around 2%.

Lower TuneAway in December could be attributed to a number of factors.  Perhaps viewers had holiday shopping on their minds and decided to pay more attention to commercials.  Advertisements also might have been more interesting or better targeted in December compared to other months.  The subject requires further analysis.

Forthcoming blog entries will analyze TuneAway from a number of other angles.  One interesting problem to address is TuneAway by spot position in pod, and whether high TuneAway numbers are always undesirable.

Data from Kantar Media’s InfoSys product using DIRECTV data.

Programs with high ratings can command high advertising fees, since large numbers of people will be available to watch advertisements during the show.  The ultimate value of a marginal viewer to a program depends on the number of ads that the new viewer will watch while tuned in to the program.  In other words, new viewers mean more people to watch ads, which raises the price that advertisers can charge for future spots.

In this context, we can analyze promotional campaigns based on the number of ad impressions and the seconds of ads watched per promo spot.  This “return on investment” or ROI of a promotional campaign can be useful in determining which campaigns are the most effective and efficient, and how to better structure campaigns in the future.  Set-top box data proves to be an essential component of these evaluations.

To investigate these ideas, we selected four new broadcast programs on several different broadcast channels.  The series all premiered in September 2009 and had promotional campaigns that began in late May or early June 2009.  We can start by examining the number of promotions in each of these campaigns, starting from the beginning of the campaign and ending with the night of the series premiere:

Non-paid promos are promos run on-channel or on sister channels, while paid promos are run on competing channels.  Each of these campaigns was significant in size, featuring over 1,500 non-paid promos in all cases.  The campaign for program #1 used almost no paid promos, while the other campaigns each used around 500 or 600.  Since these are new programs, we can assume that the promotions are mostly responsible for driving viewers to the series premiere, rather than, say, loyalty to or awareness of the program from a previous season.

The next step is quantifying the return on investment generated by each of these campaigns.  Clearly, the number of ad impressions matters a lot—marketers want viewers to be exposed to as many ads as possible.  There are other variables to consider, though, such as total seconds of advertisements watched. 

In terms of time frames, there are various ways that the data could be sliced.  Do we only give a pre-premiere promotional campaign credit for ads viewed during the series premiere of the program?  Can we take some amount of credit for ads viewed later that night on the same network, or on future episodes of the program?  Finally, do we take credit for everyone who tunes into the series premiere, or only people who tuned in after first seeing a promotion? 

There are many ways to consider the data, but for now we can start with two metrics: Total ad impressions per promo spot, and total seconds of advertisements viewed per promo spot.  We’ll look only at the premiere night, but take credit for ads viewed both during and after our programs on the relevant networks.

We will examine only people who saw at least one promotion and then tuned into the series premiere; it doesn’t make sense to take credit for viewers who made it to the premiere without having seen a promo.

The following two graphs show the initial ROI results for the four programs:

Ad time viewed is divided between promotions and commercials, since marketers will likely be interested not only in the number of advertisements watched or the time spent watching advertisements, but also in the relative split between ads for, say, restaurants and basketball shoes compared to promotions for future programs.  The graphs include ad time for people who saw at least one promotion before the series premiere and then tuned into at least six minutes of the premiere.

The first graph demonstrates that the promo campaign for Program #1 was the most efficient, in the sense that it yielded the highest number of ad impressions per promo spot.  The majority of these ad impressions were commercials as opposed to promos.  The second graph looks at seconds of commercials viewed, rather than impressions.  Not surprisingly, there seems to be a very strong correlation between impressions per promo spot and total seconds per promo spot.  If we did happen to find a campaign that yielded relatively high impressions but relatively low total seconds viewed, there would be two possible explanations: Viewers of that program were much more likely than other viewers to tune away from advertisements, or the ads aired during that program just tended to be shorter in duration than the ads on the other programs.

Overall, the promo campaign for Program #1 seems to have been the most efficient, while the campaign for Program #3 was the least efficient.  The campaign for Program #3 had far more promos than any of the other campaigns, which led to the low efficiency.  There is certainly a difference between efficiency and effectiveness, but the two items do tend to be related.

ROI is a good way to measure the effectiveness of a promotional campaign.  There are many different decisions that need to be made in such a calculation, but the bottom line is that set-top box data can be used to compare different promotional campaigns and in doing so, determine the best ways to promote a program.

Data based on KantarMedia’s InfoSys data product using DIRECTV data.

Attentiveness and loyalty to a program are positively correlated, as explained in an early blog post.  Viewers who enjoy a show will watch a large number of minutes in a given episode and then return for future episodes, while viewers who do not enjoy the show will exhibit low attentiveness when they are watching it and will likely not return for additional episodes.

One interesting consequence of the relationship between attentiveness and loyalty is that the future viewing behavior of an important segment of a show’s viewers can be predicted by simply analyzing data from a single episode, such as the series premiere.  Viewers of the series premiere will return for a predictable number of future episodes based on their attentiveness to the premiere.  This information could be used by advertising sales teams looking to sell sponsorships based on expected future ratings.

To further investigate these ideas, we selected several new programs on broadcast networks and divided viewers of the series premieres into groups based on the number of minutes of the premiere that they watched.  For each of these series premiere attentiveness groups, we looked at the number of additional episodes watched during the season.  The results support the original hypothesis of a positive correlation between attentiveness and program loyalty.  The results were fairly constant across programs, and are shown for one program in the graph below:

Viewers who watch only a small portion of the series premiere are far more likely than other series premiere viewers to tune in to zero additional episodes during the season.  Viewers who watch a large percentage of the series premiere, on the other hand, are much more likely than other series premiere viewers to return for a large number of additional episodes.  The graphs below combine data from the several programs to summarize these observations:

One key variable to consider is the relative distribution of low, medium, and high attentiveness viewers within a series premiere or other episode.  In the programs analyzed, viewers qualified for the highest attentiveness category if they watched 42-60 minutes of the hour-long premiere.  On average around 55% of series premiere viewers were placed in this category.  If a certain series premiere yields a high attentiveness group that is significantly larger or smaller than the 55% baseline that we have found, marketers and advertising sales teams might choose to act accordingly.  A very large high attentiveness group, for example, would suggest that viewers will be unusually loyal to the program, which might mean that fewer promotions need to be run.

What is still unclear is whether marketing should be targeted toward high-attentiveness viewers or low-attentiveness viewers.  Are the high-attentiveness viewers going to tune in to future episodes regardless of promotion?  If so, perhaps the low-attentiveness viewers would be a better target.  Further study would help clarify these questions.

Data based on a nationally-projectable DIRECTV dataset from TNS Media Research’s InfoSys data product.