Context should be Scene and Not Referred

With privacy initiatives gaining more public attention and converting into legal requirements, targeted advertising is shifting from demographic, identity-based audiences to contextual audience segments. By aligning ads with the content rather than targeting the viewer, advertisers bypass privacy headaches while still delivering relevant ad experiences. As an example, sports drink advertisers that might have used third party data providers to target the homes of people that bought a competitor’s sports drink last month would have to forego that slightly intrusive (and possibly creepy, if you convert “sports drink” to something more private) targeting method and derive a similar benefit from targeting sports shows.

In streaming contextual targeting can be based on various types of data: 

• App or Channel-Level: This is the broad-strokes approach. As an example, advertisers of sports cars, place ads on an automotive FAST channel. 
• Content or Show-Level: A step deeper. Now those sports car ads are placed specifically during a documentary about Formula 1, for example. 
• Scene-Level: Best of all, scene-level targeting pinpoints the exact moment in a show to place an ad. For example, the sports car ad plays exactly after the protagonist of a movie triumphs in a high-speed chase. 

News: An example of where scene-level targeting matters

For some types of context, e.g. brand safety, using coarse contexts like channel or show can leave ad opportunities unfilled. News, for example, consists of both typically safe segments (think “happy stories about the local children’s baseball team”) and unsafe segments (think “war” or polarizing political stories). An advertiser can avoid News completely and miss good opportunities, or take the risk of their ad being placed next to content they don’t want their brand associated with.

Figure 1. A six hour period (along the x axis) of a news channel scene analysis, showing brand safety categories, with red being unsafe. A significant amount of the time is actually safe in all brand safety categories. 

Unfortunately for advertisers, even the first two types of context (Channel or Show-Level),  are tough to come by, irrespective of what contextual targeting offerings imply.  Publishers, who typically have both high-value content and “less-high-value content,” *cough*, like to bundle everything together. They don’t like to share channel or show-level information to avoid cherry picking of their best content, which could leave the rest poorly monetized. 

Genres: Contextual targeting’s mirage of false hope

What tends to happen is that show- or channel-level content is converted to a genre, which advertisers commonly use for targeting. This approach falls somewhere between “highly inefficient” and a “total waste of time.” For example, about 45% of FAST requests have either no channel name or a uselessly obfuscated channel name. Additionally, 89% of FAST requests do not have a meaningful program name. And while 98% of requests have genre information, 15% is generally useless information such as “entertainment” or “drama” or “tv,” while 35% carried a declared genre that did not align with what Wurl detected in the actual program. To read more about these findings, check out our recent report, “Addressing the Content Signal Gap.”

Unfortunately, even safe-seeming genres, say Sports, can lead to unintended consequences, where content discusses non-brand-safe, non-suitable topics. Check out the chart below for an illustration of this challenge.

Brand safety on a Sports channel, showing that even presumably safe channels may have unsafe segments.

Attentive, astute readers know that context goes beyond brand-safety. Context can be granular (think “watches”), driven by emotional resonance, or theme (think “gifts” or “holiday gifts”). The more granular the context, the less available coarse-grained inventory there is, because there isn’t a “gift” genre. As advertisers seek ever better audiences, they need more granular exposure to contextual segments – genre targeting just doesn’t meet what the market needs.

Even channel or program–level genre data leaves a lot of opportunity on the table. For example, take a look at the chart below. It represents a 6 hour period in a single News channel with “News” genre. The missed value of the contextual signals on the channel is obvious:

Contextual Categories on a News channel, showing that context categories provide many more advertising opportunities compared to a single “news” genre. Categories not shown in this window occurred elsewhere that day. 

Doing Contextual Advertising Right

So what’s the solution for better timed ads? Scene-Level targeting. 

With Scene-Level targeting, advertisers can precisely target audiences based on the content that is being shown on the screen right before the ad break. We’ve written a lot about this in the past, but in summary, when ads and content align, viewers are more engaged, ads feel more relevant, campaign outcomes improve, and viewers are less likely to tune away. 

If you want to leverage scene-level data, you can, but shouldn’t, just partner with anyone. Analyzing individual scenes across millions of hours of video requires serious technical muscle. Select a partner that checks three major boxes:

1. They know how to analyze scenes at scale. Scene-level data is useless without a massive footprint. You want a partner who does a massive amount of stream distribution and actually has the AI and machine-learning chops to evaluate the emotional, genre, and brand safety contexts of individual scenes.
2. They probably already deliver your stream. Why overcomplicate your tech stack with unnecessary integrations? The ideal partner is likely the one already handling FAST distribution, scheduling, and server-side ad insertion (SSAI). When the distribution and contextual analysis happen under one roof, things work better.
3. They play nice with others. A great partner doesn’t just build walled gardens. They should be able to serve as a platform for third-party contextual partners who can contribute their context to the stream. A contextual data partner offers you their own context. A contextual data platform offers you their context and the context capabilities of their partners. 

The Bottom Line

Virtually everyone claims to do contextual targeting these days. But slapping a broad channel-level genre filter on an ad buy doesn’t give you the granular, emotion-driven results that scene-level analysis does. Find a partner that understands the nuances of the video asset, has the distribution footprint to make it count, and gives you the exact tools you need to put your ads in the perfect context.

Footnotes:
¹ Wurl has access to roughly 75% of FAST inventory, from which this data is generated.
 ² In this case, the segment was about former boston red sox star David Ortiz, who was ambushed and shot in the back, yikes!