Navigating the Ad Tech Maze
In an ideal world, digital advertising should work like broadcast advertising, but with even better targeting and variable ad pod lengths. However, the ad tech business has the nuances of stock market trading, combined with the technical challenges to support all devices of the splintered multi-platform world that is streaming. And everyone involved knows there are gaps between where things are and where they should be.
What follows is a summary of the Streaming Media West Connect 2020 panel Streaming Ad Tech Relationship Advice, with our experts weighing in on their key issues. If you're not where our speakers are, you'll be making a lot of choices and will likely have to create your own path.
The panel consisted of industry leaders from a variety of segments of the OTT and advertising ecosystem:
- Larry Allen, general manager and VP of addressable enablement at Comcast
- Jeremy Brown, associate director of video delivery at Optus Sport
- Jarred Wilichinsky, SVP of global video monetization and operations at ViacomCBS
- Yazmin Wickham, senior director of digital platform management at Katz Networks
- Olga Kornienko, co-founder and COO of EZDRM
Allen set the stage: "The foundation of how we all trade and execute today, whether it's client side or server side, is based off of the IAB specs," he said. "The buy side, the sell side, and the technology that sits in between that is the common language that we all execute on regardless of how we actually do it." In other words, industry guidelines come from IAB, but all of the players in the ecosystem implement their own variation.
Viewers today want a lean-back experience in which content and ads arrive in a packaged bundle, with well-timed ad breaks (and with fewer ads than on broadcast) targeted to their interests. Media companies want the promise of getting a good price and selling out their inventory to media buyers. Brands expect accurate measurement for their ad buys—an audited mix of frequency and reach (so the same ad isn't being served to the same viewers over and over again)—and would prefer that all advertising on digital and broadcast use the same kind of measurement KPIs.
CSAI vs. SSAI
On the tech side, the discussion around client-side ad insertion (CSAI) or server-side ad insertion (SSAI) has been going on for a long time. Everyone on the panel primarily uses SSAI, except in some cases like connected television (CTV) and mobile, where it isn't possible. "It has to work like TV," said Wilichinsky. "If you have your choice, server-side, but there are plenty of scenarios where you can't do it and you've got to do client-side ad insertion."
SSAI has a number of benefits, including a better viewer experience, the ability to deliver a single video feed merging ad and content (which foils ad blockers), consistent video quality, fewer customized development requirements, and loudness control. "Server-side ad insertion is [that] you request your advertisements in the cloud, and you are stitching those responses back with your content," said Brown. "[In] client-side ad insertion, before a video actually starts, there's an SDK [software development kit] that sits on top of the player. It's typically Google's IMA [Interactive Media Ads] SDK, and you have to install it onto every single app. During playback, when we get to these timed breaks, the SDK will take over the content player and let its own player fetch ads and manage the impressions. It's almost like a player on top of the player, and they're fighting for your attention."
A server-side ad insertion workflow (Sketch by Optus Sport's Jeremy Brown)
"I think the distinction is where you track, and that's where you can really pivot and do things in a lot of different ways," said Wilichinsky. Some advertisers prefer specific measurement partners, and each measurement company's SDK has to be integrated and supported on all platforms. This can mean the ad will be delivered server side, but the tracking is done client side. "The more tracking you can bring to the client, the better," said Wilichinsky.
"In terms of tracking as it pertains to SSAI, when we can, we always push to have tracking fire from the client," added Wilichinsky. "It does alleviate ad vendor problems and difficult conversations on false positives. However, over the years, server-side tracking and issues have gotten much better. If all the same signals are available and if issues are not there, I do think serious conversations would be had moving to a full server-side tracking model."
That tracking, also called third-party validation, verifies that the ad is reaching the audience. "In the ad that gets served to you, there's a little bit of code that checks the environment that it's landing in," said Wickham. "Then it returns that back to the ad validation company, who then spits out a report for the advertisers to make sure that they are getting their money's worth." Up until now, there have been different types of measurements used within broadcast and digital delivery. This should hopefully change with the IAB Open Measurement SDK.
"Open Measurement is enabling the programmers and the advertisers to aggregate information across all of the places where their ads are running and develop metrics around reach, incremental reach, frequency of their advertising, [and] being able to deduplicate and understand where audiences are actually being exposed to ads," said Allen. "It's really an important component for how the ecosystem is going to evolve and be managed." What this means is that instead of each measurement vendor requiring different code integration, the IAB Open Measurement SDK provides common code for validation vendors to gain access to reporting data.
An important requirement for viewers to have a great experience is frame accuracy. "In a live linear environment, it's very important when an ad is getting inserted into the stream for it to be frame-accurate so that you don't get frames of black or overlaying or clipping of your ads in the pod," said Allen.
The identification for when an ad pod starts is the SCTE marker, a timed piece of metadata that gets added to your stream, indicating a timestamp and a duration. "The duration tells you how long that ad window is going to be," said Wickham.
"Traditionally, people are using the SCTE triggers to delineate between the national inventory and the local or distributor inventory that's often shared in carriage agreements," said Allen. "Even getting the SCTE trigger out of your broadcast center is a challenge upon itself," added Wilichinsky. Does it contain the right metadata? Is the SCTE trigger on a national ad? Is it being used for dynamic ad insertion? Is it a local ad for an affiliate? Is it a local spot for an MVPD? "All these people rely on this one single trigger that goes out. Changing it is really like flapping your wings in the Sahara and causing a hurricane," said Wilichinsky.
An ad tag will call for either a VAST (Video Ad Serving Template) or VPAID (Video Player Ad Interface Definition) creative. VAST is the XML format standard, which works as a set of instructions so the player can understand what formats of the video are available to pick from and when to trigger URLs to track impressions, error codes, and measurement, according to Brown.
"[VPAID] allowed us to interrupt the player and put an interactive ad over the player. That worked really well in a website, but obviously, over time, running other people's code over your player [presents] a security issue," said Brown. New privacy requirements are also not compliant with this approach. "We are still seeing the VPAID used to sideload SDKs for measurability like viewability," said Brown.
"What we're finding with TVs using a native player is you don't get that ability," said Brown. Trying to get the same experience everywhere isn't possible yet, which is why there still can be buffering on old devices loading ads.
To DRM or Not?
The panel discussed the idea that DRM would solve ad fraud. DRM can work on SSAI or CSAI. "[DRM will provide] reporting and [keep] track of whether or not an ad actually played, because you can't crop it out," said Kornienko. "It would be beneficial to the ad provider to know that nobody took a stream, and … put somebody else's ads in." Pirates can steal and distribute live streams, substituting their own advertising so that they can monetize them.
"DRM actually does not cause latency in any player, in any environment," said Kornienko. "This is especially vital to live [content] and would not cause any latency at all, because the players should be able to handle both downloading of the license and downloading of the content at the same time."
Wilichinsky said that problems arise when content is DRM-protected but ad segments come unprotected from a different CDN, causing crashes as the switch is made between the two. Brown added that the players on many connected TVs don't support that switching.
A workflow demonstrating how DRM can be used with ad insertion. Contrary to a common misperception, DRM will not introduce any latency, says EZDRM's Olga Kornienko.
Life Without Cookies
"How does life look … without cookies? It's very bleak," said Wickham. "I guess therein lies the rub? You have to ask for the personally identifiable information [PII] in order to be able to really track somebody and serve them appropriate ads across devices. So if you can't do that, you're just shooting an arrow in the dark."
A cookie is a direct request from a client to a service at any point in the ad service ecosystem. "When you put SSAI in the middle, you've created almost a firewall between that transaction," said Brown. How can privacy policies and platform changes still allow personalized ads under the General Data Protection Regulation (GDPR), the California Consumer Privacy Act (CCPA), and iOS14? Platforms require users to log in, then target against content and the non-PII detail sourced from user registration steps. "The consensus idea seems to be that if you can get somebody to give you permission to drop that advertising tracking cookie, or provide an email address, then you can anonymize this and use that with your ad tracking across everywhere ... they are consuming content," said Wickham.
"I think it's why you're seeing this rush for almost all the media companies to move B2C, where they're building a direct relationship with consumers, so then they have an ability to do that linkage and build a household graph across their footprint and partner with other platforms that also have direct consumer relationships," said Allen.
"We have 25 million subscribers, and we have the ability to match to those households, across device types, in our set-top box environment, but also within the apps that run on our platforms and apps that run on other people's platforms," said Allen. "That is a really important thing, and I think media companies are all actively building their own brass and trying to drive more consumer relationships where they're capturing email, home address, credit card, etcetera."
"The majority of our viewership [is] always going to try to watch it on the best screen available. When you're home, that's on a 55-plus-inch TV that's either hooked up to an X1 box or Roku, or running tvOS—[there are] no cookies," said Wilichinsky. "On your phone, on an app, [there are] no cookies, so that's a cookie-less environment."
Device manufacturers have created re-certifiable identifier technologies to provide compliance to GDPR and other privacy laws. "It comes back to the resettable, privacy-safe device IDs that are available," said Wilichinsky.
"The logged-in user, the subscriber, is the next iteration of identity," said Wilichinsky. "Those profiles will become more important in the coming years as we get more relevant advertising to who we believe is watching the content."
Standardizing Digital Ads
With thousands of creatives flying around, figuring out if the ad meets an acceptable quality standard would go a long way. "If we could all just get the same asset that goes to the TV ecosystem, a lot of our problems would go away," said Wilichinsky. "This is where VAST 4.0 made the first attempt to make the mezzanine file as part of the spec and bring standards around metadata about the advertisement [Ad-ID in the U.S., Clock ID in the U.K.]."
"In streaming and in live linear, there's a whole policy process around what ads can run," said Allen. This competitive separation is the norm in broadcast, as is a standard creative distribution process that hits all of the quality standards. "That's a workflow and a process that certainly needs to be hardened now that streaming and the number of devices [are] scaling from a volume standpoint," said Allen.
Another issue is that multiple platforms have different requirements on pod durations. "Getting streams onto new platforms like Peacock and the Roku channel, they are recommending that you comply with their ad pod duration [target] length," said Wickham. "Peacock has a 5-minute limit per hour."
"If you're trying to take a linear broadcast into their digital platform, you're probably running more than 5 minutes per hour. So then the issue becomes, well, what do you cut off? Do you cut off some of your content? Do you cut off some of the ads that you promised you would serve?" said Wickham.
A third issue is data used for targeting. "Your content could run inside of YouTube TV, but if YouTube doesn't grant you rights to the user data on their platform, then you're limited to how much you're going to be able to leverage that inventory and sell it," said Allen. "That ultimately could lead to a bad user experience. … Either it goes black slate because you didn't sell it at all, or you're selling it dirt cheap, and then that creates downward pricing pressure on inventory elsewhere, potentially."
A fourth issue is actual ad formatting, as advertisers tend to focus on CPMs rather than a quality experience for the viewer. "These things are playing on 70-inch TV screens, so getting a 4-by-3 letterbox asset that got encoded 10 times before it ever makes it into a digital tag just looks like amateur hour," Wilichinsky said. "We need to be better than that."
There is a lot of work being done to improve the digital video ad experience, and three areas are key to the future.
There are efforts to standardize the use of SCTE triggers, because not all data is consistently provided. "As we're moving into addressability and we're working to consistently enable national inventory to also be insertable within these local systems or digital CTV systems, there are new SCTE triggers that are being developed," said Allen. The result will be better personalization of ads to match content, viewer demographics, and replacement ads for both digital and linear.
The privacy-compliant future means navigating a world of three competing potential data owners: the programmer, the distributor, and now the device manufacturer that controls the resettable identifier. "It's open from a technology [position, but] the business ecosystem of TV is the most complicated spider web we'll ever encounter," said Wilichinsky. "You have three layers of owning the user. How [do] we all work together at the end of the day to provide the best user experience, because identity might be three layers above you as a programmer, or you might have it and they don't have it."
Figuring out the best approach both from an ad stack and ad delivery perspective remains a challenge. "This is not getting any simpler," said Allen. "All these things need to work together so you can decision across them in a unified manner. … I think about [interoperability] a lot and talk about it with the team every day. I'm making sure that the programmers and the advertisers have a scaled opportunity across all of these devices. There is a lot of fragmentation, and there are still these kinds of business impediments based on how you've negotiated with a platform or how some of the platforms are holding their data back."
These experts have managed to navigate their specific ad tech requirements, but keeping up with the Joneses is getting more and more complicated.
It's time to move beyond our outdated, Web 1.0 approach to advertising and embrace stateful ad podding. The future of connected television depends on it.
As ad-supported OTT services grow, the video engineering world needs to understand the complicated advertising technology landscape.
Companies and Suppliers Mentioned