Programmatic Buying: What it Means, How It's Changing Advertising

The glamour of the Mad Men advertising world starkly contrasts with how online advertising is now bought and sold, since much of it happens programmatically. One expert says that without programmatic assistance, it would be physically impossible to place 40 insertion orders for a campaign to appear everywhere a client desires.

"Programmatic in many ways is leveraging real market realities; it's data, upping your bid, lowering your bid, and really working more in an environment of near real time," says Noah Fenn, head of video sales and strategy for AOL. "In some ways, it's like trading stocks. In quick succession you are trying to make sure you get your marketing message across and that you're scaling and reaching an audience."

Although subscription video on demand (SVOD) and transactional video on demand (TVOD) have been successful, ad-supported content isn't going away anytime soon, no matter the viewing device.

"Despite what some digerati might tell you, consumers regularly accept advertising in exchange for free stuff. The ad vertical is very healthy, and for Roku it's one of our fastest-growing segments," says Scott Rosenberg, vice president of advertising for Roku. "As the OTT ecosystem has matured, the ad-supported segments—apps and channels that are free—have grown tremendously."

Previously, advertising was bought and sold through direct buys between media companies and brands. In contrast, eMarketer reports that $2.91 billion was spent on digital video programmatic buying in 2015. The company forecasts an increase to $5.37 billion in 2016. Almost all online display advertising is traded programmatically. Currently, 39 percent of video ads are transacted this way, which leaves a lot of room for growth in this area. The good news is that 9 out of 10 TV ad buyers are shifting an average of 10 percent of their budgets away from linear TV to digital, according to AOL's 2015 "State of the Video Industry" report.

In the ideal programmatic transaction, a user clicks on a website, and her internet address and browsing history are packaged and whisked off to an auction site. On behalf of advertisers, software scrutinizes her profile (or an anonymized version of it) and determines whether to bid for the right to place an ad next to the media she is about to view. If you're looking for affluent women between 30 and 35 who own houses and dogs in a specific ZIP code in Dallas, you may hit a premium price. If you take a broader view—say, all viewers between 30 and 35—your pricing may go down, and the supply of viewers could go up substantially.

Big Data

Today, brands and media companies have deep knowledge about what type of audience they are reaching, based on all the discrete data users leave as they navigate around the internet. This data helps programmatic advertising thrive.

Programmatic buying can spread ad buys over multiple platforms and content types to reach a targeted demographic. Online advertising benefits advertisers by allowing them to reach a specific segment of viewers, engage directly with these viewers, and measure their responses. In the OTT world, advertisers use contextual targeting based on a range of data to identify their best possible audience.

Publishers that require subscriptions have a lot of information about their viewers. Brands may own even more data. "P&G probably owns a lot more data about its customers and its target audience than ABC does," says Mike Shehan, co-founder and CEO of SpotX.

That data can provide insight about viewer activity, ranging from basic demographics to complex psychographic information. The audience's clickstream is big business. Advertisers know a viewer's age, gender, location, browsing history, shopping habits, registration information, third-party cookie information, trending keywords, purchase history, interests, device information, and the size and location of the player within the frame.

Merging first- and third-party data with behavioral and contextual information and adding in machine learning based on predictive modeling yields in-depth information about the audience. This is something TV ad buyers can only dream about.

Brands use their data to decide whether they want to buy an impression on a case-by-case basis, in real time. If one parameter isn't working well, they can turn off that buy and move on to another segment.

"We allow you to determine on each individual basis how much you want to pay for a particular impression," says James Shears, general manager of advanced TV at The Trade Desk. "Essentially, you put in metrics you were looking to hit, you find your inventory services, and the system runs algorithms to determine what type of mix from a media perspective actually would benefit you the most, measuring against those metrics."

Both buyers and sellers are able to track the entire life of the transaction and transparently see all details associated with a campaign. Previously, this type of information was held in secret with a middleman closely protecting many of the transaction details.

It's known that advertising is big business, but the size of the advertising technology industry may come as a surprise. There are more than 2,000 companies in the ad tech space. To keep business transparent on a basic level, publishers list their supply of inventory available with third-party supply-side platforms (SSPs). Brand advertisers use another third-party service called a demand-side platform (DSP) to set the buying parameters and monitor the performance of a programmatic campaign.

"If I'm a website [and] I have cooking videos and someone comes to my site and I have a pre-roll ad slot, as the page is loading, an ad call will go out from my SSP. It will send out a signal to all of those DSPs and say, ‘We have an ad spot, do you want to buy it?' They pass along as much information as they can," says Brian Rifkin, co-founder of JW Player. "Then a bidding war starts, and those advertisers signal and say they're interested and they want to pay X. They'll send back what they're willing to pay and the advertising creative. The SSP will decide in real time who the winner is; it's usually based on price [but] there may be other factors. That advertising will be served on to the website. All that happens quicker than I can snap my fingers if it's done correctly.

"I'm a big fan of, ‘Pick your type [of] technology and stick with it,'" says Rifkin. "A lot of people bounce back and forth all the time (to SSPs), but if you think about it, the same buyers are on all of these."

There are a number of flavors of programmatic advertising. In a private marketplace, a media company and one or more brands have a private auction open only to those invited to bid. Programmatic direct is a one-to-one interaction for reserved inventory at a fixed price. In an open marketplace, inventory is aggregated from multiple partners for buyers to bid to purchase impressions via an open auction.

"Programmatic, in particular the open marketplace, is often incorrectly linked with driving down CPMs [cost per thousand impressions]," says Shehan. "In the video market where supply is short, it can allow publishers to charge a pricing premium on a much larger chunk of inventory. In some regions around the globe, we're seeing some publishers charge as much as 25 percent more than rate cards to buy programmatically in a private marketplace."

Advertisers are looking for premium inventory. "It's inventory that is created by a well-known source, that is fraud-free and has a desirable audience," says Alanna Gombert, deputy general manager of the IAB Tech Lab and vice president of technology and ad operations of the Interactive Advertising Bureau (IAB)." Fraud-free to me means there's no mechanical device or ill intent within that webpage to mislead a consumer."

The biggest issues for brand buyers according to the AOL report is the inability to access premium inventory at scale, cited by 52 percent of the respondents. The next biggest concern is safety and quality, cited by 47 percent. More than half (58 percent) of the publishers said the perceived risk of commoditization of content is their biggest concern, followed closely by what they felt was a lack of existing processes and systems at 53 percent.

The IAB Tech Lab is hoping to address some of these concerns. It runs workshops where buyers and sellers can converse with their peers about best practices. "For the most part, programmatic is on the media plan now," Gombert says. "As a publisher, you're going to be sourcing from advertisers you would never speak to in a normal sales process because they're buying from all over. They're also trying out new things, so be open and aware that this is a channel where you can glean new buyers.

"There is a fear factor when something new comes out into the market, especially among traditional companies," says Gombert. "When a new technology comes out and it is widely adopted by the industry, you need to adopt it. You can't wait, because you're going to miss the budget if you're on the supply side. You're going to miss this inventory and client expectations if you're on the buy side."

Billions of ad transactions go through SSPs and DSPs every day. Here are a few technical pointers to make sure your ad doesn't result in the dreaded spinning icon.

(Ad) Tech Talk

"Implementing programmatic advertising involves connecting your video playback to an SSP such as SpotX, YuMe, or DoubleClick. These SSPs in turn host an ad exchange and manage the programmatic fill against various DSPs such as AppNexus, DoubleClick, and Bid Manager," says Michael Dale, vice president of engineering at Ellation.

"The most basic implementation involves associating an ad tag URL to a given playback slot such as pre-roll, midroll, and post-roll within your video player ad configuration," he says. "Then the SSP normally delivers a VPAID creative that manages the ad decisioning and tracking on the client side."

VPAID, or video player-ad interface definition, is a set of industry standards developed with the IAB for communications between the in-stream video ad and the video player, informing the video player about control information and reporting how the user interacts. If a player isn't programmed to accept VPAID ads, the ad cannot be executed, according to an IAB guideline. More often, however, there is a surplus of VPAID tags used by everyone involved.

Originally, VPAID was promoted to track user interactivity. But VPAID tags are being used for something different than what was originally intended. "Everyone in the chain is delivering their piece with VPAID. It goes through the whole stack with each company in the chain using it because it gives them the most information. The main consequence is the ad takes forever to deliver," says Rifkin. You'll see a VPAID, wrapped in a VPAID, wrapped in a VPAID—it slows the experience down for the end user and is often responsible for that blank screen while your video waits for an ad.

The future standard VAST 4.0 (video ad-serving template) should address some of this problem. "VAST is a video ad-serving template for structuring ad tags that serve ads to video players. Using an XML schema, VAST transfers important metadata about an ad from the ad server to a video player," according to IAB guidelines. Rifkin says he anticipates VAST 4.0 will address the VPAID loop problem.

Other Technical Issues

There are many scenarios in which ad delivery can fail, especially in the context of programmatic fulfillment, where it's difficult for an SSP to ensure quality for all of the DSP ad creatives it traffics.

"This is not always a matter of lack of attention to detail on the DSP side. There are changes in platforms, playback engines, and ad loaders that can create issues with ad creatives. This is even more apparent recently as the ad industry has finally started to shift over to HTML5-based creatives where things are less predictable then the relatively monolithic Flash platform," says Dale.

"Publishers should employ quality of service (QoS) tools that can provide additional data points around ad success and failure in addition to working closely with their SSP to track ad failures," he adds. Conviva and Youbora are two leaders in the space.

Measurement

There are a number of measurement issues to be aware of—the first is the ad breaks themselves. The OTT industry still hasn't settled on a de facto standard for signaling ad and program boundaries in HLS (HTTP live streaming) and DASH. Most dynamic ad insertion (DAI) providers still largely follow their own signaling specifications. Those who author content for OTT distribution expecting to be able to monetize midroll ads need to make sure their streams conform to their DAI provider's specification.

"SCTE-35 is a DPI (digital program insertion) standard which broadcasters and MVPDs use to signal ad break and program boundaries in MPEG transport streams," says Alex Zambelli, principal product manager at iStreamPlanet. "Anybody who is distributing live linear OTT content in North America is likely to run into SCTE-35 as the industry standard for ad signaling, so learning how to decode it and interpret it can be essential to monetizing live linear OTT streams.

"The encoding and packaging components have to ensure that the streams are properly conditioned for downstream ad insertion. That specifically means that the video encoder must insert IDR frames (key frames) into the video streams, and the packager must break up HLS segments at ad break boundaries," says Zambelli.

Another measurement area is viewer numbers. The traditional broadcast workflow might have Nielsen watermarks in the streams to identify who is watching the ads. "When you replace broadcast ads with digital-only ads, those digital-only ads might not have the Nielsen watermark. You might need to go to whoever is providing the server-side ad insertion service and have them give you a report on which ads were delivered. So it makes the reporting a little bit trickier because it's not all coming from one place," says Zambelli.

"One of the most important and first uses of data in our opinion was the ability to score that data against a Nielsen demographic. There is $75 billion of TV ad money that moves around which is exclusively measured by Nielsen ad ratings," says Rosenberg.

Roku has recently created a video advertising network. "We have something called the Roku ad framework," says Rosenberg. "It's an IAB VAST stack, but it also embeds a Nielsen digital ad rating solution. We are the only OTT platform in the ecosystem to do that. We use our registration data and a partnership with Nielsen to enable Nielsen to measure demography on Roku and embed two rich media interactive vendors, BrightLine and Innovid." In this way, a publisher gets credit toward the demographic campaign it sold to an advertiser.

Server- or Client-Side Ad Insertion?

The age-old debate continues. Defeating ad blockers was originally a key value proposition of server-side ad insertion (SSAI). Recently, this is less of a unique value proposition, as viewership has shifted toward living room and mobile devices, where ad blockers are less common. "Additionally, robust client side tools with server ad proxies have emerged to address the ad block issue while retaining client-side interactive code in ad responses," says Dale.

"It's difficult to say if there's an optimal approach," says Zambelli. "The choice between client-side and server-side ad insertion is often dictated by workflow requirements and largely depends on which part of the workflow is willing or ready to take on the burden of performing ad insertion. From the perspective of consumer applications, server-side ad insertion is definitely easier, but that doesn't mean it's guaranteed to be flawless.

"If the dynamic ad insertion is performed server-side (SSAI), applications need only limited ability to identify ad breaks for the purposes of constraining playback controls (or perform analytics). Applications performing client-side ad insertion, on the other hand, will need a very intimate understanding of how event boundaries are signaled in the media streams," says Zambelli.

In its 2015 year-end report, FreeWheel said live streams were the fastest growing content segment, with a 129 percent year-over-year increase in ad views. Therefore, getting the live stream ad serving working is one of the more pressing areas to develop.

"Server-side ad insertion is ideal for live broadcast where frame accurate ad timing is important to the quality of experience, or in platforms and syndication contexts where no client side code can be run. In contrast, any time you can run client side ad insertion, it's generally better to do so," says Dale. "Programmatic fill logic is generally optimized against client-side creatives that run client-side code to better track users, manage fulfillment, provide interactive features, and better measure engagement."

"The pros (of SSAI) are basically that it's one seamless experience to the user; it's not that slight delay that users are often frustrated by when there's loading," says Fenn. Obviously, no delay makes both viewers and publishers happy. "It's a challenge from a standpoint of the buyers that are traditionally pixeling and monitoring for viewability, because the commercial breaks for these pods are not as easy to define as an actual ad call."

Fraud

A number of people I spoke with said they considered viewabilty to be table stakes. "The definition of viewability the Media Rating Council uses in the video space is at least 50 percent of pixels in view for at least 2 seconds in an active tab," says Fenn. It's hard to imagine a video using autoplay not being counted as a view based on this measurement. Using this viewability guideline, it sounds as if video advertising measurement is taking a few liberties.

There's no polite way to put it: The online ad industry has had a history of fraud. In the Association of National Advertisers study, "The Bot Baseline: Fraud in Digital Advertising," digital advertising security firm White Ops studied 5.5 billion impressions. Almost one-quarter (23 percent) of video impressions, twice as much as display ad impressions, came from bot traffic. Video impressions are big-ticket items compared with display ads, so it's a natural target for illegal activity.

Bot traffic inflated audience monetization between 5 percent and 50 percent (depending on site measured) by pretending to be real users, moving mice over ads, putting items in shopping carts, and traveling around the web creating a phony digital trail. In short, bots make their traffic look like it comes from real people, whom publishers and advertisers covet. If you combine the viewability measurement and the bot traffic estimates, you could be hard-pressed to figure out an accurate measurement of your reach.

Looking to the Future

So how does the story continue? Individual programming, in which viewers will be able to program their own content experience, is on the horizon. Advertisers and publishers need to innovate to stay relevant in this new universe. The companies with the most accurate data will be ahead of the pack when it comes to having consumers respond to their ads. Data derived from contextually relevant user profiling puts brands and publishers in a much stronger position to create ads viewers want to watch.

Because so much is now known about viewers from their clickstreams, advertisers should have no excuse to display off-brand or repeated ads. If you know what a viewer does, what she buys, where she lives, what she watches, and what devices she watches on, that viewer will have a higher expectation of receiving a relevant ad experience. Any company not using its data to serve appropriate ads will have to fight hard to stay relevant and keep viewers from clicking to different content. If advertisers can't get this right, maybe we should root for the bots.

This article originally ran in the June 2016 issue of Streaming Media magazine as "And Now a Word From Our Sponsors."

Nadine Krefetz's article first appeared on OnlineVideo.net