The In Crowd: Leveraging Data and Tools to Meet Video Audience Guarantees

'It's all about minimizing waste'

“There’s so much waste,” a publisher says with a sigh.

We’re speaking about guaranteeing digital video against panel-based metrics, and one can’t help but shake their head at the amount publishers have to over-deliver to hit demographics. It’s tough to watch a grown ops professional bawl over their lost margins as a result of serving an abundance of free impressions.

For years, digital video publishers glared in envy as their TV cousins swam in buckets of ad spend. When, oh when, they cried, will those dollars shift over to our awesome, extremely flexible and measurable medium? Sure enough, advertiser interest picked up as consumer eyeballs switched over to digital – the latest report from Nielsen claims television viewership was down 4% last quarter while online video streaming was up 60%. But did digital video providers know that increased video ad spend would be tied to clunky and inefficient gross ratings points?

According to eMarketer, US digital video ad spend will hit $7.7 billion in 2015. Nothing to sneeze at – unless you consider that US TV ad spend is around $70 billion. However, digital video ad spend will continue to rise at double-digit rates for the foreseeable future – the 2018 estimate lands at $12.7 billion. TV spend is still estimated at a mighty $78.6 billion, but the growth curve is around 3.5%

Engaging Audiences on Multiple Screens with YuMe’s Video-First Innovation

Much of the credit for the meteoric rise in video ad spend goes to the introduction of panel-based metrics (ahem, Nielsen OCR and comScore VCE), the language of TV buyers. And though GRPs may seem quaint and old-fashioned, advertisers love them – they can be easily compared with TV metrics, and in turn brands connect GRP data to product sales and other tangible metrics. Connecting other online metrics offline is an ongoing slog while digital attribution tools have lackluster reputations – GRPs have been trusted for decades.

Unfortunately for digital video providers, wooing TV-esque spend via demo-talk requires audience data (age and gender), when many publishers have thrived on first-party behavioral segments. In a recent YuMe-Digiday survey, only 42% of responding publishers reported the ability to collect age and gender data. Moreover, the study showed that only 39% are currently guaranteeing against audience, and of that group 28% can always meet guarantees while 72% “usually” do. Even if a publisher collects demo data, it needs to be in-line with Nielsen or comScore’s view.

To garner bigger spends, publishers have been forced to heftily over-deliver on video campaigns to reach guaranteed demos, hence the waste my publisher friend was crying over. Publishers may “know” their audiences, but not necessarily in the way TV advertisers demand. However, a variety of methods have emerged to assist publishers in reaching target demographics – some based on algorithms and other more manual processes that attempt to match first-party data with demos.

“It’s all about the game of minimizing waste,” a publisher comments.

The Post-Impression Rub

News flash: OCR/VCE are post-impression metrics, and advertisers are only charged for impressions that were verified in the demographic. It’s just like TV: basically at the end of the month or a campaign, publishers compare their numbers to Nielsen or comScore measurement data to see what percentage of campaigns hit the target demographic. For, if an advertiser buys a million impressions in a certain demo, the publisher may need to run 1.5 million to meet the goal.

If the composite numbers come in below the agreed-upon amount, the publisher must offer a makegood. If it’s above that amount, bully for the advertiser – they received bonus views at no added cost. They also probably received a great deal of additional reach that may be handy for analysis and future planning, but your TV peers will reluctantly inform you when it comes to buying, advertisers only really care about what’s in demo.

So publishers wisely raise CPMs to hedge against the wasted impressions, giving GRP-guaranteed buys a higher price point. But what was that about digital being the most measurable medium? Haven’t you guys been boasting about all this priceless data you collect? Why can’t you work that into better targeting and inventory management so we advertisers can pay “reasonable” CPMs?

It looks like the onus is on the publisher to slash the waste, with yield and ad ops leading the charge. Sadly, completely accurate pre-impression or real-time decisioning based on demo is still a pipe dream, though a variety of service providers are offering tools that promise to lift GRP composition and cut waste.

Broadcasters also have a distinct advantage here – demo performance on the TV side can give them a baseline for online GRPs. Analyzing the offline metrics against strategically fired OCR/VCE beacons draws a decent picture of program performance from a comp perspective. Good old contextual targeting can come in quite handy here: advertisers can be shepherded to the programs most likely to match their demands.

Unsurprisingly, the data for online viewers skews younger, but this analysis is really only a starting part for meeting demo demands without major inventory shedding. Video publishers need to dive deeper into their data.

Data Does It

For guaranteeing against demos, user registration data can be a powerful asset. Of course, personal identification information is never used, but age and gender are confirmed, unlike in behavioral data where inferences are made based on tracking. One publisher suggested that employing registration data would be the only way to push GRP composition into 90%-range.

Subscription-based streaming services typically have that vaunted registration data, but many video publishers don’t require log-ins for viewing. To encourage additional data sharing, some of the pubs I spoke to are introducing incentives for users to log in, including content recommendations based on viewing history and the ability to remember playback locations across devices.

However, broadcasters are wary of log-in fatigue – to view a majority of digital TV content, users must have their cable or satellite subscription authenticated. That log-in and the data associated with it belongs to the MSO. Therefore, individual site log-ins would be on top of the authentication – this might prove annoying to users, but those with high brand loyalty may accept it.

The question is, will that be enough to get scale? Publishers may also have age and gender data in email newsletter and site commerce data (the shop), but these are even further limited in scale.

Also, while age and demo data are “confirmed” in registration data, it’s not guaranteed that Nielsen and comScore will agree in its measurements. As one pub told me, sometimes they don’t agree with each other – Nielsen and comScore data will say very different things about the same user. There is an opportunity here similar to viewability – insist advertisers use the provider most favorable to you. Also, because Nielsen is measured via Facebook data to verify their panel data, a user who is logged into a site through Facebook Connect should always register accurately.

So even if there’s a great deal of it, registration data cannot be relied upon alone. The next step would be finding the intersection of first-party behavioral data (typically on-site viewing) and demos. Many publishers I spoke to were getting their DMP gameplans together, with particular focus on raising GRP compositions.

Other publishers already have their DMPs spinning: one publisher runs test line items on a weekly basis against OCR and VCE pixels to figure out the percentage of data that matches. The buffers are then adjusted manually. Publishers can also employ lookalike and act-alike modeling on users that match demos to build larger segments. Once again, these must be tested thoroughly and modified based on performance.

With or without a DMP, third-party age and gender data is available from many sources. This can be layered into the DMP or simply used However, third-party data can be highly unreliable (“Some of them perform great, some of them not so much”) and may not scale to meet business demands. Oh, and the costs will mount up, particularly since they’re additive: getting granular beyond age and gender can get very pricey very fast.

Publishers fuse all of these data together into a line of targeting to optimize GRP compositions. But there are further options beyond this to enhance comps offered through ad servers and SSPs. This is particularly handy since many pubs may not have an abundance of first-party audience data or a DMP for major data crunching.

Ad Server and SSP Offerings

“The thing that plagues our industry is additive technology,” a pub suggests. Right out of the box, the publisher ad server should have the technology to optimize for GRPs, rather than forcing pubs to overlay other vendor tools.

Indeed, every supply-side video ad server offers some kind of tool for maximizing GRP exposure – though some come at an additional cost. SSPs and DSPs with supply-side services offer solutions are offering additional solutions that promise reductions in waste.

Typically, ad server optimization services apply Nielsen and comScore data into an algorithm to judge user demo fit based on a variety of signals beyond cookies. Internal forecasting and prediction modeling are applied before and during execution.

To work across platforms and devices, they use their own identification systems and employ machine learning to get smarter as they roll through campaigns. They can also be employed on a campaign-by-campaign basis or in aggregate to gain better knowledge of publisher audience and advertiser demands over time. In addition, the ad server may employ surveys and other tools to garner first-party data, as well as pull in third-party data for better audience modeling.

On the other hand, SSPs or DSPs may serve as impression distributors, not necessarily viewing all of the impressions going through the ad server. For direct deals, it’s not exactly real-time, but does work on a bid-ask model. Basically, these services will take an active role in lining up impressions with advertisers seeking the demo.

An SSP/DSP deal can work on a fee basis or an arbitrage/reselling basis. The latter option might be better for pubs concerned about pricing: the SSP/DSP is paying a set amount to the publisher while selling for more. However, the publisher will not know the final price, which makes setting a priority level tough.

In addition, there’s less transparency into the advertisers coming through when competitive separation is essential in video pods (e.g., placing Coke and Pepsi ads back-to-back is a serious faux pas). Also, formatting issues can become a problem, particularly when transacting across platforms – “We can’t transcode on the fly,” one pub told me.

SSPs and DSPs also add great complexity to managing inventory hierarchy – you might have numerous services competing for the same impression, but who wins out? The highest CPM or the one most likely to fill the demo? Also, keep in mind forecasting video is a bear as the channel is more subject to traffic peaks and troughs.

The advantages are there, but the management is the tricky part. “It’s the same as having a handful of solutions for programmatic – this is just for demographics instead,” comments one publisher.

Several publishers I spoke to were using such tools, sometimes in tandem with first-party data crunching. However, one pub found in an evaluation that the cost of ad server and SSP services was not worth the brand lift. Granted, if used on an aggregate rather than on a campaign basis, these tools claim to learn and update their models.

100% In-Demo?

“If TV never existed, we’d have a much more efficient digital model,” a publisher laments. “The problem is we’re imposing these longstanding TV processes – the agency being the largest one – into a system that could easily be direct to advertiser and direct to consumer, with the publisher serving as middleman… The lack of a more streamlined buying model hurts.”

While some of the folks I spoke to had been guaranteeing the majority of their video inventory against audience for a few years, several publishers told me that they had been running a limited amount of GRP-guaranteed campaigns. They’re expecting that amount to rise dramatically with the 2015 upfronts and newfronts.

The proliferation of OCR/VCE is also increasingly tapping into video with no TV backbone – and the scale non-broadcasting video publishers may not equal a huge uptick in prices, particularly without solid waste management. The entire digital ad ecosystem will likely be taking a crash course in TV operations.

One publisher suggests that while it’s not perfect, time and experience have improved the company’s ability to estimate what it will take to deliver against a demo guarantee. But delivering 100% in-demo is simply unfeasible.

“In a perfect world, you [the viewer] are logged in, I know exactly who you are, and my waste for OCR/VCE is zero,” he says. “That’s the end goal. Are we ever going to get there? No.”

However, given how much data is being collected and processed, another pub feels we should be able to get damn close to 100%. Service providers should be smart enough to develop better capabilities or partner up to offer them. Layer in publisher data-crunching efforts and that 100% doesn’t seem too far off.

“Even now, it should be 98% or above,” he says. “It’s not near that, but there are smart enough people out there who can figure it out.”