Ensuring Brand Safety with Online Programmatic Advertising
As YouTube’s advertising crisis seems to show no signs of abating in the near future we decided to dig a little deeper into the issue and offer our readers a more detailed explanation of what is actually happening.
It’s rare for the ad tech industry to be thrust into the mainstream media spotlight, such is the complexity of the systems that power the programmatic advertising ecosphere. Not since the TV program The Gruen Transfer has the advertising industry captured public attention in quite the same way. Alphabet’s stock fell over 5% and senior management continue to be dragged before government panels for terse questioning. Each day, major brands continue to either pause their campaigns or pull out completely from the world’s most popular video platform.
This reaction comes as no surprise to players within the industry as we know risk shy management will always tend to err on the side of caution when it comes to brand safety. Best to overreact than appear nonchalant.
However, is YouTube to blame, or is the real cause of this crisis less obvious? With so many vested interests at play, we decided to clear the air and separate fact from fiction.
The recent Google brand safety crisis is yet another example of how the bigger picture is lost amongst a knee-jerk propensity to finger-point or deflect blame in the advertising industry.
Everyone needs to have a more realistic view of brand safety when using all advertising mediums regardless of whether they are online or offline. Whether it’s the risk of outdoor advertising being defaced, in-store promotions failing to update, TV spots being re-scheduled or brand ambassadors failing to embody a company’s core values. The digital world is not without the same risks, but this risk can be mitigated with the right approach and know-how.
Technical Solutions Already Existed
There is a whole range of tools available on the measurement end, the buy side as well as sell side that enable agencies, trading desks and publishers to make sure ads are delivered in a way that avoids association with undesirable elements. However, these tools come at an additional expense, compromising the CPM bids that traffickers can place. For some reason, big brands are won over by agency proposals that can offer the most reach for the lowest CPM. They might win the account, but you’re not doing your brand any favours. Costs will be cut somewhere along the line and brand safety mechanisms might be one. This work also requires manual intervention from highly skilled programmatic advertising staff who are in increasingly high demand and short supply.
There is a huge difference between someone who is familiar with a branded ad server such as Google’s Double Click Campaign Manager versus a specialist who understands all the moving parts of the online advertising system. A much deeper level of understanding is needed to avoid undesirable brand associations consistently.
We asked a trafficker (programmatic media buyer who executes the campaign at a DSP level) how he would prevent his client’s ads appearing where they should not. Joel from, On Point Programmatic explains, “There are many filters available to keep ads safe, usually with 100% success rate as long as the filters are activated correctly on each and every media buy. Video is more of a challenge especially when dealing with User Generated Content [such as YouTube] and the volume of ad spaces it delivers.”
Like to read more articles like this? Register for email alerts
The User Generated Content Dilemma
The bulk of content on YouTube’s platform is ‘user-generated’ and we can’t stress enough that you understand what this means.
Anyone with a Google account can start a channel on YouTube and with enough activity, they can then proceed to monetize this content by allowing brands to advertise on the channel. The problem lies with YouTube’s sheer popularity. Over 300 hours of video is uploaded every minute and this content is almost instantly available to watch. YouTube simply makes more money, the more inventory (ad spaces) they can release for advertisers to buy. Perhaps YouTube is a victim of its own popularity. Popular videos tend to be more interesting to viewers, but if the topic of that interest may be unsuitable or offend certain viewers, the advertiser has a dilemma. To associate or not associate? YouTube also has to decide which videos will be added to the available pool of advertising inventory for purchase by the market and which will be excluded.
Currently, there are automated algorithm-based systems that scan the video for any overabundance of skin shades or nipples to automatically filter out any explicit adult content during the upload process. Further checks occur to ensure it doesn’t infringe upon a large database of copyrighted content such as movie clips or music tracks. However up until this recent crisis, there has been little need to censor political or personal opinions. Some argue, they have turned a blind eye to some more controversial content in the name of free speech and to capitalize on the ability to generate additional revenue by offering more advertising inventory.
There are limits to the detection powers of any automatic algorithm so a second type of safety mechanism has been created on a user level. This relies on viewers to flag videos they consider to contain inappropriate content. Of course, to prevent spammers from exploiting these features, other parameters are applied (such as unique flag frequency and flag to view ratios) before the video is manually reviewed or removed entirely.
Facebook relies on a similar user-based flagging system to deal with inappropriate content and comments, but this is far easier to police due to the fact that you need to be signed in to even start consuming Facebook’s content in the first place. This way, Facebook can learn your preferences and tailor content to your desires in an effort to avoid exposing you to content you will not respond desirably to.
YouTube viewers on the other hand, can be completely anonymous is they wish. There’s also nothing stopping someone with a large following from suddenly changing tact or going rogue with their next content piece, which brings us to the exact catalyst for this most recent controversy.
When this crisis started
Many people don’t realise that this YouTube crisis has actually been brewing for quite some time. The flashpoint came in earlier this year in February. PewDiePie (YouTube’s most popular channel) was lambasted by the media when it was revealed that his channel contained a selection of videos which featured ‘anti-semitic references’. Journalists from the Wall Street Journal ‘uncovered’ popular videos which they deemed unsuitable for YouTube or its advertisers and asked for comment from PewDePie’s sponsors. One of these just happened to be Disney-owned Maker Studios who quickly pulled their support before others followed suit. You can read the full story here https://www.wsj.com/articles/disney-severs-ties-with-youtube-star-pewdiepie-after-anti-semitic-posts-1487034533
A kosher salt advertisement preceding one of these videos would not be the best match for an agency’s client. However, what was lost amongst the ensuing PewDiePie scandal was that the offending videos in question were uploaded nearly 6 months prior to the WSJ article, far before they made headline news around the world. In all the years prior and up until mid February, thousands of brands were happily leveraging his viewership to communicate their advertising messages en-masse.
There are only a few explanations for past behaviour. Either advertisers were unaware, didn’t understand how video ads work on YouTube or didn’t care.
What makes this situation even more surprising is the context. If you’re familiar with this particular YouTube channel, you would know his personal brand consistently highlights a propensity to show performative incoherence to content that is typically shocking, crazy, ridiculous or contextually random. Wall Street Journal’s investigative journalism revealed nothing new or surprising to anyone who has watched more than one of his videos. PewDiePie’s irreverence to anything serious is what made the channel famous in the first place. Once the controversy deepened however, YouTube was forced to react by cancelling the PewDiePie web TV show and removing the channel from Google Preferred (a packaged selection of YouTube’s most popular channels available for purchase by advertisers).
Takeaway point – YouTube is a ‘user generated content’ platform. Ensure your client understands what this means and the risks prior to advertising on the platform. Advertisers have been associating their clients’ brand messages with ‘unsuitable’ content for quite some time. This recent crisis probably highlights the fact they were completely unaware this was occurring. Clients and agencies must first understand the media they buy before complaining about its unsuitability after the fact.
The challenge of Classifying Video Content
When a video is uploaded, the video’s category is chosen by the user, with the assumption that they would logically select the most appropriate category to give the content its best chance of attracting viewers. However, even this presents issues. If a video is about different food types types that can enhance libido, should the video’s primary category full under food, self-improvement or mature/adult?
Embedded within all reputable publishing networks, there are a bevy of writers constantly adding to the collective volume of content that people are drawn towards. Very rarely do any brand safety issues arise out of campaigns that leverage the popularity of written content. Digital verification tools make it a breeze to target these pieces via keyword analysis, content categorisation and other contextual features. However, video content platforms pose a unique problem for digital advertisers. Subtle lifts of the eyebrow can turn a simple statement into a question. Just modifying the style of a laugh can convey sincerity or aloofness. Similarly, one’s interpretation of the same video can vary markedly. Where one person may laugh, another may take offence.
Joel again comments, “Until there is robust vetting technology for this available, short term solutions may be needed on the publisher side such as approving users or, in this case, ‘YouTuber’s’ for advertising.”
The digital marketing myths startups treat as Gospel Truth
Typically startups try to avoid spending resources on strategies that don’t work. Performance is essential for continued funding. However in the Silicon Valley bubble, too many firms have fallen for these 3 SEO myths
Surefire ways to ensure your next Advertising Campaign is successful
Want to know the way to make your next marketing campaign successful? Of course you do. Cover these 7 critical bases.
Marketing Managers’ Quotable Statistics Cheat List
Trump opinionated discussions with these handy factual Digital Marketing statistics based on recent studies by reputable U.S organisations.
Audiences not placements
I find it difficult to comprehend that a brand could be unaware they were advertising on Youtube’s most popular channel unless their agency kept them in complete darkness or they didn’t care to ask. However, programmatic RTB’s very nature is perhaps part of the problem.
You see, perhaps the most omitted topic from this whole industry discussion so far is the fact that all of the largest advertisers do not typically purchase online media tied to particular channels or websites. In reality they buy audiences.
While purchasing ad space with a particular publisher was customary in the print world and in the early days of online advertising, the digital programmatic world does not work this way. A network of data companies, harvest the browsing habits of users across the internet and sell this data to Data Management Platforms (DMP’s). The data are used by programmatic advertising traffickers to form audiences which match the target market of the client’s product or service. Think of audiences like you would a selection of internet users grouped together based on a number of similarities they share.
Digital advertising systems will serve certain ads to these audiences as they browse around the internet. This means that if I was targeting middle aged women looking for a new car, I would probably get very few of my ads appearing on PewDiePie’s channel due to the fact that his following is predominantly under 25yo males. My ad would only show if a user who matches the audience that I’m targeting, goes to that channel and watches one of his videos.
Takeaway point – Large advertisers and their media agencies do not generally buy ad space on particular publications or YouTube channels. They purchase available impressions associated with a pre-defined online audience instead. Unless restrictions are placed to filter out controversial placements, advertisements will have a higher likelihood of appearing regardless of which sites that user visits while browsing.
Have you ever had an online advertisement follow you around the internet, even between your mobile and desktop computers? It can be very annoying but the technology driving this bombardment is otherwise known as re-marketing or re-targeting. This is where an advertisement is shown to you depending on a previous action you have made, typically a visit to the brand advertiser’s website or app.
Once you’ve visited a particular website or contact point, you’re then more likely to receive more ads from that advertiser courtesy of the tracking cookies in your browser. The problem, however, is that you might venture to certain places on the internet that may be deemed unsuitable to the brand advertiser if they were watching over your shoulder. Without correct campaign configuration that brand’s ads would appear in all those undesirable places automatically. For this reason, re-marketing is one of the most common sources of conjecture when it comes to ads appearing where they shouldn’t. Technically speaking, if that person has previously interacted with your brand, are they not a qualified audience and is further advertising unwarranted?
To complicate things further, some brands will handle re-marketing activities internally using packaged platforms such as Criteo or Adroll. They will also using an agency to execute a separate campaign using similar creative to people who have not interacted with their brand previously. Determining who was responsible for creative appearing in undesirable locations can be more difficult to ascertain when this is the case.
Takeaway point – Ensure you apply the same filtering process if you’re using separate re-marketing platforms.
How to Approach Brand Safety with UGC
At the current point in time it would seem that whatever content doesn’t get flagged on YouTube is suitable for viewers to consume. You may have noticed recently the addition of an automatic transcribing function if the video contains speech. This transcribed text will be invaluable for advertisers as they will be able to further avoid negative associations much like they do with certain news articles.
The role of the programmatic trafficker is to interpret what a brand wants and doesn’t want to be associated with and apply the necessary available technology to avoid undesirable associations from occurring. Some aspects of what a brand considers brand safe can be missed, but others should be applied nearly every time as they are just common sense. Unfortunately an automated system doesn’t work based on common sense so cracks can form in any brand safety strategy.
Everyone needs to have a basic understanding of how the programmatic buying process works online. Perhaps the programmatic world’s automated features and comprehension complexity is the real issue?
Takeaway point – YouTube has already pulled any controversial content from their advertising inventory, meaning there is very little chance brand safety concerns should be an issue with future advertising.
Reacting to the future
Google are making progress on automated systems to caption images. Their systems can tell the difference between an image of a ‘bear sitting on rocks in a stream’ and a ‘bear in a forest’. You’ve all seen facial recognition software automatically suggest a tag for your friend’s face when uploading images to Facebook. The problem with video is that this complexity is compounded exponentially. There are 25-30 still images per second and usually a complementary audio track of variable quality combined with voice. To automatically categorize and interpret video that is ‘brand safe’ or not, they will need a massively complicated neural network with synaptic intelligence for audio and intelligent visual perception that strings together every frame. This exercise is neither quick, cheap or easy.
Perhaps it is best to pull some available inventory from the platform, roll out a new feature or two and turn up to the list of government inquiries until the storm passes. In fact, this is exactly what has happened (see below)
All channels who had monetized accounts, are now finding some of their videos are now de-monetized. This prevents advertisers from buying inventory associated with those particular videos. You can view the video below from one such channel owner who explains the situation from his perspective in more detail and Google’s recent changes in response to the crisis.
A final note
Until more advanced technology is able to classify user generated video content to a greater degree of accuracy, brand managers will just need take a step back and accept there are risks when advertising online. Pulling all your media spend from the world’s most watched video platform is a premature knee-jerk reaction. The programmatic advertising world is still an area very few people comprehensively understand and many marketers need to learn more about.
Until an algorithm can accurately caption a silent video of two cute kittens falling over onto a sleeping dog, humans will still have a crucial role to play in the programmatic media buying process.