No one can blame advertisers for exercising extreme caution right now. When news emerged early last month that publishers’ revenues were suffering due to the ad-blocking technology that brands were using to keep their ads away from Covid-19 content, it wasn’t surprising.
In an environment as febrile as the past couple of months, businesses have been in a state of shock. Fearing the reputational effects of their ads sitting next to content that may be about death, fear, and illness, they have opted for what feels like the safest option: Simple keyword blocking to avoid content containing words like “Covid-19”, “pandemic”, and “quarantine”.
The blow this strategy delivered to publishers’ revenues was so severe that the UK’s culture secretary Oliver Dowden was forced to make a statement encouraging some media placement leniency in order to protect what he referred to as the UK’s “Fourth Emergency Service”.
Scanning the bigger picture
But now we can show for the first time how damaging this policy has been to advertisers themselves. Starting in late March, we have used our proprietary machine-learning tool Verity to analyze over 6m unique pages of internet content containing various Covid-related words.
The majority of these pages would have been ad-free because they contained “unsafe” words associated with the pandemic. But running them through our machine learning contextual analysis system, which takes into account the broader context in which these words appear, we have discovered that a massive 4m plus pages – 66% of the total – were completely safe environments for advertising.
Part of the problem is that because the virus is affecting every facet of life at the moment, even the most innocuous content includes references to it. Imagine, for example, an article reviewing newly published books that begins by mentioning that more people are reading now due to the quarantine. The ad inventory on that story would have gone unutilized, blocked by keyword brand safety systems solely because of the word ‘quarantine’, despite the fact that it actually would be perfect inventory for advertising products suited to people who enjoy reading––a new e-book device, for example, or anything else appropriate to that audience segment.
Distinguishing between safe and unsafe content requires a nuanced understanding of context – something the human brain does easily – which keyword-blocking technologies fail to deliver in their quest to classify great volumes of online content quickly.
Thankfully, systems modeled on the human brain exist which deliver something like human contextual understanding but at superhuman speed and capacity. Contextual brand safety solutions like Verity use cutting edge AI-powered tech to develop big picture contextual compression of online articles. Using a combination of computer vision to look at images and video and Natural Language Processing (NLP) we have trained machine learning systems to effectively ingest an article and identify broader meaning rather than isolated keywords. For example, it can easily differentiate an article about apples (the fruit) and Apple (the tech company).
The net effect of this is that we can avoid unsafe media environments for brands but without culling so much content that advertisers’ campaigns become significantly less effective and publishers face dwindling revenues (which could ultimately drive CPMs up for advertisers).
Safety is relative
With citizens globally sitting bored in lockdown, there has been an enormous leap in media consumption as they hungrily devour news and entertainment. The reality is that across many content categories, the vast majority of articles that people are reading represent a safe environment, even when there is some reference to the current crisis.
If brands are looking for a quick and easy brand safety solution right now – rather than keyword blocking or whitelisting everything – a simple approach might be to advertise on content categories like technology, pop culture, video gaming, fitness & exercise whose content is overwhelmingly safe. Advertisers will get plenty of reach and over 80% of COVID-related content within those categories is safe.
The real beauty of using a contextual tech approach to sifting media is that it is designed to be customized for each brand, allowing advertisers and their agencies to set different levels of sensitivity to virus-related content. As a result, advertisers can truly begin to achieve brand suitability, rather than the broad-brush promise of brand safety.
What this means in practice is bespoke safety for individual brands, where one advertiser’s brand and the audience may embrace a greater degree of exposure to content about medical, political or social aspects of the current crisis, another may only be able to bear oblique references to its effect on pop culture.
Contextual brand safety is about understanding these needs and delivering a media plan that reflects each brand’s desired level of risk in maximizing media exposure.
As we are hearing now in relation to plans to ease the lockdown, it is impossible to entirely erase the risk without to some degree also compromising individual freedoms. Likewise, brands cannot be entirely safe without compromising their commercial and brand reach. With sophisticated contextual analysis tech, however, brands can achieve a much more favorable balance, where they can both advertise in safe media environments and reach audiences at scale.