Understanding brand safety & brand suitability

As said in the guide, ‘Understanding brand safety & brand suitability in a contemporary media landscape,’ by IAB US: “2020 has been the year of the unexpected. From a pandemic and quarantine to economic crisis to a resurgent civil rights movement and the most polarized our nation has been since the Civil War, this chaotic time impacts business decisions at every level, including advertising." 

In the face of this chaos, each individual advertiser needed to become shrewd about how to run campaigns suitable to its brand's identity during a historic moment when a large amount of publisher content was focused on hard news - such as growing death tolls due to coronavirus, #BlackLivesMatter, and protests that might turn violent.

At the same time, issues of brand safety have begun to evolve beyond avoidance of inappropriate sites, malware, spam, and adult content, to include larger and more difficult-to-pin-down considerations of brand suitability. Brand suitability refers to specific targeting parameters unique to each brand, as determined by its values. For some brands, an ad placed outside a specific targeting parameter might lead to a brand suitability incident with potential negative consequences that extend beyond reputational damage. 

It appears, however, that despite brands' best efforts to run campaigns suitable to brand identity, damage is being done. As evidenced in the Integral Ad Science’s Media Quality Report H2 2020, published in April 2021, UK media quality has been compromised amid the unprecedented circumstances of the last year.

The data, which examines advertising campaigns that ran between 1 July and 31 December 2020, shows UK brand risk increased across all media environments analysed (mobile and desktop display and video) compared to the same period in 2019. Desktop display brand risk rose the most over this time, jumping 3.2% to 5.8%, representing the highest level of brand risk in this environment since 2017 - a year laden with brand safety controversy. 

Meanwhile, brand risk on mobile video rose to 8%, making it the highest risk environment of all, but it experienced the smallest change at +0.2% year-on-year.

“As substantiated by this report, brand safety continues to remain a real issue,” says Smartology CEO Mark Bembridge. “At Smartology we recognise the serious impacts of brand associations and have automated exclusions to ensure there are minimal risks of brand content sitting alongside inappropriate content. Marketing professionals can be unknowingly exposing their advertisements alongside unsavoury content and in many instances it is extremely difficult to prevent.”

Brand suitability technology has been available for several years but many offer blunt keyword solutions. We often see agencies request keyword exclusion lists for campaigns consisting of hundreds of manually entered keywords. These are ill-thought through, incomplete and full of spelling mistakes which, they argue, are important to cover instances where words are spelt incorrectly. We often question whether brands themselves are actually aware of just how porous these lists are. Often they include words that conflict with a brand’s objectives.

Smartology’s SmartMatch SaaS platform not only manages brand safety so brands know which media owners their campaigns will be running across at all times, but it has built in semantic exclusion lists that cover all the main inappropriate subjects most brands would wish to avoid. The semantic nature of SmartMatch exclusion is important since it means that ‘natural disasters’ relates to all sub-categories such as hurricanes, tsunamis, volcanic eruptions etc and ‘job losses’ relates to sub-categories including unemployment, layoffs etc. While this is much more bulletproof than a list of keywords, it is also much more effective when taking into consideration a brand’s objectives. 

Negative associations can sometimes be important to a brand’s positioning. During the pandemic, for example, SmartMatch saw the majority of its brands move rapidly to create new branded content suitable to the more challenging news cycle. A global management consultancy, for example, wrote about market volatility during the pandemic. This matched well alongside the news cycle when aligned alongside articles about markets being affected by the crisis, but would risk reputational damage when matched alongside news about Covid hospitalisations.

These examples highlight how keyword brand suitability is not fit for purpose - a blunt instrument that simply blocks all negative associations causing publishers problems selling their inventory and problems for brands when pushing valid opinions and positions on important themes.

Semantic brand suitability can go much further by up weighting or down weighting semantic categories. It recognises how brands can avoid inappropriate articles in the above examples associated with death, injuries and casualties (and their associated terms) while allowing them to align with important subjects to them featuring in the news cycle where their expertise, learnings and thought leadership has high impact and value - despite appearing alongside challenging subject areas. 

That is not to say that appending brand safety with keywords is not valuable. Semantic brand suitability can only go so far. Negative sentiment is not something any vendor can manage and this is unlikely to change. So where a client wants to avoid being matched alongside any mention of their share price falling this can be described in too many ways to protect effectively; share price dropped, collapsed, fell off a cliff... In these scenarios brands are advised to add their own brand name to a keyword list to append to a semantic approach.

More often than not, brands also have specific associations that they are keen to avoid where they may wish to add specific company names and other sensitive topics peculiar to them to blanket protect against. In these scenarios, it’s helpful to append a semantic approach with keywords, which SmartMatch allows clients to append to their brand suitability before and during campaigns. 

However, consider a recent third-party vendor keyword list shared, which included several names of terrorists. While, at face value, this seems sensible from an advertising campaign perspective, in what scenario would these terrorists be mentioned outside of a wider piece referencing terrorism? The same list had scant reference to terrorism and its associated terms. Just adding keywords without thought to overall campaign effectiveness is not helpful.

‘Trump’ was a term often added to keyword exclusion lists during the last US election, which blocked huge swathes of news. While associated references to allegations of fraud, tax evasion, sexual misconduct and violent protests in the press would understandably cause brands concern, several brands sought to position content around how the financial markets were reacting to the election and yet managed to block any of their valid opinions from appearing alongside relevant articles. 

Of course, we must also not underestimate bad actors intentionally seeking to position their content to nefarious ends and it is important to police strict rules and guidelines to avoid misinformation being amplified and to only accept brands running campaigns that have been properly vetted. Where scale is balanced with focus around premium brands and titles, this is more straightforward than operating where individual user-generated content at scale becomes impossible to effectively police. 

It has been widely documented that during the pandemic, brands have been busy writing content. In fact, it is estimated that 63% will invest in content creation in 2021 and 32% shifted traditional paid advertising dollars to content marketing in the last 12 months. As the opening paragraph states, news sites have been and are continuing to reel with hard, sometimes startling news. News by its very nature is often negative to a large extent and good journalism should position both sides of stories. However, there are many branded content pieces that would be well placed alongside some of these articles. 

For example, those who can share resolutions, breakthroughs, good deeds or positive opinions that potentially counter the negative undertones of an article. The challenge here is that some ad tech solutions are not able to understand the true context of an article in its entirety at a granular enough level, and therefore to determine the appropriateness of aligning uplifting CSR content directly alongside publisher content that they often incorrectly define as unacceptable. While understandable, it is also counterintuitive if CMOs choose to shrink away from display advertising spend due to these concerns.  

Mark Bembridge explains: “CMOs need not shrink away but are now in a position to embrace content advertising and effectively minimise consequences of ineffective brand suitability. The SmartMatch platform has been developed as a deep learning matching model that is trained using a content set of thousands of proprietary manually-curated matches. On top of this sits a brand suitability model, which evaluates publisher content with regard to various unacceptable semantic categories - such as job loss, natural disasters and racism.

“This model, also trained on thousands of examples that are manually curated, is able to quickly analyse and weight words and phrases used in publisher content to determine the probability of it being unsuitable. This level of sophistication enables brands to not only confidently deliver their content-generated advertisements aligned to trusted media owner articles, to trusted platforms, but also to be safe in the knowledge that advanced semantic brand suitability and brand safety keywords are prevalent and will more effectively prevent matches with any unsuitable articles while providing a more nuanced approach than simple blunt keyword exclusion list tools. So, whilst more is being spent on creating content, we can help you amplify and share at the right time, in the right place to the right audience.”