What's missing from the brand-safety conversation: Context

Campaign's recent article “Brand Safety: How the YouTube kerfuffle is playing out in APAC” is timely in raising some key issues. You’re perfectly right in saying that reaction in the region has been rather muted compared to the US and Europe.

However, as the conversation around “#GoogleGate" both deepens and widens, advertisers in APAC are now increasingly conscious of the intimate relationship between brand, messaging and environment as well as what brands should do to increase brand safety.

While there is discussion around accountability and transparency of programmatic technology, the brand-safety debate has largely left out the importance of context.

Effective brand safety models go beyond matching individual keywords on a page, as a single word is not necessarily indicative of overall meaning. A keyword out of context can mean a missed opportunity or a branding disaster.

This is why it’s important that programmatic software evaluates complete texts and does not rely on the appearance of individual indicators. Understanding the context of individual web pages before a programmatic ad is placed enables advertisers to ensure an individual page is relevant, safe and that copy is in the same language as the article on the page.

The key to accurate targeting in brand-safe environments is context. A single keyword can mean different things depending on its context, which is derived from its relationship with other keywords within the content.

My company's technology, for instance, goes beyond individual keywords and analyses all significant keywords within a piece of content to develop 'word clouds’, in over 170 languages. The relationship between all significant keywords in the keyword cloud grouping is then mathematically analysed and evaluated against all our standard and custom segments. Finally, statistical confidence scores are given to each segment, and the strongest matches returned as best matches (the number of returned matches and significance strength can be determined by the buyer).

Categorisations are returned in the pre-bid programmatic environment within a matter of miliseconds, with checks and recategorisation happening every time a page is seen to make sure the data remain fresh and relevant. This also enables advertisers to ensure an individual page is relevant, safe and that copy is in the same language as the article on the page.

Safety within context is actually relative. While most advertisers don’t want their ads appearing next to an article on HIV, a condom brand aiming to communicate safe sex might. Understanding the relevance of context and relativity of brand safety opens up new opportunities—such as live audience targeting. Once context becomes a decisive factor in programmatic placement we can advertise to consumers based on what they’re reading at any given moment, rather than a search they made a month ago.

Keyword blacklisting, the capability to use 'high-speed word clouds' and effective holistic contextual targeting, is also a remarkably efficient way to ensure a brand, company, or product remains safe from unfavorable or harmful content specific to them.

We recommend brands work with their teams to implement a thoughtful and comprehensive brand-safety protocol, starting with custom blacklists that include brand-incompatible words and phrases, inclusion of broad standard brand-safety categorise as appropriate, then using ad quality/verification software to scan websites for other potential issues such as fraud and to report on viewability of placed messages.

When brands work with their technology partners to vet potentially inappropriate placements before bidding, brand safety increases, and potential landmines can be easily avoided.

Chris Pattinson is senior vice president of Grapeshot Asia-Pacific