By David Terdjman, associate director, Head of Buyside Services, Xandr
Third-party cookies disappearing is nothing new but what marketers do next is still an ongoing discussion. Marketers will have to rethink their programmatic media plan to continue to target their audiences and while multiple solutions have already emerged, it is clear that the easiest one to activate is contextual.
Back to basics
Contextual is not something new. It has been existing for a while now. However, the level of sophistication of contextual technologies has tremendously grown over the last few years. For many years, contextual was restricted to extracting a few keywords from a web page: it was like bringing in an elephant to kill a mouse. Let’s take an example: April 2020, in the middle of the covid crisis, you are a travel company wanting to target your audiences. Although you want to avoid pages dealing with covid deaths or travel restrictions you would like to be on pages talking about hotels compliance with covid measures. In this scenario, banning the keyword “covid” would prevent you, the travel company, from buying impressions on relevant pages, thus becoming ineffective. This has since advanced drastically with new contextual technologies now involving semantics analysis and natural language processing which allows media traders to target these relevant contexts.
Finding out the secret sauce
Contextual technology can be separated into two phases: qualification of context and activation.
To qualify their audiences, contextual technologies can use one or several methodologies:
- Semantic analysis – leveraging natural language processing, contextual data providers can run an offline analysis on millions of URLs, discovered by crawlers or using bid streams. The next step will be extracting a meaning from the content of the web page: instead of just identifying keywords, the meaning of a web page can be seen as a graph of keywords, each of them with a specific score depending on their relevance.
- Panels and navigation history – in this scenario, the contextual technology can leverage navigation history from anonymised users, and extract patterns in their behaviours.
- Campaign delivery – analysing the delivery of a campaign, in particular clicks and conversions, is also an efficient way to identify relevant contexts for a brand, which can be used as seed URLs before using semantic analysis to find lookalike contexts.
- Cookies – these can still be used to qualify the audiences. When a data provider has a deep understanding of its cookie-based users, they may leverage this knowledge to identify the most visited URLs by their audiences.
Now that contexts have been identified and qualified, they need to be activated.
First, these contexts must be grouped. Depending on the methodology applied, contextual data providers can either package their URLs into categories, which can be very easy to activate for an advertiser. The catalogues can contain audiences such as ‘car enthusiast’ which can be used for either branding purpose or brand safety.
Contextual partners can create a tailor-made audience, with specific rules on the selection of relevant contexts. These custom audiences usually come with better performances as the contextual data provider can concentrate on the most relevant context for the brand.
Finally, when dealing with final activation of these contexts, advertisers will have the choice:
- Deals – contextual data partners can directly package the relevant contexts into a deal shared with the trading desk. This can be done either on SSP side or with data partners leveraging curation.
- Contextual segments – shared via a data marketplace or directly on the trading desk seat – the same way cookie segments are activated.
- Managed service – contextual technologies may also provide a managed service where they will directly operate the contextual activation on their DSP of choice.
What’s next?
In a post-cookie world, contextual technologies will play an important role for advertisers to address their audiences. These technologies are constantly improving to address more uses cases such as retrieving contexts from image or videos and enabling brands to efficiently target their customers. But contextual is only part of the solution, and marketers will also need to get familiar with other technologies such as shared IDs in order to cover use cases such as frequency capping or retargeting that cannot be addressed by contextual technologies.