John Snyder, Grapeshot co-founder, joins us to discuss the future of contextual advertising, and how contextual tech solves some of the industry’s trickiest challenges.
I was deeply saddened to hear that Oracle Data Cloud is closing its advertising data services, including Grapeshot, the contextual targeting and brand safety business I founded in Cambridge, UK, in 2006 and later sold to Oracle.
At the time of acquisition in 2018, Grapeshot was breaking new ground for contextual. A dedicated team of people was making significant headway in positive targeting, defensive brand safety, and autonomous, trends-based content predictions for individual brand campaigns.
Back in 2016, it felt amazing that we could crawl hundreds of millions of new web pages per day, adding them into a massive index of the open internet from where we translated them into queries to find relevant campaign briefs. I am so proud of the engineers who put together systems driving 2m queries per second, or 4bn searches an hour, across 148 countries and 70 languages – with just milliseconds to give DSPs the answer, per bid.
Grapeshot was truly global in reach and delivered precision targeting based only on the URL, and the words written into each page. It hurts me to see this world-class search technology buried beneath the rubble of closure because of our sector’s pre-occupation with behavioural targeting over the last ten years.
Facing the addressability crisis
Meanwhile, of course, the addressability crisis continues - Google may no longer be axing cookies, but an unstoppable wave of privacy and awareness was started with GDPR, and customers will most likely be switching cookies off in Chrome themselves.
There remains now a huge opportunity for technologies which can identify and expand campaign audiences without privacy concerns - and the opportunity is greater than just a workaround. We have the chance to reset the entire ecosystem and to do a better and more respectful job of understanding audience behaviours; to be more accurate with how we categorise content, and to improve and extend the capabilities of brand safety.
Contextual technologies always had a higher purpose in my mind; a potential to do so much more than target pre-bid segments, and contextual is now at the centre of solving these challenges.
I’m especially excited to see the impact of advanced technologies such as AI and Machine Learning, which are bringing live behavioural data into contextual operations for the first time, without breaching privacy rules.
One company in particular, in which I have now invested - Illuma - has been working at the interface of context and AI for the last 10 years. After considerable research, Illuma has introduced a recommendation engine into its contextual targeting system, which adapts placements during live campaigns according to real-time performance and engagement. This new way of understanding audience behaviour, and acting upon it, is enabling cookieless audience expansion across the ecosystem, including in CTV, and is paired with AI-powered categorisation for brand suitability.
In other developments, first-party audiences are being profiled based not on IDs but on the contexts they habitually read or like. With third-party data from distributors waning or publicly dying, like Oracle’s decision on advertising, the swing to first-party data is the obvious kernel for developing larger look-alike models – but mediated through context. If you know your existing customers’ contexts well, then those contextual insights coupled with feedback loops on campaign performance, deliver you an open sea of ad inventory on which to bid, without having a first-party data relationship at all.
Context has always been the essence of media-buying, and it matters more to everyone as we enter the AI era, in which contextual intelligence will be vital for AI to be able to respond and create. Context is a driving power behind ChatGPT, for example, matching the context of AI prompts with content trained from the internet, so just imagine how context will drive communications in the planning, creation, execution and distribution of ads.
Oracle has walked away from this highly valuable contextual capability, while Illuma is perpetually pushing forwards with its dynamic performance feedback loop enabling real-time decisioning. The future capabilities are endless, and soon anything to do with context will be better than tracking cookie IDs.
So, as I reflect on the shuttering of Grapeshot and the end of an era for a dedicated team, I remind myself that this vision for context and intelligence continues - and I’m hugely excited about what lies ahead.