Oracle Data Cloud
This project focused on evolving Grapeshot’s contextual intelligence technology into Oracle’s Contextual Intelligence offering, providing pre‑bid brand safety, suitability and contextual targeting across web and video inventory for thousands of global marketers.
By integrating real‑time page and video‑level analysis into Oracle Data Cloud, the platform enabled advertisers to avoid unsafe environments and reach more relevant content contexts at the scale of tens of billions of programmatic impressions each month, without relying on user‑level identifiers
Timeframe
2018 - 2019
Client
Oracle Data Cloud & Grapeshot
Global brands were increasing spend on programmatic advertising but struggled to control where their ads appeared, leading to reputational risk and wasted media spend when ads ran alongside unsafe or off‑brand content.
At the same time, privacy regulations and the erosion of cookie‑based targeting created a need for scalable, ID‑free contextual solutions that could work consistently across web pages and video platforms like YouTube.
Oracle wanted to strengthen its Data Cloud offering with a unified contextual intelligence capability that could deliver brand safety, suitability and contextual targeting at massive scale for thousands of marketers.
The solution was to industrialise and integrate Grapeshot’s Contextual Intelligence Platform into Oracle Data Cloud as “Oracle Contextual Intelligence” / “Oracle Context”, exposing it as a standard brand safety and contextual targeting service across display, video and programmatic buying platforms.
This included the web page product for contextual targeting and brand suitability, and a video‑focused offering (“Contextual Intelligence for Video”) that classified videos using titles, descriptions and on‑page text to drive pre‑bid brand safety controls.
My contribution was on the core contextual platform and its advertiser‑facing surfaces, ensuring traders and marketers could define custom taxonomies and inclusion/exclusion keyword sets, then activate those segments reliably via “Context Segment Builder” integrations in programmatic platforms such as Xandr/Microsoft Invest.
Challenges
We had to maintain low‑latency, high‑throughput classification while processing tens of billions of impressions per month in dozens of languages, so that contextual decisions could be made pre‑bid without degrading auction performance.
Aligning product naming and behaviour under Oracle’s “Contextual Intelligence” umbrella while preserving existing “Grapeshot” workflows for thousands of active customers required careful design of segment models, UI language and migration paths.
The video product added further complexity, since we needed to infer context and risk from sparse metadata and on‑page text rather than full article bodies, while meeting increasingly strict brand‑safety expectations around topics such as extremism, misinformation and sensitive news.
Outcomes
The integrated contextual platform enabled over 5,000 marketers to apply brand safety, suitability and contextual targeting consistently across web and video, enhancing more than 38 billion programmatic ad impressions every month and growing at over 100% year‑on‑year.
Advertisers were able to avoid unsafe or controversial content that could damage brand equity, while also extending reach by targeting late‑breaking news and highly relevant contextual themes without relying on user‑level IDs.
For Oracle, Grapeshot’s technology became a core part of the Data Cloud / Advertising stack, closing the loop with Moat measurement and strengthening its competitive position in privacy‑first, context‑driven advertising.