When Prime Video ads guess your mood before you watch

WPP OpenDoor — Amazon’s dedicated WPP unit — decided that if algorithms know what we want to watch, they might as well know how we feel before telling us what to watch. The result? Last month, it rolled out two campaigns for Prime Video that trade traditional ‘because you watched…’ logic for something closer to ‘because you feel…’

The first, in partnership with Wootag, threads an AI-powered emotion-sensing layer into Prime Video’s digital ad ecosystem. Encounter a banner on a news, lifestyle, health or entertainment site? A quick mood check via your camera — happy, sad, exciting or neutral — instantly serves up a list of titles to match that exact emotional state. One click later, and the viewer is on Prime Video, mood-matched and ready to stream.

The second move folds entertainment into an entirely different ritual — ordering food. Teaming up with Zomato, WPP OpenDoor mapped food choices to viewing moods, turning biryani, pizza, chaat or dessert into launchpads for customised film and series suggestions.

A dessert order could yield The Family Man (‘Dosti wala sweet’), Panchayat (‘Rishton wala sweet’) or Bhool Chuk Maaf (‘Vaadon wala sweet’), delivered via a Prime Video carousel right on Zomato’s order confirmation page.

The strategy addresses a familiar streaming headache: content overload and choice paralysis. Instead of surfacing recommendations from cold behavioural data, it taps into the insight that viewing decisions are often mood-led. These activations were curated to anchor in the OTT’s brand promise, ‘Every kind of emotion… it’s on Amazon Prime’. The agency worked towards weaving emotional intelligence into everyday digital journeys, from a lunchtime scroll to a midnight snack order.

In this exclusive conversation with Campaign, Deepa Jatkar, India lead of WPP OpenDoor, unpacks how mood detection works across publisher platforms, the challenges of integrating it into ad ecosystems, and what ‘emotion-first’ means for the future of content discovery. Here are edited excerpts:

What emotional intelligence layer powers this campaign, and how did WPP OpenDoor integrate real-time mood detection into Prime Video’s digital ad ecosystem across diverse publisher platforms?

The core insight is that ‘watch what you’re feeling’, so streaming content is a thorough emotion-led behaviour. The emotional intelligence layer developed by our tech partner Wootag is like a super-smart detective for feelings. It can instantly read tiny cues from your face like your expressions, gestures, even those quick, subtle reactions.

It then sorts these signals into categories like 'Happy,' 'Excited,' or ‘Angry’! The magic happens right after when based on your detected mood, the tech instantly scans Prime Video's huge library and suggests the perfect show or movie right there in the ad.

What truly sets this apart isn't just how smart it is, but how incredibly fast and scalable it is. This whole process happens in seconds, and it works across IAB-standard display formats, enabling wide reach across devices, audiences, and publisher platforms, without compromising on performance or personalisation.

What were the biggest technical or behavioural hurdle you encountered in this initiative?

Our biggest technical hurdle, which we tackled in partnership with Wootag, was getting advanced AI to work smoothly on a mobile phone within the limitations of a mobile banner ad. AI usually needs a lot of processing power. But the face and emotion detection module identified expressions in milliseconds using only the front-facing camera and the phone's own processing.

This meant making the AI model incredibly lightweight and efficient, small enough to operate within those mobile-ad limits while staying super accurate. Basically, we had to put a powerful brain into a tiny, fast package. That was a serious engineering feat, and Wootag's expertise was crucial in enabling this on-device processing for interactive ad units.

Given users often engage with mood-based inputs casually or randomly, what safeguards were built to ensure the recommendations still feel relevant and deeply personal?

That's a great question, because people's expressions can change quickly! We tackled this by firstly using a technical intervention with our tech partner, what they call 'Multi-signal Emotion Detection.' Instead of just looking at one quick snapshot, their system captures a series of tiny reactions over a few milliseconds.

It combines different signals — like facial cues, where your eyes are looking, and how engaged you seem. This helps them filter out those quick, random expressions. It's about getting a more complete picture of your emotion.

The second one was emotional nuances and we had a creative intervention to tackle this. For instance, if a user shows a sad face, we didn’t want to suggest a sad watchlist.  Instead, we suggested, 'Here’s a feel-good watchlist for you!'

What consumer insights drove the collaboration of mapping Zomato food cravings to content moods?

Apart from this team being foodies, the core idea was simple that we noticed a strong connection between the food people order and the kind of content they might want to watch. For example, if someone's ordering something spicy and tangy like 'chaat,' they're probably in the mood for a chatpata, or 'snappy' show!

So, we worked with Zomato to identify their most popular food cuisines or dishes, like biryani, pizza, chaat, Chinese, and desserts. Then, we carefully matched the Prime Video content to each of these food moods. This extended our overall campaign message i.e. 'Whatever kind of emotion... it's on Amazon Prime.'

Which other platforms are you targeting next for contextual discovery touchpoints?

Beyond Zomato, we've already expanded this idea to digital platforms like ‘The Times of India’. There, users can also get recommendations based on their current mood, again offering multiple options that fit that broader emotion.

For example, if a user chooses 'Dramatic,' they might get options like 'Ghar ka drama – The Family Man' or 'Lauki drama with Panchayat!' Each of these was hyperlinked to the show links.

You aim to transcend data-driven targeting and lean into lived human experiences. How are you designing this to be emotionally resonant yet ethically sound, especially around privacy and consent?

That's a really important point for us. Our goal is to create a more human and relevant experience, connecting with people's actual feelings, not just data points.

Ethically, we've built this with a 'privacy-first' mindset. All mood detection happens locally on the user's device, no facial images or personal emotional data are stored on the servers. It's processed instantly and then discarded.

Crucially, consent is paramount. Users must opt in to this experience. If they don't, the ad simply functions as a standard experience. We believe true emotional resonance comes from trust and respect for individual privacy and choice.