KERV's Gary Mittman On The Benefits Of Making Every Moment Actionable
This is the last of five Innovator Spotlight articles from our new Special Report, The Contextual Revolution, Five Companies Rewriting The Rules Of CTV Advertising. You can download the report for free thanks to KERV.ai and our other sponsors.
“Moment based metadata is a game-changer because it allows us to get super granular with our targeting,” says Gary Mittman, CEO of KERV. “Instead of tagging an entire episode with broad categories, we can pinpoint specific moments—even on live content.”
ALAN WOLK (AW): Can you explain how KERV’s polygon-based object recognition works?
GARY MITTMAN (GM): Traditional bounding boxes are the default for a lot of video analysis, but they’re really limited. Think of them as drawing a rectangle around an object—it’s quick, but it often picks up parts of the background. If you’re trying to isolate something specific, like a shirt or a car, the bounding box might include other elements, like someone’s arm or a patch of grass in the background, which throws off the results.
Polygons are different. They let us trace the exact shape of an object, down to the pixel edge. So, if we’re analyzing a shoe, for example, the polygon will only pick up the shoe, ignoring everything else. This makes the match much more accurate when we’re looking for similar products in a retailer’s catalog. It’s especially important for categories like clothing, where shapes and angles can vary so much. Without this level of precision, you’re stuck with a lot of mismatched results, which isn’t great for shoppable ads or anything else.
AW: How are you able to process content in real-time, especially around live content?
GM: We rely on neural networks, which are like artificial brains—they’re trained to recognize patterns in data by analyzing thousands, sometimes millions, of examples. Our neural networks are trained to identify objects, actions, and even broader contexts within video content. When it comes to live content, the challenge is speed. There’s no room for lag when you’re analyzing a live sports game or breaking news. Our AI has to analyze each frame as it’s coming in, identify the key elements, and make decisions about what they mean—all in real-time.
One of the ways we make this work is by focusing only on the most important frames. We don’t need to process every single second of a live feed; we’re looking for key moments where something actionable happens. That’s how we can provide the kind of precision brands and publishers need for real-time applications like live sports sponsorships or ensuring brand safety during a live news event.
AW: How does your technology allow you to deliver shoppable ads so quickly?
GM: The speed comes from how we’ve streamlined the process. For example, if a retailer like Macy’s provides a product feed, our system immediately starts matching those products to what it sees in the video. If there’s a 30-second ad with multiple objects, like clothing, furniture, or accessories, we can scan that ad and connect it to the retailer’s inventory in just a few days.
Earlier solutions could take two to three weeks to deliver something similar, but we’ve built our technology to handle this in hours. That kind of turnaround is critical for time-sensitive campaigns, like a flash sale or a holiday promotion. And it’s not just about speed—we’re also able to maintain accuracy, which is just as important when you’re delivering shoppable ads.
AW: How does KERV integrate entire product catalogs into its shoppable content technology, and what are the benefits of that?
GM: The integration starts with our ability to handle massive amounts of data. Take Walmart’s catalog, for example—it includes over 1.3 billion SKUs. Our system can process that catalog and match objects from video content to individual products, all in real time.
It’s a dynamic process too. If an item goes out of stock or is only available in certain regions, our technology accounts for that automatically. This ensures that the products consumers see in an ad are not only relevant but also available to purchase. It’s a win for both the retailer, who gets to showcase their entire inventory, and the consumer, who has a seamless shopping experience.
AW: What is the advantage of KERV’s moment based metadata?
GM: Let’s say there’s a scene in a cooking show where the host is preparing a meal at scene 5. We can tag that exact moment and serve an ad for a meal kit subscription or cooking utensils during the next ad break. The same goes for live content—we can analyze what’s happening in real time and adjust the ads accordingly. This level of precision not only makes the ads more relevant but also improves engagement, which is what advertisers are really looking for.
AW: How are you helping advertisers with brand safety and suitability issues
GM: This is one of our core strengths. Let’s use live news as an example. News content can be unpredictable, which makes some advertisers hesitant to buy spots there. But our technology can analyze the feed in real-time and distinguish between different types of content.
For instance, we can flag a segment where a chef is showing off recipes as lifestyle programming and keep it separate from a breaking news segment about a natural disaster. This way, brands can confidently advertise on news channels without worrying about their ads appearing next to something inappropriate. It’s about more than just avoiding bad placements—it’s about giving brands the tools to be proactive in managing their image.
AW: Can you talk me through KERV’s API suite and how it works?
GM: Our API suite is what makes everything run smoothly. The Content API is the foundation—it generates the metadata for video content. From there, we have specialized APIs for different use cases.
For example, the Pause API is designed for shoppable ads. When someone pauses a show, it can pull up a menu of products related to what’s on screen.
The Moments API is another one—it helps identify specific scenes or moments that align with an advertiser’s goals, like a family dinner or a scenic outdoor shot.
These APIs give our clients a lot of flexibility to customize their campaigns and make the most of our technology.
AW: Is KERV working with publishers, advertisers, or both?
GM: We work with both, and that’s a big part of what makes us effective. For publishers, we help them monetize their content more effectively. Whether it’s live news, sports, or on-demand shows, we provide tools that make their inventory more valuableFor advertisers, it’s about ensuring their campaigns hit the right audience at the right time. By working across both sides, we’re able to connect the dots and create a seamless experience for everyone involved.
AW: Looking ahead, how do you see contextual targeting evolving?
GM: Contextual targeting is becoming the go-to solution as privacy concerns push the industry away from traditional audience targeting. What’s exciting is how much potential there is to make ads feel more natural and engaging.
At KERV, we’re focused on pushing this forward. Our ability to target specific moments in content—not just broad categories—is a big step in that direction. And as more brands and publishers adopt this approach, we’ll continue to innovate, whether it’s by improving our AI, expanding into new markets, or finding new ways to connect advertisers with viewers.