Cookies, signals and AI – Vijay Kunduri’s insights into the world of video

The combination of first-party intelligence with engagement signals is where the real cookie-less value is built, says Dailymotion managing director APAC Vijay Kunduri. After delivering a keynote speech at IAB’s Video Summit, he shares his thoughts on how the world of video and measurement is evolving.

StopPress: Engagement signals, watch time, completion and interaction are increasingly doing the work that cookies used to do in video advertising. How reliable are they, and what are their limits?

Vijay Kunduri: Engagement signals are not a direct replacement for cookies and framing them that way misses the point. They are a different kind of signal, more relevant in some ways, more limited in others. A cookie told you who someone was, whereas signals like watch time, completion rate and micro-interactions tell you what they actually cared about and how deeply.

But the real conversation needs to start one level upstream, because the most durable signal in a cookie-less world is not what happens inside an ad creative, it is what happens inside the video player itself.

This first-party data includes what content they are watching, on which device, at which moment of the day, how long they watch, what they skip, what they return to. That is a continuous, consent-based behavioural signal that builds a genuine picture of audience intent and consumption context without relying on any third-party identifier.

The combination of first-party intelligence with engagement signals is where the real cookie-less value is built. The honest limits remain: passive formats give you thin signal; cross-device stitching still requires a first-party data layer to be coherent; and engagement alone does not close the attribution loop. To prove business impact, you still need lower-funnel infrastructure. Anyone selling signals as a standalone solution is oversimplifying.

SP: Where are you seeing AI-driven optimisation genuinely improve outcomes for video advertisers, and where is it still more promise than reality?

VJ: The clearest wins are in three areas we can substantiate with actual data, not roadmap ambition.

• First: Pre-flight creative validation. AI tools trained on tens of thousands of tested campaigns can now generate frame-level attention heatmaps before a single impression is served. This means advertisers walk into activation already knowing their creative will land and reduce wasted spend measurably.
• Second: Contextual targeting. AI now analyses text, audio and visual content simultaneously, across the video and the surrounding webpage, to understand the full emotional and thematic register of a placement. That is fundamentally different from keyword-based blocking, it operates at a granularity and speed no human team could match.
• Third: Continuous optimisation engines. Running these 24/7 can reallocate budget, adjust creative delivery, and manage frequency in real time without configuration lag. This delivers compounding efficiency gains that close the gap between campaign intent and actual execution.

Where it is still more promise than reality is guaranteed ROAS outcomes at scale. The attribution infrastructure to prove a direct line from video impression to purchase is maturing but not globally ubiquitous. It requires specific data partnerships that take time to build, and AI creative generation for video still needs human strategic judgment to determine what to produce, not just how fast to produce it.

SP: How quickly are things changing in the world of video, and what changes or trends are you expecting to see as the year goes on?

VJ: The pace is structural and accelerating. The open web and walled garden environments are diverging aggressively: the global video ad market grows at roughly 10% CAGR while open web video revenue declines sharply. Forrester is projecting a 30% cut in open web display budgets in 2026 alone, and CTV absorbing the growth at over 14% YoY with a clear trajectory to surpass traditional TV ad revenue by 2028.

The battles to watch are these:

• CTV moving from a reach conversation to an outcomes conversation, where the question is no longer whether to activate on the big screen but who can prove measurable business impact in a living room context.
• Attention becoming the dominant currency, with CPM and viewability increasingly treated as floor requirements rather than proof points, and measurement conversations accelerating toward brand lift connected to actual purchase behaviour; and AI-native campaign workflows becoming a genuine competitive differentiator, not the technology itself, which is commoditising fast, but the system built around it that creates data loops, workflow lock-in, and execution speed.

The middle ground is collapsing: players who cannot prove outcomes and cannot compete on walled garden reach are being squeezed out. The surviving independent video players will be those who can match walled garden precision while offering the transparency the open web promises.

SP: What impact has AI had on the video space and what is it most useful for here?

VJ: AI’s most concrete impact in video advertising is in the infrastructure work that does not make headlines but drives real commercial leverage.

• Content understanding at scale. AI can now analyse a video’s full semantic content, tone, language, and visual composition to determine brand suitability, emotional register, and audience profile with a granularity no manual taxonomy ever could.
• Audience modelling without using personal data. Instead lookalike modelling on first-party cohorts and behavioural signals from contextual patterns delivers precision targeting that is privacy-safe and increasingly accurate as the data loop matures.
• Production acceleration. Generative AI is compressing creative workflows by 70% in live deployments through automated subtitles, AI chaptering, thumbnail generation, and multiformat content repurposing. None of these are experimental, they are already running at scale with measurable usage.

What AI is not yet solving is the trust gap: advertisers still want human expertise to validate AI-driven decisions, and the transition from AI-assisted to AI-autonomous campaign management is real but the market is not ready to remove the human from the loop entirely, which means the platforms that win will be those using AI to amplify expert judgment rather than replace it, at least for now.

SP: What do you see looking at the New Zealand video market from the outside that we might not see ourselves?

VJ: From the outside, the New Zealand video market looks too concentrated and too conservative.
Audience behaviour is already fragmented across devices, platforms and formats, much like the rest of APAC. But media investment has not fully caught up. There is still a heavy dependence on YouTube for reach and BVOD for quality, while the broader video ecosystem is under-tested.

In larger markets, advertisers routinely use test budgets to widen the partner mix, learn what works, and improve effectiveness. New Zealand marketers could benefit from doing the same: trusting agencies to test more partners, challenge old assumptions and optimise beyond the obvious channels.

The real unlock is measurement. If marketers are clear on what actually matters, whether that is attention, incremental reach, business outcomes or efficiency, they can make smarter decisions instead of defaulting to the safest options.