← Marketecture: Get Smart. Fast.
Loading summary
Transcript33 lines
- [00:01]
A
This podcast is brought to you by audiohook, the leading independent audio dsp. Audiohook has direct publisher integrations into all major podcast and streaming radio platforms, providing 40% more inventory than what could be accessed in omnichannel DSPs. What's more, audiobook has full transcripts on more than 90% of all podcast inventory, enabling advanced contextual targeting and brand suitability. Audio Hook is so confident that in addition to CPM buys, they offer the industry's only pay for performance option where brands can scale audio and podcasting with peace of mind mind knowing they are only paying for outcomes. Visit audiohook.com to learn more. That's audiohook.com welcome to Market Tecture where you can get smart fast with in depth interviews of leading technology executives. I'm joined today by Nikhil Jain who is the VP of Insights and Marketing Solutions at dailymotion Ads. Nikhil, thank you so much for being here.
- [01:05]
B
Thanks a lot Ari for your time and inviting me. Excited to have this conversation?
- [01:10]
A
Yeah, exactly. I want to hear what's going on. So why don't we start with the basics. Like for those not that familiar with dailymotion, dailymotion video website people know that for a long time video app. But what's the what is the current state of Dailymotion?
- [01:22]
B
Ah, that's a great question. So I think Dailymotion, most of us know it about the Dailymotion.com, the platform where we where people can stream videos. However, when we talk about dailymotion advertising, the platform has evolved a lot over time and so is our technology stack. One of the core USPs that we have is our video player. Video player is present in more than 5,000 plus publishers around the world and what this helps us do is this helps us activate videos at scale. Meaning whenever publishers upload videos on their websites, people are watching it actually using a dailymotion player and so they are users of dailymotion without even realizing. And everything is connected to our ad tech stack that makes us activate campaigns on dailymotion.com but also on this huge layers of publishers that we have partnership with. So premium in stream inventory at scale.
- [02:20]
A
Right. So you're effectively able to sell and monetize ads on the website but also on a big network of websites.
- [02:29]
B
Exactly 100% correct. Right.
- [02:32]
A
So Dailymotion is often talking about attention. I think that's what we wanted to talk about specifically. So what's your definition of attention? What are you working on? What's available to the customers?
- [02:43]
B
That's a great question. So I'll start off with the first question. So what is attention like? Attention is anyone's ability to captivate or like breakthrough noise in order to make sure your message is passing through. And I think that's where when we talk about attention. When we launched our attention program, we focused on using attention as a strategy and not just as a KPI that we wanted to optimize. Today we try to think of attention from different perspectives, from understanding audiences by doing deep consumer behavior researches, by touching these audiences, by identifying the best messages and then creative ad formats that resonate with them. And in order to do this, we do eye tracking tests and AI powered analysis in order to understand what are the pros and cons of an ad format, what is attracting attention and how we can optimize those. Of course, when we talk about media activation, we focus on the quality of inventory and that's where we have our own. Instead of just measuring attention and optimizing it post impression, we have built our own models that help us predict the attention that would, thanks to our partnership with Explain and Lumen, that help us identify whether a bid will actually be a high attention bid or not. And then try to focus on those inventories and then post campaign. Of course we do focus on media performances and everything. But the notion is to try to understand what was the impact on the brand itself. And that's where we try to focus on doing brand lift studies or like KPIs beyond what we started to call vanity metrics or media KPIs such as the impact of your brand image or brand preference. So trying to focus on outcomes and that's where like introducing this notion on planning activation and then measurement to then focus on the next campaign.
- [04:30]
A
Right, that's really interesting. So let's break that down. So is attention primarily an input like a predictive value or is it also an output KPI?
- [04:41]
B
That's a great question. So I think it is definitely an output. That's something that is easy to understand. When you watch an ad, are you focusing or not? It's easy to identify that that is an output. When you are trying to measure attention at an inventory level, that's actually an output as well. Because you would be able to know thanks to different criteria what is impacting attention, what is not impacting attention. However, what we have tried to do is we have tried to use it as an input based on the learnings that we have had over these different periods. So what we have tried to do is thanks to our partnership with explain, we have tried to send them a lot of granular data associated with past impressions and try to create a model that identifies these variables in real time in order to make and then predict attention. So based on these different variables, such as what is the proportion of your ad compared to the screen of the ad, whether the ad is visible or not, if the sound is on or not. So we have a list of 40 such variables that we take into account in order to predict attention. And that's where attention is not an input, but these variables become the input that help us predict attention almost in real time.
- [05:56]
A
It's hard to get attention if you can't see the video, right?
- [06:00]
B
Yeah, I think, I think that's where like there's this level of inventory. Like there are two layers of attention. The first one is inventory, but then the second one is the creative itself as well. So yeah, of course there's the quality of inventory. But then is your ad captivating enough? Like maybe if you just. The example I like to give is you have a video player and it's blank screen, you're going to get 100% attention. But does that add any value?
- [06:26]
A
Do you help the advertisers figure out if they're getting attention based on the creative?
- [06:31]
B
Yeah, that's something we try to do. That's something we work with Lumen in order to do eye tracking tests in order to understand whether an ad is actually captivating attention or not and or how can it be improved. And that's where now we are working with AI powered solutions that help us actually do these tests at scale in real time. So you upload a creative, you see the different neuroscience KPIs thanks to our partnership with neurons and that helps us understand and analyze what's working and what's not. But that's where we are doing still descriptive analysis and that's where we have our own in house studio that actually identifies how can we improve our digital ad experience. So taking an MP4 copy that could be a TV ad, but making it digital friendly by either adding certain layers of enrichment that could be static brand logos, an additional cta, a key message that we want people to remember. Or it could be interactive elements that help us better engage the audiences and maximize digital experiences and it works.
- [07:44]
A
Yeah, that makes sense. So is there a standard for attention or is it still in sort of the IB trying to figure out what it means?
- [07:53]
B
No, the IB is still trying to figure out what it means. However, I think a few months back they did come up with the first set of guidelines opened it for everyone who is interested in order to contribute so that they could start making it a norm or identify things. However, certain elements are very similar to what we built our attention program two months back. They also focus on the quality of inventory, the creative itself, the audience speed, the contextual piece. So yeah, I think the basic building blocks everyone is aware of and the next step is to standardize them across the ecosystem.
- [08:32]
A
Right. And we've had other guests on the show talk about attention. So I just want to ask you sort of the similar question what I asked them, which is how do you have any evidence besides the obvious that attention is correlated with other outcomes and KPIs?
- [08:47]
B
So that's a great question and I think one of the things that we have tried to do is we have tried to do a lot of brand lift studies in order to measure outcomes across the entire funnel. So if we talk about recall, interest till consideration or purchase intentions, and a few things that we have noticed are whenever we are optimizing on attention, we are able to naturally impact the funnel. However, if you only focus on attention as a KPI, meaning you have high attentive seconds, you are able to optimize short term KPIs. Like my definition of a short term KPI is a recall or a consideration. However, when you start using attention as a strategy, and these are researches that we did internally, we are able to optimize longer term KPIs such as awareness or preference. So simply optimizing on the KPI is not enough. However, if you focus on your long term strategy and start keeping the basics of reaching the right audience with the right message, but using attention as a core element of it. That's where the long term KPIs of brand preference or even your awareness after a certain extent moves the needle.
- [09:58]
A
Interesting. So, onto a new topic. So AI is what everyone's talking about. It was pretty much the only subject it can. So tell us about your AI efforts and what you're working on.
- [10:10]
B
I think one of the major things that we are working on using AI is how can we simplify everything associated with planning the campaign, activating the campaign and then measuring the campaign in order to close the loop in terms of recommendations and working along. One of the first things that we want to work on and are working on is trying to create something that helps eases out the life of a media planner, meaning today, researching about an audience is a very tedious task because data comes from multiple different elements. And then the question that everyone asks Is, okay, I have this marketing Persona. How do I translate into an activable Persona? And our goal is to bridge this gap using all the first party video data that telemotion has, its capacity, but also the different consumer behavior studies that we have worked on over the past four, say four years. Train an AI where as an end user I can start talking to it in terms of giving a brief, this is the audience I am looking for, what are the different elements, these are my campaign elements, etc. And in a few clicks it first does a complete analysis in terms of what the brand and the objectives are, associates it with the most relevant video signals that we have, and then identifies the most relevant Personas, but not based on generic interest, based on the exact nomenclature that dailymotion has so that they could be directly activated as part of your campaign.
- [11:48]
A
That's interesting. So when we're talking about video, is video a good source material for AI training or is it hard? You have to analyze it at the frame level, give us some insights about that.
- [12:00]
B
Yeah, so I think I would just answer your question in one sentence. Like when I was studying data science, the first thing that our professor taught is like garbage in, garbage out. So if you're not, if you're not actually analyzing the videos in a logical way, everything else would be a function of that. And that's where we have put in core efforts. We try to analyze videos based on the titles and the description in order to do a textual analysis to understand what they are talking about. Then we take it a step further by doing a complete video analysis or frame by frame analysis in order to understand the different elements that are present, but also the sentiments that are associated with the video in order to create a cohesive list of signals that we have. The third part is the audio. So first we do a audio to text speech to track text translation, and then we redo the entire sentiment analysis and the textual analysis and in order to find elements which are present everywhere. And what this helps us do is it helps us exactly pinpoint what a video is talking about and their context. Like I'll give you a very simple example. Let's say we are talking about a video where there's a football match that's going on and those elements in the title, you might just have the names of the teams and the score line. However, thanks to the commentary, you might be able to identify other players who are present. By doing a frame by frame analysis, you can identify which brands are present in the ecosystem and eventually the sentiments associated with the match that actually help, you know, whether something positive happened or it was a catastrophe that happened. And that's eventually helps us do precise video analysis.
- [13:53]
A
So are you doing that, you must have millions of videos on your website, your network, Right. Are you doing that in every frame of every video?
- [14:00]
B
Exactly. We are doing it on every video that's uploaded. Because I think there are two layers of analysis. Firstly, when a video is uploaded, it is assessed whether it's brand safe, whether it meets the criteria so that it can be present on our platform. And the second one, if it's deemed as monetizable, we try to classify it into IAB categories so that they follow a standard nomenclature that could be used worldwide. But we have our own dictionary of 700,000 plus topics that help us precisely understand each and every element associated with a video.
- [14:36]
A
Right. And what have you found much so far in terms of insights about what drives attention, what drives results?
- [14:44]
B
I think one of the things that has been really exciting for us is initially when we talk about contextual targeting. The idea of contextual targeting was I'm going to associate food related videos with food related advertisers. And one of the things that we have learned is like this, this idea of networks that exist or these themes which exist where food related ads might work very nicely with content that's associated with gaming. It's not the first reflex that you have. However, when you think of it, these are very complementary ideas and these are the elements which we are able to identify by understanding audiences and then growing through these connected networks of elements.
- [15:29]
A
Gamers are hungry. Gamers are hungry. They're sitting around the game and they forget to eat. It's a good time showing out.
- [15:36]
B
Exactly. Yeah.
- [15:39]
A
Sorry to interrupt.
- [15:40]
B
We also do. It's okay. We also do certain brand levels analysis in order to understand what categories are actually correlated. And then like for example, is a specific brand associated with specific football teams or certain clubs and or sentiments. And what this helps us do is take the meaning of contextual to its extreme limits where we can actually associate not only the content but also the sentiment, the intents associated with a video and then create audiences or even do contextual targeting on these broader sets of parameters.
- [16:19]
A
Right, that's all really interesting. So sentiment seems to matter a lot. Like you might upbeat content and food might go together, but dramatic or scary content might not. Have you been looking into stuff like that?
- [16:34]
B
Yes. So I think what we try to do when we talk about creatives is to associate sentiments with these creatives and then try to take a decision on how we want to improve these creatives and and that's something we want our AI to do. Like what we wanted to understand is okay based on different neuroscience KPIs. These are the strengths of a creative. But then from an emotional level this creative is more associated with these type of emotions and therefore these are the type of enrichments that we should have. Or at times sometimes something over emotional might not work with certain types of creatives and therefore like how you can nuance the balance of emotions in order to better personalize the contextual targeting that you are aiming for. So different variants of creatives on different types of different variants of creatives on different types of context associated with sentiments and emotions.
- [17:35]
A
Well, let's call it on that. So Nikhil Jain, thank you so much for giving us all these insights and telling us about what dailymotion is up to in attention and AI. Thank you for listening to the marketecture podcast. New episodes come out every Friday and an insightful vendor interview is published each Monday. You can subscribe to our library of hundreds of executive interviews at marketecture tv. You can also sign up for free for our weekly newsletter with my original strategic insights on the week's news at News Marketing Architecture tv. And if you're feeling social, we operate a vibrant Slack community that you can apply to join@adtechgod.com.