00:00Je m'appelle Hamza Kouhimat, le chef marketing officier de Dailymotion pour le business.
00:17Generating un vrai résultat pour nos clients
00:19requires plus que ajouter un layer d'AI sur nos solutions existent.
00:23C'est pourquoi je crois que l'AI doit transformer le cœur de notre business
00:27par réinventing trois main pillars de marketing.
00:31The first one is market research and media planning
00:33where AI can speed up the process from a multi-week journey to a matter of minutes
00:39while removing the human bias and analyst bias
00:42and also introducing a concept of synthetic data
00:45on top of the first-party data that we use to analyze consumer behavior.
00:49The second one is KPI definition and optimization
00:52where AI can help us move beyond vanity metrics
00:55like completion rates, CTR, to optimize both media efficiency
01:00and marketing impact on real outcome KPIs
01:04like brand lifts, awareness, sales, for example.
01:08The third one is ad experience or creative experience
01:12where AI can reinvent how brands experience connections with their audiences
01:17moving from mass communication to one-to-one interactions.
01:20I'm quite critical about what I call vanity metrics.
01:30I really believe AI can help advertisers move beyond those vanity metrics
01:35by closing the gap between having an efficient media optimization plan
01:39and optimizing real impact on their marketing outcome that they measure themselves.
01:44The problem with our industry is that media KPIs today are really simple to optimize
01:50and do not really guarantee a measurement of impact, real impact.
01:54So today it's no longer about only optimizing reach and completion rates
01:58but finding the perfect balance between media performance on one hand
02:02and reaching the right audience and optimizing the correct outcome on the other hand.
02:06We are entering today an era where consumers accustomed to interacting with tools like ChatGPT
02:18expect the same level of personalized interactions with brands.
02:21I have introduced the concept of promptable brand
02:24which is a brand that one can simply ask a question, express a need
02:29and receive personalized information and responses
02:32or any piece of content within the ad experience itself.
02:36So today it's the end definitely of a top-down, one-to-many communication process
02:42to a one-to-one communications
02:44where brands can interact and get instant feedback from their audiences
02:48and vice versa by the way
02:50and which changes completely the codes of user engagement within the ad experience itself.
02:55We have launched Ray, the agentic advertising platform by Dailymotion, which basically helps marketing leaders, media buyers, optimize their campaigns
03:11by leveraging four main pillars.
03:14The first one is the agent that helps do the media planning and market research.
03:20The AI leverages tons of first-party data from Dailymotion, from video consumption, demographics, browsing patterns
03:29but also data that comes from their interactions from existing or past video ads.
03:35The data that comes from our DMP is leveraged to transform those signals
03:40into real consumer behavioral patterns that can help identify and define audience segments.
03:46The second pillar is the engagement pillar
03:49which basically transforms the video asset coming from a client
03:53into a way more engaging video ad experience, leveraging neuroscience KPIs, sentiment analysis
03:59to make sure that we are hitting the right concept and experience
04:04that will engage with the audience that we have defined.
04:06The third one is the activation part, which is our smart reach activation engine
04:10that finds the perfect balance between optimizing the media KPI
04:14which is required by the agency and the media buyer
04:17and reaching the right audience and getting the right outcome.
04:19And finally, the fourth one, the post-campaign measurement part
04:23is a way for us to provide consistency
04:27and a clear and comprehensive media report or campaign report for our clients
04:32that move beyond only showing how the vanity metrics and media KPIs have performed
04:38but more towards understanding the outcome that the campaign itself has generated for the client.
04:44I do believe that those four pillars are fueled by one unique selling point of Dailymotion
04:51which is our first-party data added to the synthetic data
04:54that we're able to capture leveraging AI today.
04:56And the combination of those data points are what makes our AI unique.
05:01AI is as good as the data you give it.
05:03Ray today helps marketing professionals and media buyers focus on what really matters.
05:09Moving from audiences to outcome
05:12while optimizing KPIs on both the media part and the marketing part.
05:16Hey, hi, I'm Cyril Brun. I'm a Chief Data and AI Officer at Dailymotion.
05:20Our strength at Dailymotion doesn't come from the fact that we're building our own large language models.
05:31Our strength comes from the fact that we'll build many highly specialized AI agents
05:37that are called upon by supervisors, but also AI agents,
05:41that are able to understand language and nuances within questions to call the specialized agent
05:47that is the best suitable for the task requested.
05:50Some agents will actually use machine learning models to answer specific tasks,
05:55but we also have other AI agents that just understand language
05:59that are plugged with our own data or build a little bit differently.
06:02Agentic AI doesn't replace machine learning models.
06:05However, it brings a more holistic view on solving a problem with different tasks.
06:12Usually, machine learning models will actually optimize for specific metrics
06:16in a very greedy and crude way, going to optimize that metric at all costs,
06:21while AI agentic will try to find different ways to optimize in a more human way the question being asked.
06:32One key use case of the AI agentic is that it helps us solve the vanity metrics problem
06:42by optimizing for more complex objectives for advertisers.
06:47Classical machine learning models will optimize for specific metrics like completion rate in a very crude way.
06:55For example, by delivering massively on contexts where the player is on autoplay,
07:01where the attention rate is pretty low.
07:04So it will hit the KPI, but not the advertiser objectives.
07:09So the idea is to be able, through the supervisor work,
07:14to balance a little bit different objectives,
07:17whether it's the KPI for completion rate,
07:20but also more complex marketing objectives like attention or brain lift.
07:27The idea is to build a system that is able to understand those advertiser marketing KPI
07:33and to ask upon the right specialist to solve the right problem at the right time with Nuance.
07:39The supervisor will actually guide the different agents towards different objectives,
07:44sometimes pushing more for the completion rate and sometimes more for more advertiser marketing objectives.
07:57Our key advantage today in the market is that we are able to combine the AI agentic system,
08:04machine learning expertise that we've built over the years,
08:07and historical data we gathered over the past years.
08:11So large language models, they are great to answer natural language questions,
08:17but usually they are not fine-tuned to our business use cases, nor to our own data.
08:23So Dailymotion has invested over the past years in delivering machine learning product at scale,
08:29whether it's for optimizing advertising campaign or recommendation algorithms.
08:34We also have a lot of data, for example panel data, creative data, campaigns data,
08:40which we use to fine-tune the large one language models we use.
08:44And overall, what we see as a key advantage is the combination of those three assets.
08:56We see in the market a big shift in the way we define our jobs.
09:00For example, developers will actually spend much less time with tools like cursor, with cloud code,
09:08defining basic functions, they will spend much more time thinking about overall architecture.
09:14The media planner, for example, will spend much less time in back and forth with panels, with agencies,
09:22which we've built, spend much more time answering, getting questions from the tools to iterate on better understanding their personas.
09:34Before, back and forth with agencies and panels was limiting the quality of their work.
09:39Now, they will spend time where it's most interesting, understanding their persona, fine-tuning them and going into strategic thinking about how to use them.
09:50The jobs become less operational and much more strategic.
09:59Make AI work for you.
10:01Make every impression count today.
10:03Reach out to us to schedule a demo and see how Ray can be beneficial for your business.
Commentaires
1