Creative Intel for Dentsu Aegis Network

In partnership with Google, Dentsu Aegis Network identified an area of opportunity for a Creative Intelligence Unit. The DAN Creative Intelligence Unit would be able to iterate through all of a client’s YouTube creative assets (at scale), identify the elements featured in each video and gather activity data to further inform creative production discussions. The analysis would provide reports to help their clients ensure that the video advertising they are running across YouTube is relevant, drives attention and is fit for purpose. The pairing of DAN’s media, creative and data science expertise with Google’s proven AI and ML tools has the power to make transformational gains.

The following clients/brands were selected for analyses:

  • Phase 1 - Heineken
  • Phase 2 - Ashley Furniture
  • Phase 3 - Part 1 - Telia
  • Phase 3 - Part 2 - Kelloggs

The Tools

In order to derive the most actionable learnings, the following was required from each brand in order to conduct the analysis:

  • Access to the brand's YouTube brand channel and Adwords account. This will enable our DAN Technology Data Science experts to access the backend YouTube Analytics API and pull the appropriate retention report that will allow for deeper analysis and insights.
  • Raw video creative assets (mp4 format) associated with their corresponding TrueView campaign Video IDs. These assets are run through Google’s Video Intelligence API which detects and labels entities (dogs, flowers, people, etc.) within each video.


For the initial phase of our Creative Intelligence engagement, our data science team was provided with YouTube’s Analytics Audience retention report for a set number of videos centered around a specific campaign for Heineken. These videos were subsequently passed through Google’s Video API to provide information about their content. Using machine learning and statistics, the team was able to provide an advanced analysis of key dimensions within each video and compare them against its viewership and retention performance.

Broadly, these dimensions within each video were:

  1. The role of people and product,
  2. the impact of the first 5 seconds,
  3. an analysis of the text on screen versus spoken words,
  4. the impact of visual imagery,
  5. the impact of static vs dynamic videos on viewership, and
  6. an analysis of the impact of discount/promotional mentions.

This information helped to uncover trends behind what had performed well in the past and where there were opportunities for creative refinement. Upon completion of the analysis, an insights report was developed and provided to the client to use for relevant creative strategy and development discussions.

Ultimately, our goal was to help DAN's clients understand what makes consumers watch video creatives in their entirety versus clicking or tapping the “Skip Ad” as soon as the option becomes available. It is about striving to make YouTube ads that are "unskippable" in the eyes of the consumer.


Each AI/ML creative analysis report we are able to generate features:

  • Bespoke ”YouTube Creative Best Practice Insights": Which creative elements allowed the advertiser in question to breakthrough? How to brand effectively and include elements that help to foster consumer connections? What encourages consumers to take action?

  • Assessment Matrix: Beyond the elements that help brands to influence brand metrics, what is hurting brands? How often? What are the opportunities or baseline choices? What should be avoided?

  • A Framework for Actionability: When are these insights relevant? What are the key windows for optimization?