November 6, 2023

Using BrandGuard to Assess Coca-Cola’s Real Magic

Sagittis sollicitudin turpis nullam rutrum commodo sagittis mattis donec adipiscing rhoncus quis feugiat quis venenatis tristique mi tempus

Using BrandGuard to Assess Coca-Cola’s Real Magic

New mobile apps to keep an eye on

Auctor purus, aliquet risus tincidunt erat nulla sed quam blandit mattis id gravida elementum, amet id libero nibh urna nisi sit sed. Velit enim at purus arcu sed ac. Viverra maecenas id netus euismod phasellus et tempus rutrum tellus nisi, amet porttitor facilisis aenean faucibus eu nec pellentesque id. Volutpat, pellentesque cursus sit at ut a imperdiet duis turpis duis ultrices gravida at aenean amet mattis sed aliquam augue nisl cras suscipit.

  1. Commodo scelerisque convallis placerat venenatis et enim ullamcorper eros.
  2. Proin cursus tellus iaculis arcu quam egestas enim volutpat suspendisse
  3. Sit enim porttitor vehicula consequat urna, eleifend tincidunt vulputate turpis

What new social media mobile apps are available in 2022?

At elit elementum consectetur interdum venenatis et id vestibulum id imperdiet elit urna sed vulputate bibendum aliquam. Tristique lectus tellus amet, mauris lorem venenatis vulputate morbi condimentum felis et lobortis urna amet odio leo tincidunt semper sed bibendum metus, malesuada scelerisque laoreet risus duis.

Sit enim porttitor vehicula consequat urna, eleifend tincidunt vulputate turpis

Use new social media apps as marketing funnels

Ullamcorper pellentesque a ultrices maecenas fermentum neque eget. Habitant cum esat ornare sed. Tristique semper est diam mattis elit. Viverra adipiscing vulputate nibh neque at. Adipiscing tempus id sed arcu accumsan ullamcorper dignissim pulvinar ullamcorper urna, habitasse. Lectus scelerisque euismod risus tristique nullam elementum diam libero sit sed diam rhoncus, accumsan proin amet eu nunc vel turpis eu orci sit fames.

  • Eget velit tristique magna convallis orci pellentesque amet non aenean diam
  • Duis vitae a cras morbi  volutpat et nunc at accumsan ullamcorper enim
  • Neque, amet urna lacus tempor, dolor lorem pulvinar quis lacus adipiscing
  • Cursus aliquam pharetra amet vehicula elit lectus vivamus orci morbi sollicitudin
“Sit enim porttitor vehicula consequat urna, eleifend tincidunt vulputate turpis, dignissim pulvinar ullamcorper”
Try out Twitter Spaces or Clubhouse on iPhone

Nisi in sem ipsum fermentum massa quisque cursus risus sociis sit massa suspendisse. Neque vulputate sed purus, dui sit diam praesent ullamcorper at in non dignissim iaculis velit nibh eu vitae. Bibendum euismod ipsum euismod urna vestibulum ut ligula. In faucibus egestas  dui integer tempor feugiat lorem venenatis sollicitudin quis ultrices cras feugiat iaculis eget.

Try out Twitter Spaces or Clubhouse on iPhone

Id ac imperdiet est eget justo viverra nunc faucibus tempus tempus porttitor commodo sodales sed tellus eu donec enim. Lectus eu viverra ullamcorper ultricies et lacinia nisl ut at aliquet lacus blandit dui arcu at in id amet orci egestas commodo sagittis in. Vel risus magna nibh elementum pellentesque feugiat netus sit donec tellus nunc gravida feugiat nullam dignissim rutrum lacus felis morbi nisi interdum tincidunt. Vestibulum pellentesque cursus magna pulvinar est at quis nisi nam et sed in hac quis vulputate vitae in et sit. Interdum etiam nulla lorem lorem feugiat cursus etiam massa facilisi ut.

Earlier this year, The Coca-Cola Company invited digital artists around the world to “Create Real Magic” by creating new original artwork from iconic creative assets from the Coca-Cola archives using generative AI. Thirty creators were then selected to join the Real Magic Creative Academy at their headquarters in Atlanta this summer to take part in a three day workshop to co-create content for the company, an initiative that highlights the company’s commitment to move quickly to test, learn and scale ideas using AI. They released many of the images that came out of the workshop in this LinkedIn post by Pratik Thakar, Global Head of Generative AI at Coca-Cola and this article in Highsnobiety.

At BrandGuard, we teach machines to understand brands. Using an array of different AI models, we understand the essence of a brand from hard and fast rules such as logos, colors and fonts to the nuances of look and feel and tone of voice. BrandGuard acts as a validation system, ensuring every piece of content conforms to the brand’s ethos and aligns with the brand’s identity, style and values. We were excited to test our technology and showcase how BrandGuard works on these generative AI assets released by Coca-Cola. 

Since we did not have an official style guide or training assets from Coca-Cola, we trained our models using 30 recent advertising and Coca-Cola creative assets readily available on the internet to mimic a general consumer’s understanding of the brand. Then we compared the results of our models against survey responses from 122 consumers who were asked to rank how on-brand they felt each image was, on a scale of 1 to 10. We then specified that an asset would be deemed on-brand if the BrandGuard score or average consumer value was above 55%.

Key Takeaways:

  • BrandGuard scores agreed with the panel of consumers 85% of the time about which assets are on-brand for Coca-Cola, using above 55% as on-brand and below 55% as off-brand.
  • In the three instances where BrandGuard disagreed with the panel, there are a few observations.  In one case, the BrandGuard and Consumer scores differ by a very small margin but they end up on different sides of the threshold.  In the other two cases, BrandGuard scores the content slightly higher, and in the other piece significantly lower.  The reason BrandGuard uses scores is so thresholds can be set for what is considered on brand or off brand (we designated 55%) at the same time allowing a person to dig deeper into the score. This offers the opportunity for any errors to be caught during human review and that person can give feedback to the system to further train and fine-tune the models.
  • To achieve these strong results, models were trained using only 30 public samples. Compare that with the thousands of exposures the consumers have had to Coca-Cola ads over the past 5 years. We would expect further model improvements from adding additional data, particularly more internal brand centric data from Coca-Cola.

It appears from the released content that this initiative was designed to explore and push the boundaries of the brand, which explains why both BrandGuard and consumers scored only about 20% of the Create Real Magic content as being on-brand for Coca-Cola when compared with a common cultural understanding of the brand built on exposure to past advertising. BrandGuard can be used both when stretching a brand, to measure how much the content strays from a brand’s identity and style or for a more common use, to validate large volumes of creative content to ensure all of it remains high quality and on-brand.

To understand how BrandGuard accomplishes this feat we can dig a bit deeper into BrandGuard’s inner workings. BrandGuard works by using over 36 different machine learning and AI models to examine various aspects of brand content. We have four main types of models - Safety, On-Brand, Style Conformance, and Compliance models. Safety models check content for nudity, violence, and other socially inappropriate material. Essentially, would a normal person find this content socially acceptable. On-brand models take existing assets for a brand and use them to fine tune several deep learning models that have been trained on tens of millions of branded assets to ensure that the content conforms to the essence of the brand.  Style Conformance models ensure that content adheres to the rules in a brand’s style guide. Compliance models are used to ensure that content doesn’t violate trademark, copyright, and any other relevant regulations.

Generative AI is an exciting new technology that enables creatives to be more creative.  However, we understand the problems facing generative AI adoption - brands need a trust and safety layer in their AI stack. BrandGuard provides a machine powered validation system that operates at the same speed and scale of a machine powered creation system.