How We’re Using AI in the US

Addressable media has had machine learning and optimization at its core since the beginning. Those algorithms have lived in silos – unportable, forgetful from campaign to campaign, and servicing the platform that built the algorithm over the advertiser.

Emerging solutions like GroupM’s Copilot, Scibids, and Chalice are making our algorithms portable, persistent, and cross-channel. EssenceMediacom and WPP are investing in privacy-safe AI solutions like emSidekick to enable our teams to provide better service and deeper insights faster than ever before. Creative solutions are emerging - taking DCO and interactivity in our media to a new (and kind of terrifying) level.

Our vision is to create a connected biddable practice for Mars that has an engine powered by AI. 2024 is our year to invest in building that AI-powered performance engine and to begin integrating tools into our daily operations.

Learn More

AI in Biddable Optimization

The Plan for the Year

On Social and Programmatic we are exploring AI powered third party custom algorithms that update based on the real-time performance of the biddable campaign.
The algorithms are portable from campaign to campaign unlike platform solutions.
They can handle multiple KPIs, are more transparent, and are integrated with partners like NCS for sales, Scope3 for carbon emissions, and Adelaide for attention.

A centralized algorithm can unlock our ability to optimize across our biddable channels in an unprecedented way.

Mars has broken through this year on Biddable optimization. Across MW, Pet, and Food we have conducted successful tests of third-party algorithms to gain efficiencies in YouTube and attention in OLV. MW also launched a first-to-market test to improve CPMs on Meta 7/23 in our first attempt to scale across channels.

Snickers used GroupM’s Copilot to optimize YouTube campaigns for lower CPMs and increased reach, which ran from May 20th to July 21st with an NCS study. Lifesaver Gummies has launched a first to market Scibids for unique reach on Meta between July 24th and Sep 22nd. Skittles Gummies will compare GroupM Copilot and Scibids for YouTube CPM reduction from August 1 to September 29, also with an NCS study.

In March, Petcare saw 12%-28% CPM improvements using Scibids for YouTube. Scaling started on May 15 across all YouTube media. Future plans include testing Copilot on Meta for conversion. Testing Chalice against Scibids for Meta CPM efficiency is considered, but constrained by Chalice’s high fee and $500k minimum test cost.

Ben’s Original launched on May 15, aiming to optimize towards an Attention metric with an associated NCS study. This is the first use of AI on non-YouTube Programmatic inventory in the US. Initial success was seen in Australia with 40% improvement in reach and 11% lower cost per completed views. Globally, NCS lift is observed on manually optimized attention metrics, highlighting the need for algorithmic optimization.

Background on the Players

Click on a logo to learn more

GroupM’s own AI optimization tool enabling omnichannel optimization. Well integrated with Adelaide and NCS with the lowest rate for optimization services. Has the capability to optimize budgets across social and programmatic based on performance and pacing.
AI-powered third party custom algorithm with global Mars results and specialization in CPM efficiency optimization.
Minimal differentiation and higher cost solution but well integrated with Adelaide and Scope3. Social solution allows for optimization to in-store sales. Currently being used by Hersheys.

AI in our Operations

Generative AI Use Cases

Select a button below to learn more about how we plan to use generative AI via emSidekick in our team operations.

Data Analysis

Platform Optimization

Meeting Transcripts

Email Formatting

Image Generation

Summarization & Case Studies

Legal Considerations

Our first round of use cases is focused on the processing of owned inputs rather than creation. This is due to the complexity of ownership of the output when the inputs used to create the content are unknown. So, functions like strategy are not allowed to be done by generative AI at this time.

Mars and EssenceMediacom are in the process of adding an AI addendum to allow us to process data through emSidekick for the use cases below. We need to have final signoff before we can scale and get broad adoption within EMC.

About emSidekick - EssenceMediacom's proprietary AI operations tool

EssenceMediacom offers a web-based tool to EMC employees called emSidekick, leveraging the advanced capabilities of GPT-4 Turbo. emSidekick has dozens of templates and is tailor made to accelerate the operations of our teams.

emSidekick operates independently from OpenAI's public services, via Microsoft Azure, ensuring that conversations and user-uploaded content remain confidential and inaccessible outside our organization. This tool does not learn from user interactions. Each conversation is confined to its own thread, eliminating the risk of data or content leakage across different client engagements.