Greg Kihlström

View Original

S6 | 504: Deep learning and the cookieless future with Jaysen Gillespie, RTB House

See this content in the original post

About the Episode

We are here at eTail Palm Springs in Palm Spring, California and talking about all things retail, e-commerce, AI, and more.

While we’ve talked quite a bit about the cookieless future on this show, and even though the impending deadline of third-party cookies may get pushed back… again… it is still happening, and marketers and advertisers need to find solutions that will enable them to maintain their advertising performance.

My guest today can talk about all of this and we’re going to talk specifically about creating a competitive advantage by using deep learning with your data and measurement in an era where the cookieless future is (almost) here.

To help me discuss this topic, I’d like to welcome Jaysen Gillespie, Head of Analytics and Data Science at RTB House.

About Jaysen Gillespie

Southern California analytics, marketing, product, revenue and data science executive with a growth mindset. Currently, I'm building revenue through analytics for RTB House, a leading DSP using Deep Learning to inform prediction and drive performance. 15+ years of experience in technology and analytics leadership roles.

Able to effectively communicate complex concepts at a level commensurate with the sophistication of the audience, and able to keep team members focused on analysis that drives relevant decisions. Deep understanding of how to create competitive advantage using best-in-class analytics and how to message the story to the marketplace.

I enjoy working with all types of data and systems, from high-volume "big data" transactional logs to aggregate KPIs. Experienced speaker, creator of executive content, press interviewee, and global evangelist. I love the whiteboard but will happily take the "hot seat" on a panel if I can help there too! Analytics is far too interesting -- and surprisingly fun -- to keep in the back room. Let's bring your company story to life with data.

See this content in the original post

Resources

Synopsis

Deep learning, as discussed in the podcast episode, plays a crucial role in improving programmatic advertising results by driving incremental sales and providing valuable insights into customer behavior.

  1. Incremental Sales: Deep learning models have the capability to process a vast amount of data compared to traditional machine learning models. By leveraging deep learning, marketers can identify truly incremental sales that wouldn't have occurred without the marketing efforts. This is achieved by analyzing a significant amount of data and understanding the impact of marketing activities on driving additional sales. Deep learning models, with their ability to process large datasets, can identify the malleable middle of customers who are likely to make a purchase with the right marketing approach.

  2. Customer Behavior Insights: Deep learning excels in understanding and segmenting customer behavior based on on-site activities. Unlike traditional machine learning approaches, deep learning can identify patterns in customer behavior that may not be immediately obvious. For instance, it can recognize segments of customers such as gift buyers who exhibit specific browsing and purchasing behaviors. By analyzing these behaviors, marketers can tailor their advertising strategies to target these segments effectively, leading to improved campaign performance and higher conversion rates.

Deep learning offers marketers a powerful tool to enhance programmatic advertising results by driving incremental sales and providing deeper insights into customer behavior. By leveraging the capabilities of deep learning models, marketers can optimize their campaigns, target the right audience segments, and ultimately achieve better outcomes in the evolving landscape of digital advertising.

Shifting Focus to Incrementality and Media Mix Modeling in Preparation for the Cookie-less Future

In the podcast episode, Jaysen Gillespie, Head of Analytics and Data Science at RTB House, emphasized the importance of marketers shifting their focus from traditional measurement methods like Multi-Touch Attribution (MTA) to incrementality and Media Mix Modeling (MMM) in preparation for the cookie-less future. With the impending demise of third-party cookies, marketers need to adapt their measurement strategies to ensure continued advertising performance.

Incrementality as the Ground Truth

Gillespie highlighted that incrementality is considered the ground truth in measuring marketing effectiveness. By conducting incrementality studies and utilizing a true control group, marketers can accurately attribute the impact of their marketing efforts. This approach provides a clear understanding of the incremental sales generated by marketing activities, allowing for more informed decision-making.

Challenges with Multi-Touch Attribution (MTA)

Gillespie pointed out the limitations and challenges associated with MTA. He mentioned that MTA often relies on arbitrary attribution models that assign credit to various touchpoints without a clear rationale. This lack of transparency and consistency in attribution methods can lead to inaccurate measurement of marketing performance. Additionally, the impending cookie-less future further complicates the use of MTA, making it less viable for accurate measurement.

Transition to Media Mix Modeling (MMM)

In light of the shortcomings of MTA and the evolving digital landscape, Gillespie suggested that marketers should consider transitioning to Media Mix Modeling (MMM). MMM, grounded in strong mathematical underpinnings, offers a more robust and data-driven approach to measuring marketing effectiveness. By leveraging MMM, marketers can gain deeper insights into the impact of different marketing channels and optimize their media allocation strategies for better results.

Deep Learning and Measurement

Gillespie also highlighted the role of deep learning in driving incremental sales and improving measurement accuracy. Deep learning models, with their ability to process vast amounts of data and understand complex user behaviors, can provide more accurate insights into the effectiveness of marketing campaigns. By utilizing deep learning for measurement and analysis, marketers can better identify the malleable middle of their audience and focus their advertising efforts on driving incremental sales.

In conclusion, the podcast episode underscores the importance of marketers embracing incrementality and Media Mix Modeling as essential measurement strategies in the face of the cookie-less future. By shifting their focus towards these more reliable and data-driven approaches, marketers can adapt to the changing digital landscape and ensure the continued success of their advertising efforts.

Deep learning plays a crucial role in customer segmentation by identifying unique segments and in personalization by understanding the long tail of products and customer behavior. In the podcast episode, Jason Gillespie, Head of Analytics and Data Science at RTB House, highlighted how deep learning can excel in customer segmentation by identifying segments that may not be immediately obvious to other AI approaches like machine learning. He provided an example where deep learning was able to identify a segment of gift buyers based on their behavior on an e-commerce website. This demonstrates how deep learning can pick up on sets of behavior that may not be easily recognizable by traditional machine learning algorithms.

Moreover, deep learning's ability to understand the long tail of products and customer behavior is essential for personalization efforts. Traditional machine learning algorithms may struggle with recommending products that do not sell quickly or have limited data. However, deep learning can create subgroups within the long tail and provide personalized recommendations based on these unique segments. This is particularly beneficial for retailers with a high SKU set, such as Macy's, as deep learning can effectively personalize recommendations for a wide range of products.

Overall, deep learning's capability to identify unique customer segments and understand the long tail of products and customer behavior makes it a valuable tool for enhancing customer segmentation and personalization efforts in marketing campaigns. By leveraging deep learning algorithms, marketers can gain deeper insights into customer behavior, optimize targeting strategies, and deliver more personalized experiences to their target audience.

Jaysen Gillespie, Head of Analytics and Data Science at RTB House