- 50
- 17 185
Recast
Приєднався 28 чер 2022
Where marketers learn modern marketing measurement frameworks.
How to Build a Media Mix Model (MMM) - Ep.2: Using Causal Models (DAGs) in MMM
What’s the secret to building a powerful statistical model? In Episode 2 of How to Build a Media Mix Model (MMM), Tom Vladeck and Antonio Vargas from Recast reveal the importance of causal models and assumptions in MMM.
Learn how to:
- Start with your core assumptions rather than fitting data into a pre-existing model
- Use Directed Acyclic Graphs (DAGs) to map causal relationships
- Balance trade-offs in model estimation for actionable results
Antonio Vargas, Senior Data Scientist at Recast, explains how their proprietary Bayesian model integrates complex assumptions into a robust causal framework for MMM.
Whether you’re using open-source models or building in-house, understanding causal relationships is key to successful modeling.
🖥️ Learn more about Recast's MMM platform: getrecast.com/
📚 Explore Recast's Bayesian model documentation: docs.getrecast.com/
🔔 Subscribe to stay updated with the series.
Learn how to:
- Start with your core assumptions rather than fitting data into a pre-existing model
- Use Directed Acyclic Graphs (DAGs) to map causal relationships
- Balance trade-offs in model estimation for actionable results
Antonio Vargas, Senior Data Scientist at Recast, explains how their proprietary Bayesian model integrates complex assumptions into a robust causal framework for MMM.
Whether you’re using open-source models or building in-house, understanding causal relationships is key to successful modeling.
🖥️ Learn more about Recast's MMM platform: getrecast.com/
📚 Explore Recast's Bayesian model documentation: docs.getrecast.com/
🔔 Subscribe to stay updated with the series.
Переглядів: 356
Відео
How to Build a Media Mix Model (MMM) - Ep.1: Introduction to Media Mix Modeling
Переглядів 35214 днів тому
Have you been tasked with building a Media Mix Model (MMM) but feel overwhelmed by the complexity? In this first episode of How to Build a Media Mix Model (MMM), Tom Vladeck, Statistician, Marketing Scientist, and Co-founder of Recast, introduces the challenges and best practices of MMM. Discover why MMM is one of the toughest statistical problems, common pitfalls to avoid, and the foundational...
Recast Office Hours #6 - All Things Uncertainty
Переглядів 8114 днів тому
Recast's Office Hours are educational sessions where marketing-focused data scientists can get answers to their measurement questions from a group of experts. Session #6 begins with a focus on uncertainty and opens up for audience Q&A from there. Interested in attending Office Hours? They're hosted monthly by Recast Co-Founders Michael Kaminsky and Tom Vladeck. Look out on LinkedIn and the Reca...
Recast's Marketing Measurement Coffee-Break Session #15 - Ryan Dew
Переглядів 706Місяць тому
Recast's Coffee Breaks include deep dives into complex topics with experts and discussion Q&A with operators in an informal setting. We'll cover everything from digital tracking and attribution to running effective experiments and MMM. In this session, our co-founder Michael Kaminsky speaks with Ryan Dew. He is an Assistant Professor of Marketing and the Govil Family Scholar at the Wharton Scho...
Recast's Marketing Measurement Coffee-Break Session #14 - Professor Elea McDonnell Feit
Переглядів 245Місяць тому
Recast's Coffee Breaks include deep dives into complex topics with experts and discussion Q&A with operators in an informal setting. We'll cover everything from digital tracking and attribution to running effective experiments and MMM. In this session, our co-founder Thomas Vladeck speaks with Professor Elea McDonnell Feit. She is an Associate Professor of Marketing and an Associate Dean of Res...
Recast Office Hours #5 - Lift Tests
Переглядів 137Місяць тому
Recast's Office Hours are educational sessions where marketing-focused data scientists can get answers to their measurement questions from a group of experts. Session #5 begins with a focus on lift tests and opens up for audience Q&A from there. Interested in attending Office Hours? They're hosted monthly by Recast Co-Founders Michael Kaminsky and Tom Vladeck. Look out on LinkedIn and the Recas...
Recast Office Hours #4 - Robustness Checks
Переглядів 873 місяці тому
Recast's Office Hours are educational sessions where marketing-focused data scientists can get answers to their measurement questions from a group of experts. Session #4 begins with a focus on robustness checks and opens up for audience Q&A from there. Interested in attending Office Hours? They're hosted monthly by Recast Co-Founders Michael Kaminsky and Tom Vladeck. Look out on LinkedIn and th...
Recast's Marketing Measurement Coffee-Break Session #13 - Chetan Sharma
Переглядів 1473 місяці тому
Recast's Coffee Breaks include deep dives into complex topics with experts and discussion Q&A with operators in an informal setting. We'll cover everything from digital tracking and attribution to running effective experiments and MMM. In this session, Chetan Sharma, CEO and founder of Eppo, an A/B experimentation platform for modern data and product teams, joins us to discuss the intricacies o...
Recast Office Hours #3 - Backtests and Holdout Forecast Accuracy Checks
Переглядів 2424 місяці тому
Recast's Office Hours are educational sessions where marketing-focused data scientists can get answers to their measurement questions from a group of experts. Session #3 begins with a focus on backtests and holdout forecast accuracy checks, and opens up for audience Q&A from there. Interested in attending Office Hours? They're hosted monthly by Recast Co-Founders Michael Kaminsky and Tom Vladec...
Recast's Marketing Measurement Coffee-Break Session #12 - Aleksander Molak
Переглядів 4394 місяці тому
Recast's Coffee Breaks include deep dives into complex topics with experts and discussion Q&A with operators in an informal setting. We'll cover everything from digital tracking and attribution to running effective experiments and MMM. In this session, Aleksander Molak, a machine learning researcher, author, advisor, educator, and entrepreneur specializing in causality, natural language proces...
Recast's Marketing Measurement Coffee-Break Session #11 - John James
Переглядів 4405 місяців тому
Recast's Coffee Breaks include deep dives into complex topics with experts and discussion Q&A with operators in an informal setting. We'll cover everything from digital tracking and attribution to running effective experiments and MMM. In this session, John James, a sales and marketing advisor to leading D2C and B2B brands, joins us to discuss brand marketing. We'll explore the history of br...
Recast Office Hours #2 - Parameter Recovery
Переглядів 2155 місяців тому
Recast's Office Hours are educational sessions where marketing-focused data scientists can get answers to their measurement questions from a group of experts. Session #2 begins with a discussion of parameter recovery (what it is, why it's important, how to do it, etc.) and opens up for audience Q&A from there. Interested in attending Office Hours? They're hosted monthly by Recast Co-Founders Mi...
Recast - How Bayesian MMM Works
Переглядів 5526 місяців тому
How does the algorithm that underlies Recast actually work? We get this question a lot, and there are a couple of different ways to explain it. In this video, our co-founder Michael Kaminsky gives a simplified explanation of one way that we like to think about how the algorithm works. We have a bunch of documentation for data scientists or statisticians who want to get really into the weeds on ...
Recast Office Hours #1 - Variables to Include in an MMM
Переглядів 3866 місяців тому
Recast's Office Hours are educational sessions where marketing-focused data scientists can get answers to their measurement questions from a group of experts. Session #1 started with a discussion on which variables to include in an MMM and opened up for audience Q&A from there. Office Hours are hosted by Recast Co-Founders Michael Kaminsky and Tom Vladeck every month.
Recast's Marketing Measurement Coffee-Break Session #10 - Gufeng Zhou
Переглядів 3317 місяців тому
Recast's Coffee Breaks include deep dives into complex topics with experts and discussion Q&A with operators in an informal setting. We'll cover everything from digital tracking and attribution to running effective experiments and MMM. In this session, Gufeng Zhou, Creator, DS lead and PM of Project Robyn, comes in to talk about how Robyn works, features, its inception, and methodology answer ...
Recast's Marketing Measurement Coffee-Break Session #9 - Prof. Peter Fader
Переглядів 4818 місяців тому
Recast's Marketing Measurement Coffee-Break Session #9 - Prof. Peter Fader
Modern Analytics for Marketers (#7): Dynamic Spend Deployment and Forecast Checking
Переглядів 1788 місяців тому
Modern Analytics for Marketers (#7): Dynamic Spend Deployment and Forecast Checking
Modern Analytics for Marketers (#6): Validating Marketing Mix Models with Lift Tests and Experiments
Переглядів 3539 місяців тому
Modern Analytics for Marketers (#6): Validating Marketing Mix Models with Lift Tests and Experiments
Modern Analytics for Marketers (#5): Stability & Robustness Checks for Marketing Mix Models
Переглядів 2529 місяців тому
Modern Analytics for Marketers (#5): Stability & Robustness Checks for Marketing Mix Models
Recast's Marketing Measurement Coffee-Break Session #8 - Prof. dr. Koen Pauwels
Переглядів 2309 місяців тому
Recast's Marketing Measurement Coffee-Break Session #8 - Prof. dr. Koen Pauwels
Modern Analytics for Marketers (#4): Parameter Recovery
Переглядів 4399 місяців тому
Modern Analytics for Marketers (#4): Parameter Recovery
Google Meridian MMM: Key Features and Limitations
Переглядів 2 тис.9 місяців тому
Google Meridian MMM: Key Features and Limitations
Recast's Marketing Measurement Coffee-Break Session #7 - Krystina Rubino
Переглядів 15910 місяців тому
Recast's Marketing Measurement Coffee-Break Session #7 - Krystina Rubino
Recast Feature Demo: Modeling Spikes (Holidays and Promotions)
Переглядів 16710 місяців тому
Recast Feature Demo: Modeling Spikes (Holidays and Promotions)
Recast Feature Demo: Handling Contextual Metrics
Переглядів 9210 місяців тому
Recast Feature Demo: Handling Contextual Metrics
Recast Feature Demo: Channel Level Deep Dives
Переглядів 12810 місяців тому
Recast Feature Demo: Channel Level Deep Dives
Recast Feature Demo: Last Week's Performance Snapshot
Переглядів 12410 місяців тому
Recast Feature Demo: Last Week's Performance Snapshot
Recast Feature Demo: Handling Lower Funnel Channels
Переглядів 14910 місяців тому
Recast Feature Demo: Handling Lower Funnel Channels
Recast Feature Demo: How We Do Holdout Testing
Переглядів 29710 місяців тому
Recast Feature Demo: How We Do Holdout Testing
Modern Analytics for Marketers (#3): Holdout Accuracy and Backtesting
Переглядів 39011 місяців тому
Modern Analytics for Marketers (#3): Holdout Accuracy and Backtesting
Great conversation!
15:14 offers a great explanation of why branded search and website visits do not contribute effectively to marketing mix modeling
Yes!
I'd never heard that story about Nike and Apple before. Almost identical situations. Explains so much
Amazing Content
21:27 has some great resources to learn more 32:35 goes over how to create your own DAGs
Very useful information.
Promo sm
Can you share the blogpost referenced at the video’s timestamp 19:30?
Here you go! www.linkedin.com/pulse/more-precision-mmm-experiment-calibration-robyn-from-meta-gufeng-zhou/
Is there anything that can be done for MMM models to better account for upper funnel activities?
Yes! You just need to model the relationships explicitly. You can read more here: getrecast.com/lower-funnel-channels/
Peter Fader's book "The Customer Centricity Playbook" sits proudly on my bookshelf. Thanks for bringing him on!
Love the way you guys aim to simplify MMM in actionable ways
Glad it's helpful!
fantastic 👏
Very good talk, Michael, thanks for sharing!
Thank you so much!!
The positive and negative points presented in the video are very interesting. I signed up for Meridian's early access list to get access to the code. Do you know if Meridian uses the Fourrie transform to give some kind of weight to trend modeling?
Unfortunately no details yet on any of the underlying transformations. Will be very curious to see / hear once the code is released!
I am excited for this! Eagerly waiting for more episodes!
Also offline channels like Tv, Radio, print are mostly used as Awareness campaigns where as meta and google are used for conversion so how do we differentiate these effects?
We'd recommend to treat them the same!
For Online channels like meta, google we have spends and relevant impressions to check if there is any incremental sales. But for Offline channels like radio, flyers, planograms etc we dont track impressions so jn this case how do we know the incremental effect? or is there any factor we took into mmm model?
Impressions can't tell you if there were incremental sales! Just because there was an impression and then the person converted doesn't mean that the impression *caused* the conversion. We generally recommend using marketing *spend* in an MMM
*promosm* 💖