Solving Real-World Business Problems with Bayesian Modeling

on YouTube ↗42m

" summary

Announces PyMC 4 with JAX and GPU sampling and shows concrete Bayesian builds: a CAC estimate that yields 25% probability a channel is profitable, Gaussian‑process time‑varying and hierarchical time‑varying process models that handle saturation and adstock, and integration of lift‑test and voucher data. Compares against Google/Uber approaches on HelloFresh data and reports the PyMC model wins; full models run in minutes. Pick this if wanting code, priors, diagnostics, and an actual deployed media‑mix model with forecasts and what‑if simulations.

AI-written summary · about

" chapters (29)

1. 0:00Welcome!
2. 1:15PyMC Labs- The Bayesian consultancy
3. 2:39Why is marketing so eager to adopt Bayesian solutions
4. 3:49Case Study: Estimating Marketing effectiveness
5. 6:00Estimating Customer Acquisition Cost (CAC) using linear regression
6. 7:36Drawbacks of linear regression in estimating CAC
7. 10:02Blackbox Machine learning and its drawbacks
8. 11:27Bayesian modelling
9. 11:52Advantages of Bayesian modelling
10. 14:12How does Bayesian modelling work?
11. 16:53Solution proposals(priors)
12. 17:26Model structure
13. 19:57Evaluate solutions
14. 20:16Plausible solutions(posterior)
15. 22:36Improving the model
16. 23:38Modelling multiple Marketing Channels
17. 24:51Modelling channel similarities with hierarchy
18. 26:13Allowing CAC to change over time
19. 28:00Hierarchical Time Varying process
20. 30:05Comparing Bayesian Media Mix Models
21. 30:47What-If Scenario Forecasting
22. 31:53Adding other data sources as a way to help improve or inform estimates
23. 33:00When does Bayesian modelling work best?
24. 33:35Intuitive Bayes course
25. 34:38Question 1: Effectiveness of including variables seasonality?
26. 36:03Question 2: What is your recommendation for the best way to choose priors?
27. 38:16Question 3: How to test if an assumption about the data is valid?
28. 39:07Question 4: Do you take the effect of different channels on each other into account?
29. 41:33Thank you!

" related talks

• 01h 26mProbabilistic Programming and Bayesian Computing with PyMCFonnesbeck, Wiecki · PyData London 2024
• 42mBayesian Non-parametric Models for Data Science using PyMC3Christopher Fonnesbeck · PyCon 2018
• 30mIs your marketing effective? Let Bayes decide!Emanuele Fabbiani · PyData Paris 2024
• 43mProbabilistic Programming in the Real WorldZach Anglin · PyData DC 2018
• 01h 27mBayesian Time SeriesChris Fonnesbeck · PyData London 2025
• 41mPyMC-Marketing: Customer and Marketing Analytics the Easy WayChristian Luhmann · PyData NYC 2024
• 39mProbablistic Programming Data Science with PyMC3Thomas Wiecki · PyData London 2016
• 34mFast, beautiful and easy Bayesian modelling: Can you have it all?Rhydwyn McGuire · PyCon Australia 2015