Review: Persuasion Analytics and Targeting with Ken Strasma

election

tl;dr Great class, learned a lot!

Persuasion Analytics and Targeting taught by Ken Strasma, covers campaign micro-targeting theory, and application using real (anonymized) campaign data. Using actual camapign data, including candidate and partisanship IDs, demographic data, and commercial data, really made this class great (even for an old political hack like me). In the class you build partisanship, candidate, and persuadability models. Additionally, there is great material on evaluating models, measuring impact, and strategy of how and when to use different models during a campaign.

There are tracks available for all levels of technical ability including an Excel track for track for the less tech savvy along with an R/Python track for more technical audiences.

Week 1
Targeting and voter contact basics.
History of modern micro-targeting.
Basics of why campaigns should use targeting.

Week 2
Microtargeting vs traditional geographic and demographic targeting
Survey and experiment design
Predictive modeling including linear regression, logistic regression, and decision trees
Build and evaluate partisanship models and voter turnout models

Week 3
Build candidate ID models and persuasion models
Measuring lift in predictive models
Deriving predictors from voter file data
Create and compare ensemble models

Week 4
Targeting for TV, online and social media
Applying previously created models to build a direct mail program
Campaign strategy using micro-targeting

There were some great discussions about polling and applying micro-targeting more effectively in campaigns. Ken Strasma is suprisingly active on the message boards with advice and help on class problems. It was fascinating to get his perspective on the recent election and the associated polling fiasco.

Pros:
Lectures are interesting
Covers a lot of topics
Practice with good examples
Instructions for Python, R, or Excel
Replies quickly to the boards

Cons:
Could have used more feedback on the modeling process
Never quite got used to the forums for conversations with classmates
TV and social media is light on practice despite good background info

FYI – The lectures are in PDFs. I like to read when learning but I can understand that not everyone does.

Rough drafts of my Jupyter notebooks from the class are on Github here.

Overall, I highly recommend the class! It was good to have a chance to learn from the master.


Comments

comments powered by Disqus