NYC Data #342: Airbnb's LTV Calculations, Surveillance Technology Oversight, ESPN, Aircall, Compass, Misleading Log-Log Charts, Mollick et al on Working with GenAI
Plus, StreetEasy's 10 Best NYC Neighborhoods for Home Buyers
Hi friends: Spring is finally in full swing! I hope you find a way to spend some time outside tomorrow, it should be in the high 70s!
As always, help me keep this space up-to-date: please send me posts, events, and job openings. If you know someone who might enjoy or benefit from this newsletter, please share it with them. [image credit: Yann Arthus-Bertrand]
Good Local Posts
Andrew Gelman & Hanno Böck have gone back and forth on the use of a log-log chart and how it’s misleading or not (in this case, on the relationship between a country’s GDP & electricity consumption). Böck’s post was interesting to me both for the data vis critique and the economics content.
Looking to buy a place? StreetEasy published its annual, data driven 10 Best NYC Neighborhoods for Buyers post.
Ethan Mollick’s lab, in partnership with Procter & Gamble, released a working paper about the impact and effectiveness of individuals and teams using GenAI I expect this will become a big part of the conventional wisdom around how AI can help augment knowledge worker performance, so I think it’s worth a read to really understand what this is (and isn’t) saying. Two charts really jumped out at me:
This was a pretty confident group: about 25% of the teams thought they’d rank in the top 10%! But people using AI were less confident they’d perform well. There are a few interpretations of this I could imagine, but clearly within this group, participants were discounting their abilities with AI, maybe because so few people are familiar with it.
The other thing that stood out to me was the dramatic underperformance in quality by people without a teammate (human or AI). As the author of a solo newsletter, this has made me realize I’m not taking enough advantage of my ability to have an ‘instant editor’, and inspired me to run drafts through an LLM before sending to catch any silly errors that I’m missing. Hopefully this will lead to an uptick in quality!
Upcoming In-Person Events (new listings in bold)
3/21 - 4/6: Corpus: Bodies of Data
3/22 - 3/30: NYC Open Data Week
4/2: Columbia’s Data Science Day 2025
4/15 - 4/16: AI in Finance Summit New York
5/15: AI Summit NYC: The Technology Conference For Non-Tech Professionals
5/15: Data Science Salon New York
5/28 - 5/30: Lifetime Data Science Conference
Open Roles
Squarespace is hiring for several data roles, including a Senior Growth Analyst and a Senior Data Scientist.
S.T.O.P (The Surveillance Technology Oversight Project) is looking for an Executive Director!
Stellar Health is seeking a Lead Data Scientist.
Aircall is hiring a Senior Data Analyst.
Disney / ESPN is looking for a Senior Data Analyst - Subscriber Analytics.
Compass is seeking a Senior Director, Data Science, ML and AI.
Miscellany
The picture at the top of this email is of Yankee Stadium from Yann Arthus-Bertrand, whose Earth From Above I saw in 2002 in Chicago and I’ve been obsessed with ever since. This picture in particular has always spoken to me for some reason, but there are many fascinating ones.
This piece on How Airbnb Measures Listing Lifetime Value was fascinating to me. My team works on LTV measurements, but we don’t run a marketplace and so I had never really considered accounting for cannibalization of other sales when calculating the incrementality of value from a customer or listing. I actually think it’s a bit tricky in that it puts the marketplace at odds with its customers. For example, I imagine there’s a point where adding additional inventory has negative net value to the marketplace, so there’s value in constraining supply. That is probably a bad thing for many non-economic reasons!
Thanks so much for being a subscriber. To see previous job listings (many of which are still open!) and blogs, check out the archive (which has emails from the tinyletter days!). Feel free to forward this to anyone: they can subscribe here: