μ · Methodology

How I work

Every data point in every story is sourced from one of the datasets below. I use a mix of original surveys, GDPR data exports, and public datasets. All raw data is available for download.

Datasets

Dataset	Organization	n	Method
Streaming Listener Panel, Q1 2026	pointofmu desk · panel survey	4,201	Voluntary panel; 12 months of streaming history via Spotify API. Weighted to age + region.
First-Date Receipt Panel, 2020–2026	pointofmu desk · panel survey	11,042	Self-reported receipts uploaded by panel members. Restaurant-tier via Yelp price tier. Outliers >$500 trimmed.
Hinge Match Outcomes, n = 200	pointofmu desk · friend-of-friend recruit	200	Friends shared 90 days of match data including "most compatible" tags. Self-selection bias acknowledged.
NBA Last-Minute Shot Attempts, 2019–2025	Basketball-Reference (public)	2,412	All shot attempts in final 60s of regulation, score within 5. Compared to season-long FG%.
Consumer Price Index, Food Away From Home	U.S. Bureau of Labor Statistics		Series CUUR0000SEFV. 2020-01 through 2026-03.
For-You Page Recommendation Audit	pointofmu desk + 80 volunteer accounts	80	Volunteer accounts logged 14 days of FYP swipes. Inter-account overlap measured by video ID.

My standards

Correlation noted, causation never claimed without a controlled study.
Sample bias is always disclosed in the methodology block.
Outliers are trimmed at the 99th percentile unless stated otherwise.
All charts can be reproduced from the linked raw data.
I round to the nearest whole number or one decimal place.