Notes from two articles about unemployment insurance (UI) benefits during COVID-19 in the US. The two articles center around the fact that Congress paid an extra flat $600/week in UI benefits on top of state UI benefits. This results in workers receiving more in total UI benefits than their previous salary.

How frequent is this phenomenon? Does it discourage returning to work? Is there a better way?

Notes from a Holistics.io book on setting up analytics and BI tools.

This post is my notes on a white paper by Michelangelo D’Agostino and Katie Malone on managing data scientists. Overall, the report is a good introduction to the basics of managing DS, providing a bunch of generally-applicable guidelines. Note however that both authors come from the same company (Civis), and are relatively junior managers (5 and 3 years of management experience respectively).

This is a list of data-related jobs I think are likely to develop in the next 5 to 10 years.

Hay Day is a mobile free-to-play farming game released in Summer 2012 by Supercell. These are notes from 2015, when I played the game for a couple months. I first go over the core mechanic of the game, fulfilling orders. Then I cover two of its very constrained, yet elegantly-designed, multi-player interactions: the Daily Dirt newspaper and neighborhoods. Finally, I’ll have to mention some less elegant parts of the game: manipulative monetization, and grinding.

Sampling is often necessary for heavy algorithms to run on large populations. Sometimes, we want to draw some elements more frequently than others. For example, we could draw US states proportionally to their population. In this case, we could use probability-proportional-to-size (PPS) sampling. This post provides an efficient SQL script to sample any number of weighted elements via PPS sampling.

Redshift is a distributed columnar database. Columnar databases scale easily by distributing data between nodes in very specific ways. This typically comes with a cost: joins are expensive, if not impossible. Redshift is no exception, although it allows a distribution key by which to spread records. This way, joins on distkeys do not require records to be shuffled between nodes, and are somewhat cheap. Surprisingly, however, joins on distkey and another column currently require a shuffle.

Let’s imagine a user browsing Facebook’s Newsfeed. This post describes how I would track basic Newsfeed user activity, and design supporting schemas in Redshift.

App users often go through processes involving several steps, like browsing > adding to cart > checking out. We want to measure the typical duration of each step and of the whole process. Averaging durations naively (dividing sum by count) skews towards users who take an abnormally long time to go through a step. What about using percentiles instead?

This post is an introduction to tracking user behavior on apps and websites. It focuses on how Netflix tracks carousel loads and user clicks on movie cards.

The Great British Bake Off (GBBO) is a 10-episode show that started in the UK in 2010.

Shop Heroes is an item-shop game released on mobile and Kongregate in mid 2014 by CloudCade. The gameplay consists of crafting equipment for heroes, who then go on quests to find materials, which are used to craft more equipment. Players can gather in Towns to share the costs incurred by leveling up buildings, which unlock more heroes and quests.

On January 13 2016, the Powerball lottery made the news by reaching a jackpot of $1.5B, the largest in US history. Lotteries are interesting because they are a gambling game used by governments to raise funds without raising taxes. Let’s have a closer look at the rules, odds, and profits made by lotteries, using Powerball as an example since it’s one of the lotteries with highest sales.