Keeping Up With Data — Week 36 Reading List
5 minutes for 5 hours’ worth of reading
What drew my attention to the image above was the latest Analytics Engineering Roundup. Having built a couple of data teams myself, I had to chuckle (a few times) when I saw this Emilie Schario’s tweet.
I haven’t had as much time for reading this week as I normally do — for family reasons 🤰 ➡👶! But there were still a couple of articles that I want to highlight.
- The 18 Generally Agreed-Upon Information Principles: Doug Laney starts his book Infonomics by telling a story about how data got disallowed by the GAAP and IFRS rules. But just because data can’t be explicitly listed in the books doesn’t mean that it’s not relevant for the P&L. Doug is one of the thought leaders in the space of data (and information) valuation and he has kicked off a set of 18 GAIP principles this week. My favourite ones are #9 and #13. I still have to wrap my head around #18. I’m just not sure why Doug got so impatient and published the list on Thursday and not on Saturday given the role of 9/11 in the IFRS rules when it comes to data. (Doug Laney)
- How to monetise your data to fuel growth in your business: Chat with Bill Schmarzo: Bill says that “you can’t determine the value of data in isolation of the business.” This goes to the very heart of economic value of data. “The more the data and analytics get used, the more […] valuable they become.” The marginal cost flattens; the economic value growths. But it’s easier said than done. What are Bill’s best practices? What to avoid? Read this article, read Bill’s book The Economics of Data, Analytics, and Digital Transformation or follow Bill on LinkedIn. (Hyperight)
- Reinforcement Learning for Recommendations and Search: Eugene is great at communicating complex concepts in an easy-to-follow way. In his latest blog posts, he talks about downsides of conventional recommendation systems and how reinforcement learning could help. You often don’t want your engine to prioritise short-term value on the expense of the long-term reward. Nor do you want to overemphasise item popularity, which can easily happen when you are estimating preferences based on the available historical data. The articles present the various approaches reinforcement learning offers for recommendation systems including bandits, value-based methods, and policy-based methods. (Eugene Yan)
And that’s it — as the shape of indifference curves between reading/writing and sleeping has changed drastically this week!