Tuesday, June 21, 2016

Book Review: Sprint by Jake Knapp

I read a book! OK, I read a lot of books, but I also like to pretend as though I'm marginally literate in light of my librarian wife who reads many more books than I do.

The book was "Sprint: How to Solve Big Problems and Test New Ideas in Just Five Days" written by some guys at Google Ventures (GV), a part of Google that buys/invests in startups and then works with the startups to grow into successful companies.

This book is essentially a business book, but I think it can be useful for data scientists especially in product/project development.  The book details the process of taking a "sprint" week where a company focuses on making very rapid design and product improvements by exclusively focusing efforts on a specific project.  I think this kind of focused "sprint" has quite a bit of potential in algorithm and data-science product development work, although the prototyping and testing phases might look quite different.  I may try this one my own, and see what happens, here are some positive and negatives:


  • Quick to Solution: The first three days are spent working through ideas and solutions, day four on prototyping, and day five with a real-life user test.  This means that within the week you both solve the problem conceptually and create a prototype-in the data science world, it's likely the code from this prototype could be somewhat easily ported to production.  
  • Early Results: The last day of the Sprint is a test of your prototype with users, which gives you a quick read on how(if) it works and what improvement are needed before going into final production.  I like this because the process provides and early insight into how end-users will see the product, and an ability to get an early gauge on its value.


  • 1:00 Lunch: The book recommends (for a few reasons-read the book) waiting until 1:00 to eat lunch.  I eat lunch at varying times and sometimes just grab a couple of sticks of string cheese and a bag of chips between meetings. But many of my coworkers have to have lunch at the same time every day, and in fact get quite agitated if they are forced to miss their normal lunch time.  Additional hypothesis-people in central time zone tend to arrive at work earlier and *culturally* eat lunch earlier, there's at least some evidence from 538 to support this, showing the only central time zone towns on the "late arriver" list are college towns.
  • Costs: Sprints can help get your product to market quickly, and fix obvious issues, but devoting a week of time from multiple employees to a singular process (with a lot of meetings, btw they suggest a no-devices-in-meetings rule.. booo)* can be costly.  Ironically, the companies that can most benefit from a Sprint are likely those who can least afford it.

*Side note: I sometimes like to on-the-fly calculate the costs of meetings I attend by estimating salaries, backing in to hourly rates, and the multiplying. My second day in my career I was placed in a 3 day meeting (7 hours a day) with 10 people who averaged $37 an hour in salary ($49 ish including benefits).  Here's what that cost of meeting looks like (this is when I worked for government, private sector meetings tend to look more expensive due to higher salaries):
Without benefits = 3 * 7 * 10 * 37 = $7,785.
With benefits = 3 * 7 * 10 * 49 = $10,380.