Data Science: For Fun And For Profit

As early as 1923, Claude C. Hopkins wrote in Scientific Advertising that advertising had reached the status of a science. Almost a hundred years later, data science is still talked about as something new. Lukas Vermeer, who works on experimentation at Booking.com, argues that data science is simply science and that the size of your data has nothing to do with it. You have to be just as careful with 40 rows as with a petabyte.

You'll learn why constantly refreshing your A/B test results will show you a winner where there is none, how people dropping out of a test skew the averages and how a check of the 50/50 split catches it. Lukas explains why a ramp-up can reverse a result through Simpson's paradox, why predicting supermarket purchases is pointless and why conversion rate is only a proxy for success. He closes with the four "Ps" every data scientist should master.