Two ways to look at it.
- Online selling of books has catered to people who have clarity on what they want. You come to buy 1–2 specific books and go. Discovery and commerce are deconstructed. In physical bookstores, discovery is inbuilt with the act of purchase. (i.e., Jobs are different: I want to buy vs I want to explore; ‘retail as logistics’ vs ‘retail as discovery / leisure’ — this from Benedict Evans). So you could say Amazon is trying to cater to a new job to be done. But I dont think this is the entire story.
- Amazon clearly wants to sell more books online, but there is only so much that online machine learning can teach it. Presently its recommendation algorithm is weak — because it is only getting data from pre-decided buyers, not exploring buyers. What if the machine is fed with data from purchases made by explorers as well as data such as do customers buy books next to each other? etc This new data that goes into the machine will enhance it dramatically, and help it make better online recommendations. Keep in mind A lot of Amazon’s efforts are aimed at removing barriers to purchase such as Prime, same-day delivery etc. (If you could get it now (such as with physical stores), would your online purchase change dramatically?). Similiarly if you can get a better recommendation online, you will buy more books. Therefore I see this physical book store as a channel to get discovery data and add it to the machine and enhance it to sell better online. A lot of the initiatives by online retailers such as Warby Parker, Bonobos can be seen in this light.