An Event Apart: Design Treasures from The Amazon

by December 8, 2009

Jared Spool's Revealing Design Treasures from The Amazon presentation at An Event Apart in San Francisco highlighted some great design lessons from Amazon's Web site.

  • Amazon used to sell Tuscan Whole milk no longer does but the site still keeps all the user reviews. 1,025 reviews are available.
  • Amazon is an attractive target for people in the design space: “do it like Amazon”
  • Why? It’s the 8th trafficked site on the Web, 71M visitors in December 2008. Made 19 billion dollars in 2008 –up 29% over last year. Amazon ranks 3rd in customer satisfaction ratings
  • Amazon has 76 million customers since they launched -17th largest population in the world. In 2008, processed they 175 million orders.
  • All search results and product pages on Amazon tell people when they will get a product. This is a crucial point of comparison to going to the store.
  • Amazon never changes because it is always changing. Subtle progression of design over time.
  • Four things Amazon is really good at: engaging through content, don’t fear new ideas, eliminating tool time, never forget about the business
  • One: Engaging through content.
  • Originally, Amazon reviews were stored chronologically. So latest was always on top. Fixed this problem with “was this review helpful?” question. Amazon community now curates reviews. This feature has been worth 2.7 billion dollars to Amazon (through derivatives research).
  • Lots of people go to bad reviews first to convince themselves not to buy something.
  • Be careful when you emulate features. Target went out and bought the Amazon reviews platform. Their reviews are either missing or very negative. Why?
  • When the new Harry Potter book came out, Barnes & Noble/Wal-mart got 300+ reviews. Amazon got 1,805. Target got 3.
  • On Amazon, one out of 1,300 purchasers will write a review. To get 20 reviews, 26,000 sales needed. If you have a 2% conversion rate (typical for e-commerce), you need 1.3 M customers for 20 reviews. That’s why Target has so few reviews.
  • Two: Don’t Fear new ideas.
  • On an Amazon product page: bundles, look inside, customers who bought carousel, what do customers ultimately buy, discussion forums for products & reviews, lists, guides, etc.
  • Not all experiments work. Amazon gold box tried to introduce you to products in categories you did not use before. But most people felt it was not relevant. Tags added to site but not that useful.
  • If the average customer spent $5 more per order, it would make Amazon an additional $875 million. So experimentation can really pay off.
  • defectivebydesign is a protest tag for products with DRM technologies. One of the most popular tags on Amazon.
  • Companies that take risks make great products. Companies that don’t take risks make crap.
  • Three: Eliminate tool time while delivering confidence.
  • Understand where people spend their time. Amazon shows you complete set of options before it asks for your password. Users are not willing to commit to things until they know enough information.
  • Level zero: Amazon does not know who you are (no cookie)
  • Level one: Amazon knows you from a cookie (recommendations & one-click)
  • Level two: need to login to change addresses, etc.
  • Goal time vs. Tool time. Goal time is making the outcome of the experience better. Tool time is when user is trying to move forward without any improvement in the outcome of the experience.
  • We always want tool time to be better.
  • Amazon does really small incremental changes. No one wants the site redesigned except for the owners. 97% of facebook users said they did not like the latest Facebook redesign.
  • Amazon navigation phase in plan. First gave it to non-cookied people. First did 5,000 per day. Then 1 in 5. Then moved to cookied customers. Whole process took 12 weeks.
  • Not every use case is the same. Amazon does a good job for specific names of media products. 70% of their revenue comes from media products. But there are many things that are hard to find on Amazon.
  • Four: Never forget about the business.
  • Can buy an iPod at Apple, Target, Best Busy for 149 dollars. Amazon sells it for 133 dollars. Can sell everything at cost and still make money.
  • Negative Operating Cycle: Amazon turns (sells out) its inventory every 20 days. It takes other retailers 74 days (Best Buy) Standard retail payment terms are 45 days.
  • On day zero, Best Buy orders inventory but has to pay on the 45th day. From the 46th day to the 74th day, they are in cash debt and often have to borrow money.
  • Amazon still orders on day zero and pays on day 45. But they are sitting on cash until they have to pay the bill. This means they earn interest on the money they have in the bank. So Amazon floats cash vs. cash debt.
  • Amazon has designed a business process that really works. To be involved in strategic decisions, you need to understand this kind of information.
  • Be careful when emulating features –some experiments will not pan out and not all use cases are equal.