I’m at An Event Apart in Boston where Jared is kicking things off. He’s pinch hitting for Eric who can’t be here, alas. (Do you like the way I’m casually using baseball references like
pinch hitting to ingratiate myself with the local audience?)
Jared’s talk is entitled Revealing Design Treasures from The Amazon and I’m guessing he’s not talking about the South American river. He begins by talking about milk. Two years ago, you could buy Tuscan whole milk on Amazon. The reviews are hilarious. Jared reads an over-the-top literary travel piece to everyone’s amusement. Another review is written as a romance novel. Another is written as Coleridge’s Kubla Khan. Another is in the style of a rock music review. Some of the shorter ones are hilarious too,
Okay product, but you have to buy a glass to use it.
Here’s a comment that comes up at lots of planning meetings,
I like the way that Amazon does this, why don’t we do it like Amazon? People don’t say that about QVC. What’s so special about Amazon? Well, for a start, it’s a very, very popular and successful site. Jared quotes the stats. But there are other little things that are almost invisible but are very appealing. For example, the search results tell you when something will be delivered—most sites provide this information on the product page, but not on the search results page. They also keep iterating lots of subtle little things like the
add to shopping cart button. So it’s understandable that people want to do things like Amazon. But to do that, you need to know exactly what it is that Amazon does. Jared will now reveal all. Jared has spent a long time watching people shop on Amazon.
It all starts with the content. In the case of Amazon, that’s reviews. People read reviews on Amazon and then sometimes buy the product somewhere else. Amazon knows this and they’re okay with that. Jared compares two reviews of a Harry Potter book. One talks about the content of the book, another talks about the experience of getting the book delivered from Amazon.
As more and more reviews get added over time, quality reviews get pushed down the page. This isn’t good because reviews are so critical to purchasing decisions. Amazon solved this problem with a small, simple device. They added a little question,
Was this review helpful to you? This small addition gained Amazon approximately 2.7 billion dollars in revenue. That vital little feature was rolled out without much fanfare. With that extra data, people can now view most valuable reviews, least valuable reviews, etc. It turns out that the only people who want to see reviews ordered by date are the author of the book and the people who wrote the reviews. It’s just not that valuable for customers.
For a lot of products, such as alarm clocks, you’re only going to write a review if you have a negative experience. How does Amazon get people to write reviews? Most people don’t leave reviews. About 0.7% of people who buy something leave a review. But because Amazon has such a huge amount of customers, that equates to quite a lot. So the next time someone says,
we should have reviews; that works really well for Amazon, you can respond with
sure, we should have customers too; that works really well for Amazon.
Frankly, review writers are the lunatic fringe of customers …and you need to encourage that lunatic fringe. This is what the discussions, listmania, and
customers ultimately bought.. features try to do. Some of these ideas and experiments work but some of them don’t.
Remember the Amazon gold box that would wiggle at you from the top of the page? The idea was to show people products in a “treasure chest” to encourage people to buy those products. The goal was to let people know about all the other stuff that Amazon sells; they already know from your shopping history what you know about so they specifically showed you stuff you didn’t know about. People got really frustrated with this. People wanted to see things they were likely to want to buy but that wasn’t the goal of the gold box. After a while, people stopped clicking on the gold box.
Amazon added tags. Two of the most popular tags are
dvd. Those aren’t very useful for navigation. Neither is a tag like
not interested but people use that tag a lot.
One of the products Amazon sells is an ethernet cable selling for $500. Boing Boing picked up on this. People starting havig fun with the tags for the ludicrously overpriced product;
waste of money, etc. Well, try clicking the
waste of money tag; there’s a lot tagged with that. Why would Amazon allow people to tag their products in this way? The tag
defective by design is a protest tag for anything that uses DRM.
Here’s the lesson. If your people are saying
we should be more like Amazon, that should really mean
don’t fear new ideas. Experiment. Some experiments don’t work. Know when to drop the ones that aren’t working; you’ll need a good metrics system to know what’s working.
There are also lessons to be learned from the experience of using Amazon. Notice that they don’t ask you to log in with your password until you absolutely need to. That really matters. Some customers never need to give their password. Amazon has amazing security levels that they’ve put a lot of effort into:
- Amazon doesn’t know who you are (no cookie).
- Amazon knows you from a cookie—you can receive recommendations.
- Amazon wants to reveal something only you should know (password required).
There was a lot of negotiations with credit card companies to make the purchasing experience as good as it could be.
This is all about increasing Goal Time and reducing Tool Time. Struggling with security and remembering passwords is Tool Time. Finding the product that’s right for you is Goal Time.
Every time Amazon changes something, they are dabbling in changing the Tool Time. Amazon has changed a lot of over time. But most people don’t notice because the redesign happens slowly over time. This is in stark contrast to, say, Facebook’s sudden redesign. People don’t like it when things change suddenly. Amazon doesn’t have that problem even though it is constantly changing. Most users didn’t even notice when the mega dropdown was replaced with link list navigation. Here’s how they rolled it out:
- Show the new design to 5000 non-cookied visitors per day. That means switching on the new design for 45 seconds. These non-cookied visitors are the least risky; they haven’t visited Amazon before.
- After three weeks of that, show the new design to 1 in 5 non-cookied customers.
- After another three weeks, show 5000 cookied customers the new design.
- Show 1 in 5 cookied visitors.
- Show everyone.
That’s twelve weeks to roll out one change.
Search can be hard. How do you find the first Tom Clancy book to feature Jack Ryan? How do you find an inexpensive but high quality SLR camera? How do you find a good toy for your six-year old niece? How do you find all the novels by Nobel Prize winning authors?
Let’s say your new to Salsa music and you want to get the best Salsa artists. First you have to limit your search to music to avoid getting food products. Even then, you get greatest hits albums but you don’t know who the artists are. CD Baby handles this better than Amazon because they have curated content.
Finally, never forget the business. Jared will now share the secret of Amazon’s business.
You can buy an iPod nano on Apple, Best Buy, etc. for about $149. Amazon sells it for $134. That’s probably cost price. It turns out that Amazon can sell almost everything at cost price and still make a product because of volume. It’s all down to the Negative Operating Cycle. Amazon turns over its inventory every 20 days whereas Best Buy takes 74 days. Standard retail term payments take 45 days. So Best Buy is in debt between day 45 and day 74. Amazon, on the other hand, are sitting on cash between day 20 and day 45. In that time, they can invest that money. That’s where their profit comes from.
You have to start with a great business model to produce a great experience.
Jared leaves us with some homework. Visit the Amazon page for the Playmobil security checkpoint. Let that be a lesson to us.
- Be careful when emulating features.
- Some experiments don’t pan out.
- Not every use case is the same.