I'd suspect not a lot. The number that gets tossed around is based on the whole year's sales divided by downtime - but that fails to account for seasonality.
The retail market in the USA can be accurately summarized as "slow trickle slow trickle slow trickle slow trickle MASSIVE DELUGE OMG OMG slow trickle slow trickle..."
It would shock people how much consumer spending is shoved into the November-December time frame. Since this downtime is in the middle of May, I don't expect them to lose much money at all.
Also, a lot of the "lost" orders during the down time will actually just be delayed. I imagine most people will just try to place their order again tomorrow
That depends too. I suspect at this time of year it's true - during the holidays I wouldn't bank on it. When people are gift shopping downtime easily means the difference between a sale and the customer jumping in their car and running down to the Toys 'R Us.
Downtime is a Big Deal during the high retail season, and not really a big deal outside of it.
I, honestly, doubt very much. I'm sure you'll have a percentage of potential customers not make a purchase at all because the site is down, but everyone that is committed to buying whatever it was they were going to look for will just wait until the site is back up.
I recall someone claiming to be an amazon developer one of the previous times when this happened claiming that they don't actually lose that much money. (sorry no link, my google-fu has failed me). A small number of orders where people are desperate are lost, but the vast majority of people generally just wait it out and make their order tomorrow instead.
10 years ago we calculated downtime at $100k/minute for the homepage. So when you tried to do an ajaxy thing on the home page and messed it up, even after testing, you had to write an email that calculated how much theoretical money was lost.
And sure, there's some elasticity, but an approximation is handy.
I don't recall the source, but I've seen that the loss due to downtime is negligible, and that customers return at a later time to complete their purchase.
I'd agree for people who are buying stuff they need or otherwise are making considered purchases. I don't believe that they don't lose at least some amount of impulse purchases.
Amazon.com does run on EC2. You can bet that if EC2 had a a complete regional outage (which is extremely rare) Amazon.com would be down.
In general Amazon.com makes extensive use of AWS services internally as well, but it's very well architected to handle service outages without it being immediately obvious to the end user.
I understand this wasn't a general AWS outage, AWS !== amazon.com, and that successful hosting in EC2 requires specific architectural decisions to make things run reliably in face of zone/region outages.
What i didn't realize was that jokes about Amazon being somewhat slow and cryptic with information about outages while they're happening were off-limits.