One should understand that at Booking, the only thing that counts is "conversion". In other words, the ability of converting a pageview into a reservation. Be

One should understand that at Booking, the only thing that counts is "conversion". In other words, the ability of converting a pageview into a reservation. Because that's the part of the process that creates value. Since hotel reservations refer to a real world, where people change their plans, hotels get overbooked, burn down, are not reachable because of strikes, volcano eruptions, etc. etc. there is a huge customer service department. These people are able to cancel reservations, or force availability (after consultation with the hotel). In other words, until people have stayed at the hotel, and the hotel gets the bill for that, it really does not matter whether something was correct or not: the customer department can fix any problems. So the only valid test is really: does it convert? And this is very closely monitored in close to real time. The financial part of the system is thoroughly tested, as Booking.com is a listed company, so the scrutiny is extra severe. > Most developers could push directly to production. This is facilitates ownership and empowerment. You need to be sure that what you're doing, will not break things. If you are, it should be ok to push to production. If you were wrong, then that's ok the first time. Just don't do it again for the same reason: always make new mistakes! This was the situation in early 2012, when I basically retired. Things have changed since then, but I still think the above is still basically true.