Had one session of phone screening and five sessions on site in one day earlier this year. Every session lasted 45-60 minutes. Passed the phone screening with very positive feedback but failed at on-site stage with good feedback on two skills and bad feedback on another two skills (the recruiter said the ‘places for improvement’ instead of ‘bad’). I do not want to specify the location but let me just say it was outside of US. All interviews were still done in English. Every session was one vs. one and every interviewer kept typing while I was answering, which you should assume in any Google interviews. This is because they were asked to give complete record of what I spoke for the hiring committee to give final decision. I was not sure if I should slow down while they were typing hard. I would ask them whether I should slow down if I could have a next chance since one of the feedback was "my communication could be more structured". Although there were differences in the personalities including a good/bad 'chemistry' to mine, everyone was nice and responsive. Any of on-site sessions were designed to mix the skill sets to test e.g. stats, data insight, coding, and communication, which teased my brain a lot. Every interviewer asked me 3-5 questions, around 20 questions in total. One question was in my back pocket already from googling “data scientist interview questions”, most of others were new but somewhat similar to them, and a few were something I never imagined. Still, the preparation based on what we can find through googling helped me a lot in answering their questions and was definitely worth doing. I knew I made one obvious mistake in stats question (which was not easy in general but I should have answered correctly because I had ever done that and even put it in my resume; also I have 10+ years of experience in applied stats). Other than that, my overall after-math impression was “my answers should be fair enough” and hoped one mistake in 20 questions did not a lot harm. In reality they pointed out not only that I did not do a good job in stats questions but also that I lacked one other skill. I guess that error in stats question was not the only reason of the consequence. As we know, they may offer another follow-up session if the candidates are close enough to the bar, which I didn’t receive. One thing which I feel a bit surprised was that every stats question came with simulation coding exercises. One stats question was obviously easy to answer in an analytical solution but they requested the solution in simulation coding (I checked with the interviewer if it was). Some other stats questions asking “probability of the event” was impossible to answer analytically and even difficult by simulation; I could not answer well on site but probably I should have defined Classes and class instances to simulate the situation (I assumed Python). Sometimes they asked in a way that “how do you get the probabilities of xxxxx” and ended up with the simulation solution, other times they directly asked “how do you code the simulation model to get the probabilities of xxxxx”. As I saw somebody else posted about Google interview experience, there is no guarantee that what the recruiter promised in advance really comes true. In my case, there were a few differences from what the recruiter had told me. 1. There was no break between the session (except the lunch break), the next interviewers were just waiting outside the room or accidentally called in before the prior session ends. Literally no break even for a single minute or water break. I might have had one if I had asked for, but it looks like every interviewer had a predetermined set of questions they had to finish during the session. They were always in a little rush. 2. The recruiter told that the coding test would be about data wrangling, but it was more like general one like list handling or a lot of simulation ones. Let me skip the details here because I already put enough above. 3. Recruiter said I would code on PC with my choice from R or Python on RStudio or Jupyter (or whatever other platforms you like), but actually I was not allowed to use my own PC and the actual choice was Google doc on office Chrome book, white board coding, or even brain coding when neither of two was available. Either way, they did not care the quality as a code. I made a lot corrections on the fly and wrote some lines in math forms but still had a positive feedback about coding skill. Overall, it was my first Google interview and though the consequence was not something I expected, I am proud of my own performance. They say there are many Googles who were hired at second interview opportunity or even later. Also, the luck does some game in any life opportunities; Google interview is not the exception. I recommend anyone with experience which can be qualified with their requirement giving a first shot. Good luck!