First there was a recruiter lead interview, which was very short because he was outside after a meeting, supposedly anyway. Took about 10 days after that to get an interview scheduled with the lead. Very slowwww response times at this point..., sure she was apologetic, but there's only so much days of non-response you can take before you think they are flaking out.
After that was a test, which is when response times got faster. The test contained 5 questions, 3 SQL questions and 3 math questions.
I don't remember all the 5 questions, but I'll try to summarize them:
Given a beauty product that claims it reduces heart disease and it is retracted 6 months later, what is the most likely reason?
Given that you are in the future with all information on people in 2012, how can you estimate human value?
Given that players are more likely to spend money while being in a guild and we upgrade the guild experience, what are the reasons for this?
What 5 statistics about Guild Wars 2 do you want to know?
I don't recall the last one.
Honestly, I have no idea what these questions were testing. The "in the future" one is very much superfluous. Wtf is human value? I wanted to can this test after answering those five questions as anyone could answer them with LOGIC.
The SQL questions were just that, SQL queries and one optimization query (extra credit) referring to the use of temp tables, distinct and count. I don't know why the heck anyone would use a temporary table instead of a subquery, which automatically does that. Anyways, I figured it would be better to count first then filter instead of creating a temp table, which is the filter, then use that filter as we are counting. The table large and given the fact that the prediction algos in the cpu would fail given the filters I would assume count 1-5 million row with a filter would be slower than just counting without a filter, but what do I know? I should have stated that exact assumption I guess.
The cpu prediction algos tell the cpu which branch to take given an if statement, so I would think counting in a big table without a filter would be quicker than with. I assume there's not as much CPU stall during the count without filter hence why it would make a difference. Wtf do statisticians know about Computer Architecture anyway.
There were 3 math questions: 1 systems of linear equations, derivative and integral. I couldn't tell what the question with the systems of l.e. was asking at the time as poorly worded as it was. Simply could have said these next two equations are part of a system of linear equations please find x and y or even find the intersection of these two lines described by the two equations instead of posting two numbered equations with " find x, y", which I misinterpreted understandably as find an x, y in these two lines. the only challenging question was the optimization query because it was on paper rather than having a cpu to test it. Poorly worded and poor test for finding suitable candidates.
I was rejected based on 'mastery' (w/e that means) after the test, which is BS in my opinion, considering how fit the position and company's position is to my background, but what can you do when these people know nothing about you or your dedication. Less risk in hiring people that you know I suppose... I was hoping to talk statistics (being a data analyst and all) in an actual face to face interview to show I know what I'm talking about and therefore lower the risk. I think they made a significant prediction error in their decision, which is fine for me because I don't believe I will be applying again (a big f-u). Reject me once and you've lost my overall interest and excitement and believe me the prospect of working for a games company is exciting for me as long as I'm not programming (it's very specialized). ;]
There is no monetary damage from posting trivial questions online therefore I will not be removing them regardless of your NDA.