After joining Upsight in late September, Nir Amram quietly became one of the most interesting people in the office. A trained physicist, his career history includes everything from military cryptography to experiments with the world’s largest particle accelerator. As the newest member of our customer success team, Nir analyzes quantitative data to identify trends in user behaviour for our enterprise partners. We sat down to talk about why, after nearly 15 years in experimental physics, he decided to study people instead.
First off, welcome to Upsight! I hear you come from an experimental physics background. What can you tell us about that?
Nir: I started out doing some work for the large hadron collider: building, testing and doing quality assurance on detectors that are going to be installed for some of the big experiments. Specifically I worked on something called muon detectors, which are the outermost detectors, and are quite important for the whole collider. That got me rolling and really enjoying that kind of work. On top of that, I did pattern recognition analysis on the detectors themselves to track particles and reconstruct their paths.
From that I moved to CERN to do physics analysis when the LHC started running. I worked on one of the big experiments there, which is called ATLAS. A lot of the work we did isn’t necessarily particle physics, but detector physics, which eventually leaks out into the industry and the public. One of the detectors invented as part of a previous CERN project is now being used for PET scans and MRIs and all that. This is what attracted me. Physics is nice and important, but the benefit for the public I think is a great thing.
And is it that interest in public benefit what drew you away from the theoretical and towards the more experimental side of things?
Nir: I think there were several reasons. Academia, even though it’s very nice, has some disadvantages. What attracted me to the experimental side of things and not the theoretical side is that I like to work with my hands and see the impact of my work on everyday life.
Were you finding academia a little impractical?
Nir: Maybe a little, but I’m just not sure I would have enjoyed more time there. I enjoyed my research immensely, but going forward in that, you end up focussing on things like teaching and administrative work, and that didn’t appeal to me so much.
Nir: (laughs) Grant writing is, I think, the main reason I didn’t want to continue there.
So what are you most looking forward to with regards to your work at Upsight?
Nir: Ah, that’s easy. So, what we did at CERN was interesting because you have so much data, and the data was simple. It described a simple behaviour, so it was easy to analyze. But most of the work was filtering it to the point where you have 0.001% of the events that you were looking at. At Upsight, I think a lot of the work is the visualization of all of the data and trying to get an overall idea of the general behaviour. You have so many different users practicing different behaviors in so many different apps that finding the underlying model to describe so many different things becomes the challenging part. I’m most excited for that. I also come from a gaming background, back when I had spare time, which is part of why I found Upsight so interesting.
What sort of games did you play?
Nir: I played a game called EVE Online. I developed some tools for users and then ran tests to analyze their behaviour and see what their behaviour said about them.
That’s super interesting. What sort of tests were you running?
Nir: EVE is about space exploration and galactic trade, so I made tools that analyzed trade routes for monetary potential. I also tracked high-level targets. There were people flying very expensive ships and we were all very keen on killing them. (laughs) You could try and track them to see where they went, but more importantly you would need to track their friends, and where their friends went. That would tell you where they were going to go.
The last thing I was doing before joining Upsight was spy hunting. We had a certain amount of spies in our group of 600 people. We had to understand, based on their behaviour and some information that was leaked, out of 600 people, who the spies were? And that was very challenging.
How did that go? Were you able to weed them out?
Nir: We did. It was actually simpler than we thought. Our group used [a chat service]. By recording the people that log into that service and then log out of that service, the times that they do that and for how long they stayed, that allowed me to determine who were the most likely candidates, and that allowed us to reduce the number from 600 to 4. We kicked two of them, saw that it wasn’t them, and then found out that the third one was our guy.
What advice would you have for any publishers that maybe don’t have your extensive experience but still want to leverage data analysis to produce results?
Nir: The first thing I would do is look at two behaviours that I think are every close together and try to correlate them and see, is there a correlation? And if there is, ask “Is this the correlation I expect?” And if not, “Why Not?” After I do that, that gives me an idea about the interconnected events or behaviours that I could see in the game and I would try to expand that. So, there’s a correlation between these two things, but what about this other thing? Is it close? Is it related? Then I’d run some rudimentary analysis. Some supervised learning algorithms or simpler things to classify people to try and see if my assumptions are correct. I would do that because I’m curious, but to get from that to actionable insight, that’s the bigger question.
A lot of the driving force for me is that these behaviours are genuinely interesting and I’m curious to see how they behave. I think a lot of us physicists are like that. My supervisor once said to me, “It’s challenging for physicists, because their profession is also their hobby. They will spend days and days on one thing because they really want to understand how that thing works.” You need to be careful not to burn out, because of that.