Unlocking User Experience: How Complex Data Visualization Transformed the Motoactive Watch

Yeah, uh, so one of the very complex technical, um, things that I had to design was for Motorola. We were designing a watch called Motoactive. So it's, it's there in the market. The first, uh, uh, Motorola was entering into this market of fitness and, uh, music. So while dine, and it's a small device, it's a small watch which you would, uh, just put on Android watch which you would, uh, put on, uh, strapped to your, uh, wrist. So it measures a lot of metrics. It has a lot of sensors embedded, um, which. Measures the heart rate, measures how fast you walk. It measures the, the blood pressure, and likewise, measures a lot of uh uh health-related metrics, matrices. So, uh, one of the, the complexity here was that uh most of the sensors collect the data every second and then, uh, the, after a certain period of time, this data is, uh, moved on into the server. So, the idea was to show a couple of the last 360 days of 1 year, more than 1 year of data on this small device. Um, in a way that people could, uh, uh, zoom in. I mean, suppose as you zoom in, as you go on zooming in, uh, 360 days, uh, days of data or the last 2-3 years of data would just come up in a small page, and then as you zoom in, uh, we had to show a few days of data. I mean you could zoom into a single day and a single hour too. Here, the complexity was that, uh, you had to bring a lot of data from the back end. You couldn't, uh, uh, keep the entire data in the, uh, in the small watch. So as you people zoom zoomed in without, uh. Allowing the user experience to go back, we had to go on streaming the data and then create graphs for it, so we had to, uh, do a lot of approximation for this, create the graphs. I mean, we are, we are, it was a visual, uh, graphs had to be created on this matrices, uh, uh. So creation of such a graphs and then not uh changing these graphs so that they show data from 360 degrees all the way to 12 hours and minutes

that's fine. That's true. So, uh, these are the graphs which are to be drawn, and, uh, from 360 degrees, 360 days, or more than 1 or 2 years, people had, we had to give them an opportunity to go on zooming to 1 hour, 2 hours. So the entire, it, it was very complex, complex because, uh, you had to bring in a lot of data from the back end at the same time you had to approximate a lot of things, uh, approximate the entire, uh, uh, graph itself so that, uh, uh, you don't have to rely too much on the network communications, so. This was a pretty, uh, pretty, pretty complex, uh, project to do, and we figured out that, uh, uh, it had a lot of research components which were being pursued on the, I mean, on the, uh, in, in the university level. So the complexity was about, uh, how do you, um, draw a graph without having too much of a data. So how do you approximate that, uh, that curve so that it looks much more similar, much similar to, uh, having the exact data. So that was a technical complexity which was there, which we solved. Apart from that, there are a lot of other people's, uh. So this was one of the complex projects, technical projects that we have handled, yeah.

And I, I was primarily there in it. I was primarily there. I was the one who, uh, figured out the algorithms for doing it, uh, came up with the wonderful workarounds which were well appreciated by the, um, my management, and, uh, we released it.

Yeah, I was a, I was a project lead here, so I devised, uh, I was the one who figured out how to do it and then um there was another two junior members whom we interacted together and then we uh uh brought the features live.

Sure. So, uh, we went through a lot of, uh, failures while doing so. So, uh, once we started doing it in a much more ideal fashion of bringing out the data. I mean, as the people zoomed in, we started hitting the uh servers and getting the data. We saw that there was a quite a lot of lag. And as we started painting a huge amount of data, the entire screen itself froze off. So there were a lot of technical challenges which we came up and then, so we started. solving it, but once we start, the first challenge was, once you really paint out, say, 360, uh, uh, days of data where every, there's a data of every 1 2nd, you start painting on the small device, device itself freezes off. You are used to freeze off. So we started approximating and in the sense, we started going through the data and then finding out which part of the data could be dropped off. So for example, there's a simple line, then then that all the. Points in the line doesn't have to be taken. The first upper point and the last point would be, uh, taken, and that would be, that would be enough to approximate the entire line. If there's a crust in the trap, then, uh, there, there were certain approximations that we thought were instead of taking all the points, we could just sample a couple of points. So, so our algorithm was successful in sampling it in such a way that, uh, the shape of the graph, was well retained. It was much more simple, which was much similar, uh, to having all the data points and then pointing up and then painting it. It's solved a lot of things like, uh, first thing is that, uh, putting in all the data, the huge amount of data on the, on the UI which would freeze it up. That that problem was solved. Second thing was, uh, suppose that as a person goes on zooming it, then we had already had a little bit of, I mean, we had enough data to show them and then approximate that rough.

So this is a complicated thing. It was a complicated thing to really find out how do you sample it in a way that uh the shape of the graph doesn't go back.

the, uh, yeah, the way we were measuring it is we had a similar product on the website too and there was a, a, browser in the browser also the graphs would drop and drop and would be present. So we would compare this. I mean, after sampling, we would compare how the shape of the graph looked to the exact graph which was drawn on the, uh, on the browser. Um, yeah, when, when it was drawn on the browser, there was no sampling done at all. All the data points were loaded in directly. So that's how we compared our success. Huh, so this was the prime thing. The second thing was, of course, this was the, uh, uh, from here we knew that we were doing something good because we were sampling something, right, so that the graph shape of the graph was not going right. Second thing was, uh, um, we measured that. Number of, as we, as we moved from one iteration to another, we saw that number of uh network hits that we had to make in order to get the data was reduced. So that reduced the amount of battery consumption also. So, these are the three things that we used in order to see, to make sure that we are going in the right way.

The way I would have done differently now is that instead of doing all the approximations and figuring out, I think I would have, uh, asked my, I would have, uh, uh, gone to my, my managers and told that this is a complex thing and we should not approach it, uh, um, with workarounds. It would be much more interesting if, uh, we go and, uh, there was a lot of research materials, so we, I would have asked for a certain amount of time to study that research materials and, uh, understand, uh, to make it much more, uh, we would have used the, uh, uh, research materials to understand. What could have been the best possible way rather than figuring it out in a hurry or in a, in a, we would have uh stood on the shoulders of those people who have done all the research and from there we would have derived something better. That would have taken a little more time but that would have created a quality which would have been, uh, far ahead of what we did. is what I would have done. Yep,