Quantifying What To Wear

So my name is Andrew Paulus and today I’m going to be talking about quantifying what to wear. So I’m one of the co-organizers of this meetup and to give you a little sense my background is in product management and product development and a few different startups and tech companies around here.

So I’ve been involve with Quantified Self for a bit of time now and I’ve done a lot of traditional tracking of sleep or you know activity, mood. And so about a year ago I was reading a book that I’m sure some of you are familiar with, it’s called The Power of Habit and it really got me thinking what I was doing on a daily basis.

One of the things I notice was meditation for me was a very foundational habit and in the morning I realized that right after I meditated I was in the habit of every single day checking my phone and checking the weather on my phone. And when I really started to think about it I realized I was doing that habit every day to figure out what I should wear for that day.

So I thought that was kind of interesting that every day I would be looking at my phone and I would see numbers much like this for September that aren’t necessarily that easy to interpret. So it got me thinking about how can I make these numbers easier or more applicable to the job I was actually trying to to solve and to figuring out what to wear that day.

So to start I didn’t really know what I was doing, but I started putting numbers down and tracking what I was wearing so I did this of a month and a half or a little bit more starting in January this year.

And after about a month you know I just did it every day in the morning once I got to work and didn’t really think too much about it just because I wanted something out there. then about a month I looked at it and started to realize some big problems with what I was doing.

So first of all, to figure out the weather data I’d just be looking at my iPhone at not a consistent time, so the weather data wasn’t really standardized. I didn’t really know what to enter so I made up categories or shorts or pants, or what kind of socks and I had been way too descriptive with what I was actually wearing; I didn’t need that level of detail.

So I tried to revise what I was doing and probably went a little too far on the other end of standardizing everything. So this is a screen shot of the second round of my data tracking that I did. So what I did I every day tracked what I wore and then around at six o’clock at night what I think I should have worn. I went through pretty detailed. I went through each pants or shorts, socks, type of coat, sunglasses – everything. So I thought I went a little too far on that side, but I also figured out a way to standardize the weather data I was receiving. So for that I actually found that getting a daily farmers forecast was very good, so that had very accurate data on the chance of precipitation, highs, lows, wind.

And so I continued to do this for many months and my idea was to do this to capture the entire weather spectrum. So you know the heart of winter in January through summer. So my nice girlfriend was nice enough to actually do the same thing for herself with slightly different data since there are skirts and tights involved in the girl side of things.

So after I guess six or seven month of data I was able to sit down and really go through the results, and figure out if there is a way that I can really create a more relevant kind of weather application for myself.

So a few takeaways is that i picked the right kind of article of clothing about 78% of the time, my girlfriend did about 74% of the time. Incorrect choices were due to various factors, sometimes I didn’t have the clothes I wanted to, other times they just weren’t clean, so that kind of reinforced what I wanted this tool for because I thought if I know what I’m going to need to wear in the future I can be better about making sure that’s washed or I had purchased it or something.

So what I did is that I took all of the various weather data and tried to look at the variation coefficients between that and each one of these kind of articles of clothing or each of these decisions, so whether it was shorts or pants, or whether you should wear a T-shirt or add a second layer or something or whether you should wear a hat or a scarf.

I looked at the correlation coefficients for each one of these kind of metrics across each weather data point, and what I realized that for me it was fairly predictable and you could predict what I should wear and create some real rules and guidelines. For instance I know it’s a little hard to see here but if the daily average is above 69 degrees I know I would be more comfortable in shorts. So these were the kind of insights that I got out of this.

Looking at the girl’s data it quickly became clear that there was an explanation why all girls looked at me like I was crazy when I said about this idea before because there is absolutely hardly any data correlation there compared to the men’s or compared to my data.

So where my correlation coefficients for most of these things were you know 0.7 and the girls data was more like 0.0, 0.2, so that signaled to me that clearly there are more things to consider probably on average for women when deciding what to wear.

So based on that I’ve kind of realized that this could possibly be a very useful tool if there are other guys like me, but if most girls are somewhat similar to my girlfriend in clothing choices then maybe it would be such a good choice for them.

So hopefully in the future I’ll be able to demo this tool, because I did all this with the hope that I can make it into you know a little tool that I can use, so that’s coming along and hopefully I’ll be able to show it in the future at a future QS meetup.

Thanks.