Archive for April, 2010
First, check out what happens if you introduce a flame to a fluffy dandelion:
I first discovered that effect last year when I was playing around with my then new Swedish Fire Starter. In general the fire starter wasn’t very powerful, in fact, it was a little bit of a disappointment, but wow do you get a spectacular result from dandelions.
In other non-flammable news, I watched Woman in the Dunes this last week and it was marvelous. The poster doesn’t really do much for me, but the luckily the movie did. The premise is that a woman is trapped living in a valley between Dunes so steep she can’t climb out. A man is lured by the local villagers down to her home under the auspices of giving him a place to stay for the night before he returns to his home in Tokyo. The man awakes only to find the one way out (a rope ladder at the top of one of the dunes) gone, and his fate now tied to his counterpart. It has a slow beginning, but it definitely delivers. Highly recommend that one.
Movies that did not meet (or even come close to my expectations): Dune, Sleep Dealer, The Talented Mr. Ripley. Worst amongst them was definitely Dune, of which I only watched about 19 minutes, but those are 19 minutes I cannot get back. I mean, look at these screenshots. Just look at them.
Work has been super-frustrating as of late due to some unavoidable delays which are hampering my ability to proceed on several projects. I’m in the wrong line of work to expect things to turn out a certain way, but it’s always frustrating when powers beyond my control stymie my best efforts at productivity.
Nick is away visiting Chloe for the weekend, and Scott has spent the last 36 hours at an Oberlin college computer science department reunion so tonight looks like a double feature movie night for George. Exceeellllleeenntt.
Way back in November 2009, I read Paul Graham’s essay, “What Startups Are Really Like” and was especially interested in his thesis about the experience being more extreme than he can really impress upon his readers. The part about the emotional highs and lows in particular caught my attention. I’ve always been an emotionally extreme person, and I think being the CEO of a small web startup (Skritter) has heightened this aspect of my personality. But while reading Paul Graham’s article I couldn’t help but wonder, are my moods really as volatile as I feel they are, and if so, is there some way I could qualitatively describe them? This had both practical and theoretical ramifications. If I could figure out if I was misrepresenting my own moods, I could better balance my working moods. On the theoretical front, I was just damn curious.
Collecting the Data
So, in early December of last year I endeavored to record my moods regularly on a 1-10 scale, 1 being absolute rock bottom “we should just quit the startup” and 10 being “we’re likely to be bought by Google tomorrow.” Part of the reason I started recording the figures was to test whether I was over representing my negative moods, and partly to see how often I fluctuated and by how much. I reported the moods purely in a qualitative way, so it’s fairly impossible for me to tell if one day’s 2 was equivalent in terms of absolute serotonin levels to the following Tuesday’s. I can say, however, that I got much better at quantifying my moods, and have even begun using the system to let others know how I’m feeling. (I can now more quickly communicate how I’m feeling by saying “its a 3.5 morning” than by trying to describe exactly how that morning is going.)
Nick suggested early in the data collection process that I might want to record my reasons for mood changes, so that I could tease out common reasons for my mood swings. However, I decided to pull a minimum viable product on the experiment just to see if I could remain committed to the project for long enough to get a dataset of any worth. It turned out I had enough tenacity, and I managed to collect 119 mood changes over the course of 51 work days (a bit more than 3 months due to thanksgiving, Christmas, and new years). I recorded the numbers in plaintext in a Notepad ++ document that I leave running on my computer 24/7. At first I wanted to record in Excel, but found that the lightest, most ubiquitous data entry method was the one I was most likely to use consistently.
As I discuss below, I would be really interested to run the experiment again, this time tagging mood changes organically to see what exactly causes startup founders happiness and angst.
For the first data set, I simply took a list of the data points per day, used the min and max numbers and graphed them. Points where the two lines converge are days for which I only had one data point. Otherwise, for each value X the data represents the high and low ebb of my emotions in a typical working day:
The two big data-less slopes are from my long Thanksgiving day weekend and Christmas/New Years vacations respectively. I didn’t do any sampling when on vacation. Also, you’ll notice that I omitted the x axis labels, this was just to keep the chart clean.
One reason I started graphing my moods was because I thought I was feeling pretty low, and I wanted to know whether actually sampling my moods frequently would yield the negative results I expected. As time went one, I definitely started using my mood sampling as a way to record my frustration, which would account for at least some of the negative bias seen above.
What is surprising is that despite these fairly wild fluctuations, the median mood is exactly 5, or neutral. In short, I was prone to reacting to news emotionally (“Oh no, the conversion numbers for the new A/B tested signup page are rubbish, I guess the business is sunk!”) before my mind had a chance to intervene (“there are only 10 datapoints in the sample so far, perhaps I should wait a week or so before making a judgment about this iteration”). To add weight to that argument, take a look at the average number of times I recorded mood changes in a given day:
Again, the dates on the X axis aren’t to scale, but you get the idea. On average, I changed my mood 2 times per day with a median change in temperament of value 4. So essentially on most days I was changing my mood the equivalent of an 8 to a 4, or a 5 to a 1. The lesson here is that when my moods change, they tend to be big sweeping changes. Given that not too much can really change within a single day, I think it’s fair to say I over-react a little bit.
What this has taught me about running a startup:
As I mentioned above, I think this dataset has taught me 2 key lessons of working as the CEO of a young startup:
1) Even though the mood swings are dramatic, they are often temporary, and I would be wise to place less weight on my current emotional status. If only I could force my brain to only react to a two-week moving average of my moods!
2) I tend to over-represent my negative moods both in characterizing my base mood later and when recording data. In actuality my median mood was a perfect 5, so despite my moments of panic and ecstasy over minor business developments, I’m balancing better than I give myself credit for.
Further Data Collection and Analysis
As I mentioned above, the next step is clean up the reporting bias, standardize records/day, record mood change tags, and record for slightly longer. In performing the experiment, Nick got interested, and now we are planning to download a notification program and have it randomly ping us 3 times a day and do another round of data collection to determine how much the original dataset was skewed by biased reporting. We’re going to organically record reasons for mood changes and derive tags retroactively.
By the end of this data set, I had actually started recording some reasons for mood changes, and based on the extremely small sample size, I noticed that mentions of the word “lawyer” typically coincided with drops in happiness (changing the business structure hasn’t been much fun), mentions of investment (or lack thereof for the time being) cause dips in mood, and accomplishing tasks/building stuff tends to make me happy.
I look forward to making a more full, scientifically correct report here several months from now.
If you would like to take a look at the raw data (stripped of the tags unfortunately, some of that’s sensitive information!), drop me a line in the commend with your contact info and I’ll send you a copy of the excel file.
I just spent several very content and happy days with Becca, who was passing through on her spring break. She called me this morning and apparently she’s got an interview with a school that looks interesting! Great news from the job front.
So, over the break Nick ate a dozen donuts and then didn’t eat again for 24 hours. His total caloric intake: 2900. Total pounds gained in the course of the week: ~3. Nick needed more weight though, so it’s no biggie. I have videos of all 12 of them going down the hatch, will post that later.
We used the trailmaster to slice a mango in half (pit and all), a coconut also met it’s end on the edge of the trailmaster’s furious strength, video of that will also be forthcoming.
We watched several movies and played a board game that Becca bought the household at my request: Dominion. It’s like Magic the Gathering minus the whole he-who-pays-the-most-wins aspect. Mom and Dad, if you are reading this, I fully intend to bring it home with me and get you guys to play with me.
We spent ~32 man/woman hours making the wiremap I mentioned two posts ago. Here are some of the pictures of us constructing it. It still doesn’t work, Nick and I have to calibrate the heck out of it. It took a long, long time to construct but the results from last night’s calibrating are promising. If we get it working the good students of Oberlin are going to have one heck of a show some time soon:
Life post-car continues to stress me out. I did the math and it cost me ~$.25/mi to drive my Subaru a few times a month. For short distances, it costs me about $.40/mi to use the CityWheels car. However if I just don’t make any large trips outside of Oberlin (it’s gonna be tough in the long term), the CityWheels vehicle will actually cost me about 50% less than the Subaru, and that assumes the Subaru only costs $700 in maintenance per year. If I were to repair it right now, it would cost that much for the first 4 months of the year. I guess older cars are a little more expensive to keep around. The experiment will continue on, and it’s incredibly freeing to not worry about fixing the car, insurance, gas, etc. Now if I only could get past the feeling of being a poor bum that comes with driving a car that says “carsharing” on the side. I feel like a hippie.
We recently viewed Bringing out the Dead again. It had been many years for me, and wow is that ever a good movie. I mean seriously. Whoa. We laughed a lot. I didn’t get a chance to show it to Becca while she was here, but Nick, Ben, and I had a good laugh at Nick Cage’s expressions. Emily, I seem to remember that you love that movie, and man you’re right.
In my next post I’ll post some of the videos and pictures of Spring break fun.
You are currently browsing the Data Addict blog archives for April, 2010.