Tech 4D

Last week, Fred Wilson wrote, in Whats wrong with Alexa that he isn’t sure Alexa can be trusted with anything.

I agree that there are some large inconsistencies with Alexa, but you can still get some useful information by cross-checking Alexa with Quantcast (or comScore if you can afford it). For instance, here’s a correlation of the two sites Fred talked about, digg and del.icio.us;

Table 1

Assume that quantcast is more accurate than alexa (I think there are good reasons). From Table 1, you can see that alexa overstates the importance of del.icio.us and mildly overstates digg.

Now lets look at some other sites:

Table 2

Based on Table 2, I would throw away the alexa results for mapquest, history.com, discovery.com and nationalgeographic but keep information from the others.

Based on these comparisons I think I can also infer something about the demographics of Alexa toolbar users: they are classic “early adopters”. It would be very interesting to make correlation of alexa to comScore or quantcast and then calculate whether alexa toolbar users are good predictors for success of a new web trend or property.

Conversely, you might use this effect to predict failure as well. Take a look at mapquest vs maps.google.com (this is really scary). A Yahoo press release quotes comScore Media Metrix saying that, as of April 07, mapquest traffic increased by 3% while google maps traffic increased by 49%. But look at the alexa chart for mapquest:

During that time mapquest suffered a 50% drop in alexa toolbar users. Unfortunately alexa doesn’t report on subdomains, so I can’t get data on maps.google.com, but if I could, I suspect I’d find a large jump in usage of google maps by alexa toobar users.

So this data tells me that the early adopters that used to be mapquest users are leaving in droves while at the same time mapquest is adding new users who aren’t early adopters. This isn’t a good trend for any web property. On the other hand, there is a huge difference between the quantcast ranking (174 vs 13) and mapquest numbers, so perhaps alexa toolbar users never liked mapquest and it is no big deal if a half of them leave in one year. Or maybe something else is going on?

I’m an early adopter (I started using google maps in Feb 2005 when someone had to tell you about it) and I’m sorry to say that I don’t use mapquest very much any more.

(Thanks to Matt Owens for helping me with the data)

When I was a mechanical engineering student, I joined a five state regional A.S.M.E. design contest. The challenge was to make a coffee-can-shaped cylinder roll, under its own power, down a 15 foot track, climb a steep little ramp, pass a finish line, then roll back, all under its own power. The fastest vehicle to complete the course would be proclaimed the winner.

I figured this simple little task was perfect for me, being a creative, clever fellow, with good grades. I knew that, if you are smart and creative and approach a design problem with an open mind, you will eventually find the best answer. I was confident my design would be superior to all others. In fact, the simple little task was perfect for me, but not in the way I had envisioned - it taught me a lesson that has shaped my behavior ever since.

My friend Mike and I teamed up to work on the project. We decided early on in the design process to forsake the conventional wisdom of using an electric motor, feeling that we could transcend the pack mentality and come up with something truly revolutionary. Our idea involved planting a ridiculously overpowering gasoline model airplane turbine inside the can. A complex series of baffles and levers would manage steering and reversing. The scheme might even have worked if we could have figured out how to prevent the propeller from sucking the whole contraption toward one wall and subsequently choking itself of air.

Our friend Andre’ approached the problem more pragmatically. He knew the real problem was to maximize the power to weight ratio, so he focused on reducing weight rather than increasing power. He was very good at model building and constructed a beautiful, meticulously crafted wooden device that was light and fast. As the six or eight weeks of design and development drew to a close, I began to see that his conventional, but well executed cylinder was much more effective than our flying can. I wasn’t completely deflated, though, because the flying can was still pretty radical and it did some things very well. Besides, if we couldn’t win, then Andre’ would!

When the great day arrived, many students piled into a few cars and drove to the neighboring evil-rival university to witness the combined industrial might of the next great generation of engineers. There were 10 or 15 schools represented, each with a broad range of designs.

Someone even had the nerve to come up with the turbine idea - baffles and all! His version didn’t work either. Before the competition, all of the devices were laid out in a display area like so many monster trucks awaiting the axle swapping, gear grinding action action action of a tractor pull! The event, however, more closely resembled a demolition derby because only five of the 20 or so entrants even completed the simple little task.

The combined efforts of all of the entrants represented hundreds of hours of devoted labor, yet only five of us even arrived at a workable solution. Andre’s design was one of the five, but he didn’t win. In the display area, nestled among the microprocessors, gas engines, aerospace aluminum, delicate wood and high strength steel was a completely homely, unpretentious cardboard oatmeal container which contained a cheap little radio shack electric motor taped to a pencil and a spool of thread. It couldn’t compare in looks to some of the more elaborately constructed entries and it certainly wasn’t high-tech. But when its owner placed that oatmeal can on the track and let it go, we all knew that we had been out-classed. That can practically flew down the track, stopped on a dime and flew back to its master. The performance wasn’t just a little bit better than the competition - it was almost 10 times faster than the nearest rival was.

Afterward, we listened with awe and respect as the winner demonstrated his device. The beauty and elegance of his design was obvious once he described it to us, but we hadn’t thought of it ourselves. Hundreds of students in five states had been given the same set of design requirements. Among all of these participants, he alone had come up with a design that was overpoweringly superior to anything else: much simpler to build, 10 times faster, and much more reliable.

Here’s a diagram of the device.

The entire parts list is: cardboard oatmeal box, pencil, pin, two spools, thread, cheap radio shack motor, two AA batteries, wire, scotch tape. Wow.

I didn’t despair of ever achieving such greatness and virtue, rather I had something to shoot for. I was forced to concede, however, that I wasn’t going to get the right solution every time. Since that competition, I have worked in many fields and on many different projects. When I finish something, I look upon it with a critical eye, enjoying what I have accomplished but not assuming that I got it right just because it works. I have had a few successes and a few failures, but through it all, I have been guided and directed by that oatmeal box.

I’ll present a statistical analysis of the competition in a later entry.

Engineering Goodness, part 2

Engineering Goodness, part 3