The past few months have been very stressful for a lot of us – wherever you went on campus, the topic of conversation was the budget issues. They are real and the plans that are being developed by the senior administration to deal with them will have serious consequences all around. Two things are clear – the problem we are currently facing will have fairly long term consequences; all of us who are supporting the core academic business of the University will be challenged to do things very differently than in the past.

During the first week of January, I plan to discuss some of the budget cuts for ITS in more detail and the idea of having to deliver services differently as a result will become clearer. Now, on to other things…

An Exciting PDA Application

I happened to hear from a friend about ShopSavvy for Google Android - it lets you take the picture of a barcode which is sent online for price comparison. What a great idea. Here you are, in Best Buy, you snap the picture of that gadget that you badly wanted, send it over the net for price comparison and within a few minutes you know exactly how much it costs elsewhere. If it is considerably cheaper, you can go ahead and order it right from your PDA.

When I mentioned this to Jim Kamm, he told me that something similar existed for iPhone from Amazon. So I downloaded Amazon.com mobile for iPhones (from the Appstore) and have been testing it. It works differently. You can take the picture of any object for which you need further information and send it to Amazon. Within a few minutes, you receive an email with a link which tells you if Amazon was able to identify the object, if so, how much it costs in Amazon. Not Bad. So far, my success rate with this has been 60% when I did it. My children and nephews have been going crazy with it (like taking the picture of thai food and beer cans) which has driven down the rate far lower. One could take the picture of a book or the ISBN number of a book and do comparison shopping too.

Barcorama for Windows Mobile devices seem to do the same thing by doing comparison through PriceGrabber or Google.

I can see how this will eventually play out – the vendors will either agree to match the lowest prices or completely block wireless data access from within the brick and mortar store (which will not work in the end!)

Super Crunchers by Ian Ayers

How about figuring out the quality of Bordeaux wine using the commonly available data and the equation:

12.145 + .00117 * Winter Rainfall + .0614 * average growing season temperature – .00386 * harvest rainfall

Make sure to specify the rainfall in millimeters and the temperature in degree Celsius. This index is more useful as a relative number than absolute, so you need to measure it relative to some of the earlier vintages. You can read more about this in Orley Ashenfelter’s paper. Apparently, this method has outperformed some of the respected wine tasters by predicting the quality way before the wines are even ready for tasting.

The same way, how about the new measure of On Base Percentage as a better indicator of batting efficiency than the simple Batting Average?

The book called Super Crunchers, which was one of my holiday gifts, written by Ian Ayers from Yale uses several examples such as these to make the point that data driven decision making is influencing a lot of fields seeped in intuition and decision making by experts.

I am a big fan of this methodology and I have been asking for a lot of data whenever possible to make informed decisions. Google Analytics is a step in that direction. For example, while confirming some of the conclusions several of us seem to have arrived at based on our own experiences, it also provides several new insights.

I have been asking academic computing services to provide me with actual usage data from the public computing labs to make decisions regarding the best way to purchase hardware for these labs. what are the usage patterns? Are these computers being used simply to read email and do routine word processing type of activities or are they actually being used to access the specialized instructional software? Collecting these types of data and analyzing them should be the driving force behind how exactly to configure these labs and not what any of us conclude based on what we “see” in the labs.

We are pretty good in tracking network stats using various tools; similarly, we are tracking the CPU and memory usage of servers; we track various budget numbers over time; I look at all of these on a regular basis. They provide so much visual information in many cases and trending can be easily understood.

So, let us become super crunchers too… Let us capture the trouble calls in ways we can learn if there is a particular question that gets repeated, so we can proactively train the users; let us use random trials