Tag: data

Did Engineering Blind RIM to iPhone’s Assault?

BlackBerry 850 photo
BlackBerry 850

Research In Motion board member Roger Martin offered the Globe and Mail a passionate, though rather odd, defense of of RIM’s actions in the long decline that led up to the departure of co-CEOs Mike Lazaridis and Jim Balsille. In the course of defending for ignoring advice he described “go bankrupt and fire our founders and bring in a moron,” Martin, dean of the University of Toronto business school, may have inadvertently put his finger on the real problem.

“People were saying we can’t make powerful phones like Apple,” Martin told the Globe and Mail.  “Yes, we can, but we couldn’t believe consumers would put up with that kind of battery inefficiency and that kind of network inefficiency.”

Mike Lazaridis is an engineer to his very core. I met with him many times over the years and he would always bring the newest BlackBerry, often a product weeks or months away from release. He was always proudest of two features: its battery life and the way it kept network traffic–initially just email and later web data–to a minimum.

In the beginning, these were huge virtues for the BlackBerry. RIM had its origins in the paging business, where long battery life and minimal data use were paramount. The original BlackBerry 850 not only looked like a pager, it ran on the very low-bandwidth DataTAC paging network and probably used less data during its lifetime than an iPhone consumes in a few minutes.

As the BlackBerry evolved into a smartphone, Lazaridis’ obsession with battery life served it well. Its early competitors the Handspring (later Palm) Treo and Windows Mobile Phones had problems in both departments. At one time, the BlackBerry proxied browser ran rings around the competition in loading pages.

The problem is that as networks got much faster, BlackBerry’s data stinginess became more of a burden than an advantage. It’s easy to forget, but the iPhone was the first smartphone to offer anything resembling a PC browsing experience, and it didn’t really get good until the introduction of the iPhone 3G in 2008. But from then on, the BlackBerry was lost. People wanted full-featured web pages, they wanted them fast, and they didn’t care much about the data consumption, especially when they were doing a lot of their browsing on fast, unlimited Wi-Fi networks.

BlackBerrys still crush the competition on battery life, often able to go through a couple days of hard use without recharging. The difficulty is that the competition has gotten good enough. The demand is for a phone that will get you through a day’s use before you need to recharge. More is nice, but it doesn’t sell hardware.

I think it is likely that the BlackBerry engineering team, led by Lazaridis, fell in love with the wrong metrics. They were still concentrating on battery life and data consumption as the world became concerned with performance and app availability. Too much of a good thing can, indeed, be very bad for you.

Wolfram Alpha: Analytics for the Rest of Us

Wolfram Alpha may finally be making the big leap from interesting curiosity to generally useful tool.

For the 2 1/2 years of its existence. Alpha has had something of an identity crisis. It’s sort of a search engine, but instead of crawling the web for results, it relies on a large set of curated databases. At its heart is the computational engine of Mathematica, so it can do some amazing things with that data. When you asked it a question that it knew what to do with, it could produce very interesting, sometimes surprising, analytical results. But it often bombed on simple queries that Google handled with aplomb. It could tell you the distance between San Francisco and Buenos Aires (6,456 miles or 35 light milliseconds) but not where to find a nearby pizza. Alpha got a considerable boost when Apple chose it to handle computation-based queries from Siri–the New York Times reports that Sire generates a quarter of Alpha’s traffic–but it  remained a bit of an oddball service.

With the launch today of Wolfram Alpha Pro ($4.95 per month, $2.95 for students), the service is promising to get a lot more useful. There are loads of new features that make the pro option well worth its modest cost. One big change is that you get a persistent account, in which past analyses are remembered and can be modified. Another the ability to upload your own data for analysis and then to download any of the analytical results, including some very sophisticated graphics.

As a simple test, I uploaded about 65 years worth of government data on U.S. growth of gross domestic product, change in the Consumer Price Index, and the unemployment rate. After an initial glitch, in which Alpha interpreted a column headed “Real GDP” as representing amounts of money in Brazilian reals, it gave me an instant deep analysis of my data. As I would have expected, it should a fairly strong relationship between GDP growth and unemployment,  but no significant correlation between inflation and either growth or unemployment.

bubble chartThe bubble chart on the left is an example of the sort of thing Alpha does automatically. It’s a sort of three-dimensional graphic, in which the y-axis represents GDP growth in a given year,  the x-axis is inflation, and the size of the circle represents the  unemployment rate.  Other data generated include scatter plots, histograms, a regression equation, a covariance matrix, and principal component analysis for analysis of variance.

Using Pro, you can also perform analysis on Alpha’s extensive databases  and  download not only the results but the raw data (there are some restrictions on data exports imposed by Wolfram’s licensing agreements with the data suppliers.) CEO Stephen Wolfram describes the service as “all the tools a data scientist would have combined with knowledge of the world.”

Another feature offered by the Pro service is analysis of images. Upload a photo and alpha will perform a range of  analyses on it, including metadata, color analysis, intensity histograms, edge detection, and even some basic editing tools. Nothing that Photoshop can’t do, but its quick and easy.

The real prize in Alpha Pro, though, is the data analytics. You can learn a lot when you start looking at your data in more depth than yu can get from a spreadsheet. Alpha makes it cheap and easy to get started.