Changes

Robert Oliver of Fair Isaac became an investor and started coming to Xamplify on a regular basis on hold meetings with Jeffrey Klein and Sumer Johal. A couple of months later Bob Burnett, one of our major investors and the president, was forced to leave. Something to do with direction of the company. There were some other changes brewing too.

One day Sumer Johal informed me that the company was happy wuth my work and wanted to bring me in. Which meant the position of a regular employer. Two decades ago I had graduated from IIT, Bombay and collected other graduate degrees and experience on the way, and all this hard work had paid off finally and I was to become a regular employee just like the gora sahibs at Xamplify. It's a tribute to my self-restraint that I didn't breakdown crying with joy, having finally made it. I had to know that the presence of civil right champion Bob Burnett, founder of Computer Professional for Social Responsibility, Terry Winograd, founding editor in chief of lefty Mother Jones magazine, our Jeffrey Klein - all of these gentlemen had something to do with this momentous decision and my soul is eternally grateful to them. It must be my good Karma of my past life otherwise who am I to sit on those $1,300 Aaron chairs when I could be sitting on the floor. Anyway, Sumer Johal cheated on salary negotiations again, he was just being himself but a 40% increase in wages to $65,000/year, allowed to take two courses at UC Berkeley, let me think of this job as a half time job to justify my salary. However, there were pockets of unrest in the company and real reasons unknown to me I was made a contract worker again soon after a meeting. I mean a contract worker without a written contract.

I heard that there was a power struggle at Fair Isaac and that Robert Oliver who was the chairman of the board there, was being forced out. Anyway, he started coming regularly to Xamplify. Knowing Sumer Johal's lack of elementary skills in statistics and finance, I was almost sure that his lack of knowledge would be caught by Robert Oliver. It was not to be and soon Sumer Johal appeared to be Robert Oliver's favorite employee, just like he was the favorite of our departed Robert Burnett.

Andrew Rudd, a wealthy Berkeley businessman, who is the cofounder and Chairman of Barra, a highly successful financial software company (symbol: BARZ)in Berkeley came one day and decided to become a major investor. I heard that his Ph.D. thesis advisor at Cal Berkeley was Robert Oliver.

I used to come to work around 7 to 8 AM and used to generally stay late - till 11 or 12 in night. One night, towards the end of 2001, I decided to stay late. After 4 AM, the fax machine started and I was curious. Seems like it was a routine transmission from CIA in Richmond, VA to Robert Oliver regarding background check on an individual. (No, that individual wasn't me.) I thought maybe Robert Oliver wanted to dig dirt on someone at Fair Isaac opposed to him, perhaps a standard practice for powerful person like him, but now I think that it could have been related to Xamplify moving into money laundering detection business. I really don't know, why don't you ask him?

The trial customer wanted us to predict potential customer purchase probabilities for certain products. Since psychometrics (psychological profiles) was our unique selling point, I decided to use the extrapolated psychological traits with the usual demographic and transactional (includes product ownership data) data. Very soon it became obvious that it was not going to work. The extrapolated psychological traits were computed as linear combination of demographic and transactional data and in a linear regression were not able to add any prediction accuracy over using the traditional data and instead made the model unworkable. (It makes the design matrix singular!) If the psychological traits were computed NOT as a linear combination but rather some non-linear combination, things would have been different.

This problem was severe since it hit us at the core of our claims. Sumer Johal was informed and he accepted the model results which didn't use any extrapolated psychological traits. The model had other problems too, but Xamplify didn't seem to care. I was able to achieve R-square values of around 0.25 or more which were quite reasonable. Since at that time we didn't know the timing of the product ownerships, it is quite possible that we were predicting past ownerships based on future ownerships!

Initially I heard rumors the model didn't work. Later I heard that it was quite successful and it seems Xamplify was able to portray that use of extrapolated pschological profiles helped cross selling significantly, say by 600%. (Remeber that if you have a stodgy customer, and only 1 person bought the product before your promotion and later 7 customers out of a potential 500,000 bought it, it will show improvement of 600%. This is just for illustration only. Actual numbers may vary.)