Models
First Phase
This was their public-stated business model.1
Xamplify was founded on December 10th, 1999 to provide Internet users with a compelling solution to their frustrations with one-size-fits-all tools and content. Our patent-pending software allows for real-time data analysis, helping companies to identify their most valuable customers and giving users the control to shape their online experience.
Reasonable enough.
Second Phase
Now the company's creative juices were really flowing. At the time I joined the company during June 2001, it claimed that its products were 2 years ahead of that of its competition which I believed based on Elwyn Berlekamp, David Donoho, Bob Burnett and similar high-caliber individuals listed as being associated with it.
Gist of what the company seemed to be claiming during this period:- The company has a list of well-designed questions designed by its world-leading experts.
- Just one question for any major psychological trait is good enough. (implicit)
- Online survey consisting of about 10 questions, one question for each psychological trait to be measured, administered to about 1000 customers of an institution can be used to accurately determine the psychological profile (the company called it psychometrics) of millions of customers of that particular institution by using sophisticated extrapolation techniques designed/audited by its world-renowned statistics "genius."
- These extrapolated psychological profiles (psychometrics) combined with traditional demographics data (like age, sex), transactional data (like time elapsed since last transaction), and product ownership data (like type of credit card owned) will result in identifying customers most likely to buy.
- This result is demonstrably superior to the result obtained by using the traditional analysis using demographics, transactional and product ownership data only. Means, adding the psychometric dimension really, really helps.
It was fascinating. I tested out the online survey by taking it myself and it took me about 30 seconds to complete. Assuming that the average test-taker took 5 minutes (absolutely the worst scenario), testing 1000 individuals should take about 5000 minutes which comes to less than 100 hours. If these 100 hours can result in significant cross-selling improvements, big institutions would be fools not to buy this at the bargain price of just half a million dollars.
A few weeks before I left Xamplify, Dr. Robert Oliver ("President"), Sumer Johal (Chief Technology Officer), and Ron Anderson (Director of Sales) apparently made business trip to several banks together. Everything was very secretive so I don't know much more, but just after the trip, Ron Anderson happened to show me an email just received from a potential client this trio had visited in this trip. The email worded somewhat like this:
This email is to request you to confirm if our understanding of what Xamplify told us during your visit to us is correct. We understand that Xamplify has been able to show improvement of over 600% in cross-selling by extrapolating psychometric data from a sample which is less than 1% of the total customer base.
I believe that the 600% figure quoted there came from results obtained by running my model on a trial customer. Xamplify's management being familiar with the model, my only conclusion is that there was certainly some serious lack of communication or...
The Spying Phase
Sanjay Dayal, a colleague I spent a lot of time with, was hired a couple of months after I was, and he proposed some database-related concepts like "observation agents." Soon Jeffrey Klein, Sumer Johal, Julian Brookes etc. were spending a lot of time on it. It was filed for patent, perhaps Jeffrey Klein as the co-patentor. This phase turned out to be a very interesting one. Fair Isaac seemed to be interested in it. Robert Oliver, Andrew Rudd and other investors joined the Xamplify team around this stage. I guess that they could still be in this significant phase.