Ravi Mohan's Blog

Sunday, May 07, 2006

Grounding "exotic" technology - The Banker's story

The machine learning system I was called in to design for a Telecom Fraud detection company is moving into production. I was told yesterday that the testing folks are "amazed at the high accuracy of the new codebase" and my client is extremely satisfied with the improvements. I was really happy to hear this. Nothing makes a programmer happier than news that his code/design is making a substantial contribution in the domain for which it was written.

My clients were leaders in their domain even before we added the new ML capability. With the enhancements (not limited to the fraud detection engines, they've added a lot of functionality in other areas) they should wipe the floor with the competition. None of their competitors has a fraud classifier system as powerful. And we have implemented only a small fraction of what could be done.

This datapoint just confirms by belief that powerful technology allied to a strong management vision can lead to massive value addition. What made the difference is the adoption of a key metric(erroneous classification/million cell phone calls) as the measure of the software's quality and the almost total freedom given to the implementation team (which included some of the best folks in the company). We had NO political or turf battles and the laser like focus of the manager in charge on the key metric helped us to evaluate alternate strategies and select the most appropriate ones. This (tying the software developed to key business metrics) is one way of "grounding" business software and preventing thrashing. (More explanation of this conmcept coming up in future blogs).

Back to Banking. One of the nuggets that came out of our conversation(see last blog) was the fact that most Indian banks still do manual verification of signatures on cheques. This has been automated to a large degree a long time ago in most banks in the West. The algorithms for this are available in the literature. Large chunks of software that made up the original solution are publicly available. All it needs is for someone to put together a hardware/software combination to provide this functionality and the banks would jump at it.

Why is this important? Here is the Industry context. In 2009, the Indian Banking sector will be opened to Foreign banks . (It is semi-open right now). That would mean(assuming India is doing well economically) that tidal wave of foreign banks would arrive and the existing banks would face real competition.

The scariest part of this impending shift in the landscape, from the point of view of Indian Banks is the massive technological advantage that western banks enjoy (relative to the Indian banks). And in the Indian banking scenario, market share is almost directly in proportion to the technology level of each bank.

Here is an example. For years, only ICICI offered an "at par cheque" facility, ie, the ability to present a cheque drawn on an account in another city and get it cashed instantaneously (vs. waiting for few days for the cheque to be couriered to the city in which the account resides). Thousands of businessmen did not have an alternative but to bank with ICICI because no one else had anything similair.

By the time the other banks caught up with this technology, ICICI came up with a feature by which transfers between accounts could be transacted on the net (even if the target account was in another bank). As of today no one else has this feature(unless they go through ICICI) and this gives them an edge.

I forget the domain term for this. But the point is that the ranking in the technical arms race almost directly determines the marketshare. And this marketshare is what is under threat when the big boys land with their deep pockets and massive mainframes. So even something as simple as automated cheque signature verification could provide an edge in the impending blood bath in the banking sector.

So there. Now you have an idea on how to make millions. If you do, cut me a cheque for the idea. :-)

No comments: