Credit Crunch - Math vs Judgement

Excellent in-depth article in NYT on whether the use of VaR led to the cavalier risk taking by investment banks or of it is just the lack of judgement. Blaming mathematical modeling for the greed and poor judgement of the management is a travesty of justice.

Games People Play!

Online gaming is already one of the most popular online activities. The demographics of gamers and their interaction with their close-knit community, and the context of their favorite games provide a great opportunity for marketers to target them with the right offers. Jeff Kaplan's article makes interesting reading, though I am not sure of the incremental opportunity that gaming portals have over Google's AdWords and its video ads.

Lending Strategy – The Delinquency Blacklist

A lender on Prosper would probably have greater confidence in funding listings with membership in a reputed group. We spent some time this week studying the reputation of high activity groups in terms of delinquency rate [what percentage of a group’s outstanding loan book is delinquent (more than one month late)?].

Our analysis indicates that the groups listed below contribute to 30% of all delinquent loans (by volume) though they make less than 17% of the total loan book. Prospective Lenders might want to keep that in mind when reviewing listing from these groups. With a lending strategy that evaluates listings from these groups carefully, you might save your money from going default.

Rank	Group Name	Delinquency Rate (Volume)	Delinquent Volume of Loans	Total Volume of Loans
1	MustardTree	13.44%	$24,651	$183,460
2	435 Supporting Lenders	10.71%	$256,740	$2,398,297
3	UtopiaLoans.com	10.53%	$107,450	$1,020,670
4	PsychDoc'sGroup	9.76%	$226,336	$2,319,594
5	KoyneLoans.com	5.68%	$49,501	$871,383

• Data is between November 2005 and March1, 2006.
• Only groups with more than ten have been analyzed.
• Delinquent loans refers to loans that are more than one month late.

We would love to hear from you on what you think, and what other analyses you would like to see on this blog.

Lending Strategy - Top 5 Groups to Lend To

As a Lender on Prosper, one is constantly on the look-out for groups of high quality borrowers, whose listings attract enough lenders to get funded and whose loans yield significant returns. We spent some time looking at groups with high activity levels, where one could find such borrowers in this context.

We started out by saying "can we identify groups that tend to have,

· Low proportion of high risk loans (<33%)

· Above average yields (>14%),

· Above average listings-to-loan conversion (> 20%)?"

Well, we believe we may have found just the groups you need, so that you can make your money work harder, minimizing idle money, while giving better returns. With a lending strategy that allocates a higher proportion of your money to listings from these groups, you might save yourself precious time finding the right listings, get them funded, and get good returns. Here are the groups, and the metrics we used:

Group Name	Weighted Av. Lending Rate	High Risk Volume %	Listings to Loan Conversion Rate
Voyagers Financial	20.21%	16.61%	37.39%
DocProsper's Credit Community	14.68%	12.71%	35.29%
America One Funding	18.59%	14.59%	29.46%
Professors and University Graduates	19.39%	33.80%	23.23%
NuBeginnings	17.87%	31.61%	21.93%

• Data is between November 2005 and December 2006.
• Only groups with more than hundred members have been analyzed.

A quick note on the metrics:

Weighted Average of Lending Rate, defined as the average of Lender’s Rate weighted by the amount borrowed for the loans in a group, is indicative of the yield that a lender can be expect from the group, taking the volume of loan into consideration as well.

High Risk Volume, defined as the percentage of volume of loans that have been lent to low credit grades E & HR, is indicative of the quality of the loan composition of a group. Lower the dollars lent to low credit grades, higher is the quality of the loans.

Listings to Loan Conversion Rate, defined as ratio of loans to listings, is indicative of the group’s contribution in getting its listings funded fully. When a lender bids for listings from groups with high success rate, chances that the lock-in costs are minimized for him.

We would love to hear from you on what you think, and what other analyses you would like to see on this blog.

How much is your home worth? Ask Zillow.com!

I recently came across this amazing tool that allows you to tap into home valuation data nationwide in the US, and get a sense for the property market. Zillow combines data analytics, visualization and mapping technology to deliver a high quality consumer service.

Order Online for Discounts, Pickup Offline at Store!

The WSJ reports on an emerging trend among cross-channel retailers such as Wal-Mart, Circuit City, etc., wherein they are offering discounts to holiday shoppers - who order online, but come to their store to collect the goods. Not just that - if you are made to wait at the offline store more than the promised time, you get gift cards and coupons as well. Apparently, the traditional brick-and-mortar retailer is only now warming up to the potential of the web to send more customers to their store, but one wonders if this would really give them the edge to grab a share of the ever growing online retaining pie.

Oracle pushing the Analytics Envelope

Oracle has announced the released Oracle Data Mining 10gR2 Code Generation Release. This augments its increasingly powerful data mining capabilities (that includes a GUI-based tool that automates repetitive processes in a data mining project) with automatic code generation functionality for PL/SQL.

With the addition of powerful features to Oracle 10g database (not the least of which are support for regular expressions, and enhancements to grouping and analytical functionality), Oracle has taken giant strides in the analytics space. Oracle's data mining tools serve as a good complement to its business intelligence platform. Its push into data mining is blurring the distinctions between traditional database servers and special purpose statistical tools.

We see many benefits of using Oracle as a Data Mining Platform:

1. Leverage the skills of a large pool of Oracle and Java programmers for data mining projects
   
2. Reduce the number of disparate platforms, and instead use Oracle's well integrated end-to-end platform for all analytical needs, including data warehousing, reporting and predictive analytics
3. Benefit from the scalability, security and versatility of Oracle database technology

If you are interested in Oracle's data mining initiatives, visit the following links:

1. Oracle's Data Mining and Analytics Blog
2. If you want to learn about Oracle Data Mining, here's the tutorial
   

Put horse before the Cart, and revenue shall follow!

In a very relevant article, Billing World highlights the importance of Telco managers understanding the power of customer analytics, while designing new offers and the need to move away from the engineering driven mentality of "build it and they will come" that is so common today. Can't agree more!

"Peopleware" Is The Secret Sauce!

In this article by Niel Mason, that builds on a report by Forrester, Mason argues that even with all the advances in hardware and software, successful customer analytics projects is more about having the right people. Simple enough! - he picks presence, commercial awareness and a detail oriented number cruncher as his ideal analyst. I would have thought business insights would pretty important as well - maybe, he considers that part of the commercial acumen.

Keeping up with Fraudsters - The Challenge

In this recent article in Intelligent Enterprise, Michael Voelker has highlighted the challenge of staying a step ahead of fraudsters, as they seem to be constantly developing new ways to evade detection. The article goes on to highlight the use of predictive analytics that has significantly helped to combat fraud, and the use of business rules engine that makes the decisions based on trends identified by the predictive model, actionable.

Social Networks + Forecasting = Prediction Markets!!

In a recent feature, BusinessWeek has covered the emergence and adoption of prediction markets by enterprises to significantly improve their forecasting ability. Instead of getting a single business manager to do the forecasts in the conventional model, firms have started to create a prediction market, where groups of employees have an opportunity to participate in the market, and hopefully lead to more accurate forecasts on product sales, consumer interest in new offers, etc. The employees get paid in cash or kind as an incentive to get it right, and this puts the power back to the employees who are knolwdge able, but who don't voice their opinion, costing enterprises lost opportunities!

Making Marketing Measurable

Yet another survey - this time by Forrester Consulting - commissioned by Unica - confirms automating brand marketing is a top priority. I wonder if Forrester could have come back and told Unica that there is no market for their products :-) Be that as it may, this is certainly an area that is underserved and needs more management attention.

Analytics - The Future of BI - say Experts

In a recent poll by NetworkWorld on the "The Future of BI", analysts and industry leaders, identified the emergence of analytics as one of the key future trends in business intelligence technology. According to these experts, One key trend to look forward to is the automation of repeatable, operational decisions that are currently in the domain of "gut feel"!

Spreadsheets R Us?

Retailers take a hard look at their internal data management processes so that they have better quality of data that they can rely upon to do advanced analytics, says a recent research study from Aberdeen Group, covered in Intelligent Enterprise.

Insurance companies get smarter!

SPSS is organizing a series of online seminars on Predictive Analytics for Insurance