Fred93's blog

Prosper's biggest hangnail is the fact that its loans perform so poorly.  One aspect of this is  collections.  Once a Prosper loan goes 1 month late, the loan is sent to collections, and it almost always defaults.    If you can't collect, then Prosper loans aren't worth much.

Collections performance has improved at various times, but it has been stagnant for the last 3 months.

I've graphed a simple ratio.  The denominator is the total number of loans Prosper has sent to collections (over all time up to the observation date) and the numerator is the total number of loans which became current after going into collections.  I have included only loans sent to Penncro and Amsher, because historically those are the numbers I collected daily from Prosper's web site.  I never had much interest in the other collection agencies, because they never were given more than a handful of loans.  The color of the curve changes on the day (2/23/08) that Prosper fired Penncro and moved all the in-collections loans to Amsher.

When the curve bends upward, it means that recent performance is better than historical, and is thereby pulling up this all-time average number.  When the curve is flat there is no improvement.  It is -- most recently -- flat. 

I started complaining about Prosper's collections operation back when the numbers were around 6%.  That's really a ridiculously low number!  Its amazing how much flak I got at the time.  Several know-it-all lenders told me that collecting on Prosper loans was inherently impossible.  "Can't get blood from a turnip." etc.  I never believed that.  I believed that Prosper's operations were simply not competent.  Over time it became clear that with a lot of pushing from vocal lenders, Prosper did pay more attention and roughly doubled this measure of collections performance during early 2007. 

By spring  2008 the curve levelled off again, as Prosper stopped paying attention.

Sometime around mid-September of 2007, Prosper hired Doug Fuller to head operations, which means "taking care of your loans".  After Doug arrived, the curve started heading up again.   October thru December of 2007 were great months.  It looked like Doug was able to make Penncro perform much better than they did before.  I called that bend in the curve in Nov'07 the "Doug Fuller bend". 

Eventually Doug expressed displeasure with Penncro.   As a result, he hired a new company, Amsher, in February 2008.  Now from everything I've read about Amsher, they seem like a great company.  You should read their web page.  Unfortunately, about the time Prosper's collections business moved from Penncro to Amsher, performance sagged.  The curve is now once again horizontal.  This means that Amsher's performance is around 17.5%, simply matching the all-time historical norm.  Stagnation.  Bummer.  There has been no explanation of this from Prosper.

Doug Fuller occasionally posts a collections status update in the official Prosper blog.  These generally have a postive message, like things are improving, and some ad-hoc statistics that you can't verify.  For example, in the most recent update, he talked about dollars per loan per month being recovered.  For example, we have no idea what loans he counted in that denominator.  He described this number as higher than in the past, but depending on which loans he counted, this may not have been a fair comparison between Penncro and Amsher.  I think my presentation makes more sense.

Background reading:

Over a year ago I wrote an open letter to Prosper about the collections problem.  Some of it is a little dated, but the main ideas are still correct.  You can read it here: Open Letter #2

Doug Fuller gave a lengthy talk on Prosper's collections at ProsperDays'08.  You can watch it here: PD08 Collections Video

Technical notes: I left this part for the end, because most people aren't interested in the details. 

When lenders attempt to understand collections performance, they face several difficulties.  Prosper publishes collections statistics, but these statistics are not very helpful when we try to compare two collection agencies.  You may have noted that Amsher's "brought current" percentages are all lower than Penncro's numbers were.  Why?  One reason is that the characteristics of the pile of loans they're working on are different.  Penncro got loans that were freshly late, whereas Amsher started with a pile of loans that had been sitting around Penncro for several months.  Its not a fair comparison.  I've chosen a methodology which avoids this sort of problem.

I've generated a measure which combines the effort of all the players: Prosper, Penncro, Amsher into one number.  As described above, the denominator is the count of all loans that have gone to collections, and the numerator is the count of such loans that got fixed.  Simple.

Here's how I form the combined statistic. 

On 2/22/08 Prosper's collections page showed us that 3348 loans had been sent to Penncro.  On 2/23/08 the same page showed us that 1181 loans had arrived at Amsher.  2167 loans "disappeared" from the statistics.  I asked Prosper about that, and I learned that only the "live" loans were transferred.  Loans that were not actively being collected (either became current, or defaulted) were not transferred.  Ok.  Whatever.  Sorta makes sense, but it makes it more difficult to understand history.  I add those 2167 loans back to the number of loans Prosper reports as sent to Amsher.   This gives me the total number of loans sent to either Penncro or Amsher, in other words, my denominator.

Forming the numerator is a similar exercise.  On 2/22/08, Prosper showed that Penncro had been sent 3348 loans, and that 17.3% of these had been brought current.  17.3% * 3348 = 578 loans brought current by Penncro.  From the percentages brought current that Prosper gives us for Amsher I compute the number of loans Amsher has brought current.  To this I add the 578 loans brought current by Penncro, giving me total number of loans cured, in other words, my numerator.

I record several of the numbers on the Prosper collection agency web page every day, do the calculations I've described, and draw the curve.

Here's the mid-April update to to my late loan statistics charts.

These charts show statistics for the performance of all prosper.com loans. Each curve represents the set of loans that were created in one calendar month. The vertical axis is the fraction of those loans that have "gone bad", in other words are 1 month late or worse (up to and including default). The horizontal axis is the observation date. All data comes from Prosper.com's performance web page.

A larger, more readable version of that chart can be found here

Explanation of methodology can be found in my prior postings in this blog, and in my postings on the old prosper.com forum archive found at http://www.prosperreport.com/

Here's the April update to to my late loan statistics charts.

These charts show statistics for the performance of all prosper.com loans. Each curve represents the set of loans that were created in one calendar month. The vertical axis is the fraction of those loans that have "gone bad", in other words are 1 month late or worse (up to and including default). The horizontal axis is the observation date. All data comes from Prosper.com's performance web page.

A larger, more readable version of that chart can be found here

It does continue to look like there's a change in the quality of loans (for the better) between Oct'07 and Nov'07. 

On the other hand, we've been fooled before.  Earlier it looked like there was a significant change in behavior between Feb'07 and Mar'07.  We thought that Mar'07 and later curves had lower slopes.  As things evolved, we now see that was an illusion.  Feb'07 was a worse than average month, and Mar'07 was a better than average month, so there was a visual "gap" between these two months.  As we now look back on Apr'07, May'07, Jun'07 we see that these curves are very similar in slope to prior months.  Loans got a little better early in '07, but the effect was slight.

Explanation of methodology can be found in my prior postings in this blog, and in my postings on the old prosper.com forum archive found at www.prosperreport.com

Here's the 03/01/08 update to my late loan statistics charts.

These charts show statistics for the performance of all prosper.com loans. Each curve represents the set of loans that were created in one calendar month. The vertical axis is the fraction of those loans that have "gone bad", in other words are 1 month late or worse (up to and including default). The horizontal axis is the observation date. All data comes from Prosper.com's performance web page.

A larger, more readable version of that chart can be found here

Here's a version with the curves slid to a common origin.  The horizontal axis is now days since month of loan origination.

Prosper threw us a curve this month.  You may recall that a group of loans were partitioned off a few weeks back into a program called "new agency test".  Its actually a set of very late loans on which has engaged some lawyers to take legal action against deadbeat borrowers.  For techncal legal reasons, prosper bought these loans back from lenders for either $0.00 or something very close to zero (depending on whether lenders opted-in to the program ... a detail that doesn't matter here) prior to initiating legal action.  After buying the loans back, prosper changed the state of these loans in their database to "repurchased".  I know thi sounds straightforward, but it is not.  The "repurchased" state was created early in Prosper's history for an entirely different purpose.  At that time, Prosper would occasionally acknowledge that a loan was a result of identity theft, at whcih time Prosper would buy back the loan from lenders at 100% of face value.  Prosper argued that by buying back the loan at face value, they restored lenders to a condition as if the loan had never existed.  Therefore, they chose to exclude these "repurchased" loans from loan performance statistics.  The prosper performance web page, for example treats "repurchased" loans as if they had never originated.  I believe thta most 3rd party statistics web sites (ericscc, lendingstats,...) do the same. 

Now however, Prosper has used the "repurchased" state for a new and different purpose.  Loans that have gone bad, and for which the lender has not been reimbursed, have been put in the "repurchased" state, and eliminated from loan performance statistics.  This error makes Prosper's ROI calculations look better than they should, and similarly boosts the lender ROI calculations at ericscc's and lenginstats' web sites.  Bummer.  This is database pollution.

A note for the conspiracy theorists: I don't think Prosper did this on purpose.  I think they were lazy, and didn't think thru the implications of their action.  The loan "states" in the database have meaning.  Software depends on that meaning, even software written by Prosper itself.  If you're gonna use an existing state for a new purpose, make sure you understand the meaning first.

I have carefully adjusted each month's stats in the curves above so that these repurchased loans are counted in both the numerator and denominator of the bad loan ratio.  I urge others who use Prosper's data to do the same, at least until Prosper fixes this problem.

Discussion of methodology can be found in earlier posts in this blog, and the late loan thread on the old prosper forum, now found here: http://www.prosperreport.com/prosper/forums/archive/threads/1/1/2/11290.0.HTM

Here's the 02/15/08 update to my late loan statistics charts.

These charts show statistics for the performance of ALL prosper.com loans. Each curve represents the set of loans that were created in one calendar month. The vertical axis is the fraction of those loans that have "gone bad", in other words are 1 month late or worse (up to and including default). The horizontal axis is the observation date. All data comes from Prosper.com's performance web page.

A larger, more readable version of that chart can be found here

Discussion of methodology can be found in earlier posts in this blog, and the late loan thread on the old prosper forum, now found here:  http://www.prosperreport.com/prosper/forums/archive/threads/1/1/2/11290.0.HTM