Author Topic: Cancelled notes: any analysis on those  (Read 4503 times)

Kombinator

  • Full Member
  • ***
  • Posts: 193
    • View Profile
Cancelled notes: any analysis on those
« on: October 22, 2013, 05:53:48 PM »
Wanted to see if anyone has done any analysis on some parameters that cause notes that pass all the screens and get funded to be ultimately not originated by the platform. 

Seems this ratio is very steady at about 32% of total notes / dollars committed month after month, and so far I have not ben able to explain it.  It is clearly creating a cash drag, so I wonder if anyone has managed to keep a better ratio of invested to originated on a stable basis, and if so is there a criteria that affects this in a meaningful fashion?

bobeubanks

  • Sr. Member
  • ****
  • Posts: 273
    • View Profile
Re: Cancelled notes: any analysis on those
« Reply #1 on: October 22, 2013, 07:21:37 PM »
I have not but would be interested in seeing any pattern too. It is indeed very annoying. I had a loan stay at stage 3 for over two weeks before it cancelled. Loans often go from stage 1 to 3 after a week and then an hour later cancel. I had a stage 3 from a previous borrower cancel after about a week. Really? I thought a repeat borrower at stage 3 would be a cinch to fund.

Kombinator

  • Full Member
  • ***
  • Posts: 193
    • View Profile
Re: Cancelled notes: any analysis on those
« Reply #2 on: October 22, 2013, 07:42:27 PM »
Well the thing is the notes that are Cancelled get fully funded, so that is not the problem.  Seems something happens during the approval process, either the borrower does not supply required documentation, some initial information does not check out, the cancellation is part of Prosper's process, not the funding...it may just be that this is something that should be addressed during the initial screening process by the platform itself, so as not to cause such a large degree of cash drag for the investors.  The percent seems steady, so perhaps they can tighten their initial criteria and thereby eliminate this odd dynamic.

agd

  • Newbie
  • *
  • Posts: 27
    • View Profile
    • Email
Re: Cancelled notes: any analysis on those
« Reply #3 on: October 22, 2013, 08:37:51 PM »
I've analyzed this, but was unable to increase my strategy's ROI by a meaningful amount to warrant the increased complexity. 

Some of the drivers that I identified are below.  Any cancellation rate that I reference is based on listings started between 6-1 and 10-1 and divide cancellations by any non-withdrawn application.  The average rate during this time period is 29.5%.

1. applicants with previous prosper loans have a very low cancellation rate. 13.1%
2. channelcode (if you are using the API): 50000/90000 have elevated cancellation rates. 35%/37.3%
3. employmentstatusdescription = 'Other' has elevated cancellation rates. 44.2%

There are a lot of other minor drivers (homeownership/DTI/inquries).

Kombinator

  • Full Member
  • ***
  • Posts: 193
    • View Profile
Re: Cancelled notes: any analysis on those
« Reply #4 on: October 22, 2013, 11:31:12 PM »
Interesting, the fact that employment description "Other" have higher cancellation rate makes good sense.

Can you please say a bit more about the channelcode 50000 and 90000, I see the field as Listing/ChannelCode, is there more detailed info on what do these codes stand for, the API documentation says that these are channels from where the applicants came...

Great info btw, thank you for sharing.
« Last Edit: October 22, 2013, 11:38:56 PM by Kombinator »

agd

  • Newbie
  • *
  • Posts: 27
    • View Profile
    • Email
Re: Cancelled notes: any analysis on those
« Reply #5 on: October 23, 2013, 02:54:03 PM »
No problem.

Prosper keeps the channelcode definitions very close to the vest.  I think its awesome that they've released what they have.  For someone with a marketing background its pretty easy to infer what channels they represent based on the credit skew and the peaky nature of some of the channels.

Kombinator

  • Full Member
  • ***
  • Posts: 193
    • View Profile
Re: Cancelled notes: any analysis on those
« Reply #6 on: October 23, 2013, 05:11:43 PM »
Yep, I asked them, but they did seem to not want to disclose it too much.  I guess I will start keeping track of these for my portfolio to see how much of 50000 and 90000 I actually get and than may determine if I stick it in my filter matrix.

Thanks again.

MarinBB

  • Jr. Member
  • **
  • Posts: 86
    • View Profile
Re: Cancelled notes: any analysis on those
« Reply #7 on: October 23, 2013, 08:20:35 PM »
I have also been wondering about Cancelled applications but hadn't had the time to look into it yet. My bet would be that a lot of loans fail to originate because the borrower's income couldn't be verified within a reasonable time or because the borrower didn't supply the proper documentation. This would also be consistent with agd's observation that loans with EmploymentStatusDescription == "Other" have a high cancellation rate. Maybe people employed in an "Other" status have a hard time documenting their income.

I'm actually happy that so many loans fail to originate after funding fully. It makes me more confident that Prosper is putting in an effort to weed out listings that don't fit the bill. I think that it makes business sense for them to do so because dealing with delinquent borrowers must eat up a lot of man hours on their behalf. It's way more profitable for both Prosper and the lender if the borrower makes each an every payment on time via an auto draft.

To put it another way, if you had a portfolio only of loans that would have otherwise been cancelled by Prosper, would you be happy? There's a lot of adverse selection here.

Kombinator

  • Full Member
  • ***
  • Posts: 193
    • View Profile
Re: Cancelled notes: any analysis on those
« Reply #8 on: October 24, 2013, 12:09:09 AM »
I have no problem with the loans being cancelled by Prosper when they cant get the documentation they need, that is indeed an excellent part of the underwriting process.  I am just trying to determine if there is a way to predict that the loan will be cancelled using some variables so I can reduce the cash drag.  Perhaps a combination of factors such as channel/"other" employment/no prior loans + some others may provide a compelling threshold.