A Sneak Preview of p2pxml.com – An Open Standard for P2P Lending Data

Posted

An Open Standard

As the P2P lending Community grows there is going to be an increased need accessing organized data. Developing a meaningful format for data will require a few key elements:

  1. Core metrics
  2. Flexible
  3. Relational

Core Metrics

There are some core commonalities between all loans. Interest, amount and term are probably the most important. Core metrics are considered data elements that every loan must have regardless of platform. Some examples are loan amount, percent funded by investors, grade, interest rate, term, origination date, application date and modification date.

Beyond that, we want to know how to assess the risk of the loan and derive meaningful data. In an ideal world the interest rate would perfectly reflect the risk. However, we are not in that world. Part of understanding these inefficiencies started with the pioneering work of LendStats.com. In an effort to augment this work, Nickel Steamroller has developed an open standard based on XML for P2P lending so that the community can start to examine platform data themselves.

Flexible

Not all platforms will collect the same data in their application/underwriting process. A result the export format needs to be flexible.  Part of the specification calls for a “Meta” tag that allows any free-form data to be entered that doesn’t fit into the core metrics. This will allow data like borrowers profession, or other data to be stored.

Relational

Most data in real life has relationships, and P2P lending is no different. The Lending Club exports are you know are very “flat”, meaning there is no relational data.  The simplest example of a relation in terms of P2P lending is payment history. Many payment events can be related back to one loan. XML has the capacity to do this well, which is why it was chosen.  The P2PXML specification fully supports payment history, which Prosper provides.

The specification is still in draft format. It will be subject to many changes as a result of your input. It’s a design goal to build a format that will be robust and platform agnostic, with the P2P lending community I believe we can.

Sneak Preview Screenshots

 

Sample Data / Launch Date

I’ve placed a sample export online of the first 1,000 records from Lending Club’s platform (Sans meta data for now). Feel free to download it and take a look. P2pxml.com should be launched in March with full support for both platforms and a few client libraries for developers who wish to start using the data.

 

Be Sociable, Share!
  • Randy Merrill

    This is an awesome contribution. I’ve been contemplating building a small site to do some of my own analysis. The data importing is a large process, especially when you have to do a separate process for each of the services.

    I would be very interested in helping with the spec and getting things going.

    Is this something that you have talked with Prosper and Lending Club about? It would be nice to have the actual sources of information follow the same spec out of the box.

    Also, the sample export has bad XML formatting. The entities are not being escaped in the xml text. ex: & -> &

    • http://www.nickelsteamroller.com/ Michael

      Randy, any help would be greately appreciated. I plan to work with Prosper and Lending Club to the extent they are willing to help out. I believe there will have good feedback.

      With respect to the amperstand, I did run all the loan descriptions through an html encoder but I did not do it on the title which I believe you might have found. I added the encoding to the title now. Did you see them anywhere else?

      I’ll shoot you an email to figure out how we can work on this best.

      • Randy Merrill

        I think that probably did it. The only reason I noticed right off is that I tried to open it and it opened in FF by default with the encoding error. :)

  • http://twitter.com/SocialLoans Peter Renton

    Great job Michael. As we discussed the datasets are so different right now that I think something like this is sorely needed.

    • http://www.nickelsteamroller.com/ Michael

      Thanks Peter.

  • Randy Merrill

    Looking at your sample XML some more, you are missing the actual value for the InterestRate and MonthlyPayment. Right now they are both the same as the LoanGrade.

    • http://www.nickelsteamroller.com/ Michael

      Sorry about that. I resolved it and published a new sample set. Thanks

      • Randy Merrill

        I tried to download it again… but I believe that there is a permission error happening on S3: AccessDenied :)

        • http://www.nickelsteamroller.com/ Michael

          Fixed. apparently re-uploading resets the permissions.

  • Pingback: Roundup of Social Lending News – February 11, 2012

  • Anonymous

    On LendingClub at least, the Payment History is related to a Note and as far as I know the only way to access it is through a Note.

    This complicated in that LendingClub does not make this information directly available from just a LoanId. You also need an OrderId. It’s possible to look at your own notes, or those for sale on FolioFn and come up with that information but that’s always going to be an incomplete list.

    This is further complicated by the fact that a $50 Note for a loan does not always match a $25 Note for a loan. I would think the values would be exactly double, but they’re not. Compare:
    https://www.lendingclub.com/foliofn/loanPerf.action?loan_id=478120&order_id=1942089
    https://www.lendingclub.com/foliofn/loanPerf.action?loan_id=478120&order_id=1942090

    In any case, it might be nice if you added a OrderId field, or a URL, or some way to access a representative note.

    LendingClub also makes Credit Score over time information as well as “Collection Log” activity available at that level. This information seems like it can be useful. Is there any room for those in your format?

    (I don’t know enough about Prosper to know if they make similar information available or what it might be called on their site.)

    • http://www.nickelsteamroller.com/ Michael

      This is a great point. I had not really considered making the export format for notes. I could potentially add a fractional note value field. One thought on the order ID is that I’m trying to be platform agnostic and the order ID is specific to the investor, not the platform. This is a very interesting tangent though. I will give it some thought. I like it.

  • Pingback: P2PXML.com Has Launched | Nickel Steamroller