On Deck Again

We have received a great deal of feedback to our recent On Deck post – a number of good points have been made and a number of good questions asked.  In this post, we will do our best to summarize two of the more pointed questions and attempt to offer constructive responses.


 Q:     Although bank account data and ACH have both been around for a long time, why can’t the use of technology aid in better decision making/ improving risk management?


 A:     We certainly believe it can and indeed believe that better risk management decisions could in fact be made by doing so.  Our only argument on this point is to ask the question “To what end?”  I think we can all agree that better information and better decisions don’t change the risk, they just offer the potential to help better manage it.  However, to be able to apply technology to an information set requires having enough good information – information that is similarly situated to the risk one is attempting to better manage – in a word “relevant” – or as a data person might say “robust”. 


In our view, this reality creates a chicken and egg problem for On Deck.  Making a multitude of loans to target customers could certainly provide a “robust” data set to observe, but by doing so,On Deck has to have already made numerous loans (i.e. risked a lot of capital) – and done so before being able to secure the knowledge of how that portfolio of loans will behave under different conditions.  On the other hand, On Deck could (and almost certainly has) run data simulations (i.e. tests where they analyze “mock” portfolios without having to risk their capital).  There is an emerging information technology field aptly named “Predictive Analytics” that does this type of work across a broad range of problem sets – including many beyond finance.


 There are a few problems on this proverbial “other hand”, but the primary one is that On Deckwould have to contend with a major flaw when using “mock” portfolio data.  That flaw is the simple fact that the “mock” portfolio would need actual “mock borrowers” in order to be “robust”.  Generic data just won’t do unless you are willing to argue that there is no significant difference in the behavior of a “business” that signs up for a 25%+ loan and one that doesn’t.


Herein lies the chicken and egg problem – To get the data necessary to effectively manage the risk, On Deck has to first risk the capital. Of course, if it risks the capital first, it is doing so without sound knowledge of how to manage that risk.


 Q:     Isn’t it true that direct access to the bank account via ACH is a highly effective risk management tool?  One that could be made even better with the use of technology?


 A:     Yes and yes!  We actually think that perhaps the greatest potential value creator in the entireOn Deck equation turns on the answer to these two questions.  The single best way to manage risk in the lending business is to have a definitive way of getting your money back and ACH is already a highly effective tool that enables this (if not in whole at least hopefully in part), and one that could certainly be improved with better information.


 The biggest trick (again in our view) of course is applying this technology in practice, and doing so effectively depends heavily on an enlightened understanding of various borrower behaviors that would signal potential problems; problems that would trigger an extraordinary recovery type response.  Also, an enlightened understanding of what types of recovery responses might be most effective.  There are numerous details involved, but in the end, they all bring us back to the same chicken and egg problem mentioned above.  On Deck needs “robust” information first before it can hazard an educated guess about the best way to identify and deal with various problems.  Absent this, the best way forward is an incremental and iterative one; one that takes quite a bit of time to develop properly (and safely).  The methodical path forward is the only one that has proven effective in capital deployment business models over the long haul.  At this stage, we don’t see any reason to think that On Deck’s path forward should be any different.


 We also think there are some useful analogues that can help to frame a better understanding of the potential for On Deck.  In the interest of time and space, we’ll plan another post on those.  Stay tuned.