Custom Loan Filters
On of the first questions that most P2P investors have is “What filter should I use to select loans that will maximize return?” In this article, take a look at how to use Peer Lending Server to create and refine your custom filters. Filtering loans are typically the most difficult portion of the peer to peer lending learning curve. Conversely, it is probably the most rewarding if done properly. We designed peer lending server to take the hassle out of maximizing return with easy to use dynamic filters.
A little Background
Lending club provides a detailed credit history and profile for each borrower. Some of the information provided are FICO score, Home Ownership, Annual Income and much more. As an investor, the information is provided to assist underwriting each loan. However, manually combing through all of the information is difficult and time consuming. It’s almost impossible even for the most experienced underwriter to review each loan and make comparisons. Making matters worst, you risk losing prime loans because others have already completely funded the loan prior to making an investment decision. Therefore, it is important to have predefined filters to automatically select loans. As new loans are listed, a properly designed filter can significantly help when comparing loans.
What is Your Personality
Most novice P2P investors start with safe A-C grade loans and eventually add more risk with grade D-G. Personally, I was no different. I started with A-C loans and slowly added more D,E and F loans. Later, we’ll take a look at why I followed this path like so many others before me. The first question you may want to ask yourself is what are you goals? For example, if you want to maximize your return and throw caution to the wind, you may want to select loans above 2o% interest rate. On the other hand, maybe you want to be a benevolent philanthropist and help others get their small businesses up and running. In this case, you might select loans with the purpose of starting a small business. The options are limitless, and is all to often the main reason why people don’t know where to start. When creating a filter, it helps to know what your target is that caters to your personality profile.
Take the Easy Route
Sometimes it helps to have GPS navigation to help find your way. In PLS, you will find preset filters that make loan selection a piece of cake. We have gathered popular filters from the Internet as well as developed some of our own. Simply assign a preset filter to one of your custom filters. Or, choose multiple filters to combine the benefits of multiple strategies.
Lets take a look at a typical scenario for a new investor. She deposits $20,000 to ensure diversity based on LC’s claim that no one has every lost money when invested in at least 800 notes. She starts looking at some notes and over time wishes there was an easier way. She checks out LC’s build portfolio tool and realizes that it essentialy spreads her money across all currently available loans. She realizes that is not such a good idea because the next batch of loans offers more choice (and she is right). She reluctantly decides to use a tool to automate her orders based on the amount of effort required, and realizes that she has to hand over her password to a third party. She definitely thinks that is not such a good idea either (and she is right). In addition, she has no idea what filter to use to get started. With PLS preset filters, not only are you provided with the portfolio’s current return on investment but also the projected return on investment. This is a unique feature, and provides real value in gauging the worthiness of a filter. You can also tweak and fine tune the preset filters to match your own strategy. Preset filters are a good choice for those that are new to peer lending and wish to start investing right away.
Developing a Filter From Scratch
Developing a filter can take considerable time as well as lots of trial and error to get one that you are comfortable with. That is, unless you have PLS to quickly drill down into each variable in a way that shows relative impact to the historical portfolio performance. Lets use an example to illustrate how to develop a filter that targets your overall objectives. As a new investor, You want to take a conservative approach and wish to make a return of 8% annually. Knowing that some loans will eventually default, you decide to set your minimum interest rate threshold to 10%. To do that, simply open a custom filter window and enter 10 in the minimum interest rate field and save the filter. Now, lets take a look at the portfolio’s overall performance
Looking at the instantaneous return on investment (iROI) we see that all historical loans with a borrower’s interest rate of 10% or higher is yielding 8.3% with an average loan age of 11 months. The projected return based on statistical models for this portfolio is is 5.1%. Obviously, this simple filter may not be a good choice to reach our 8% expected return. So lets look at how we can improve our filter. Clicking the “Purpose” tab in historical analytics shows us the projected return on investment (pROI) is given our 10% minimum interest rate by purpose.
We simply click the pROI header to sort in descending order, and quickly notice that we my find better results if we fund loans that are not Vacation and below. We simply change our filter to only include every purpose above Vacation(House, Major Purpose, Home Improvement, Debt Consolidation, Wedding, Car and Credit Card) and re-run analytics.
With this filter modification, we improved from pROI of 5.1% to 5.75%. Lets take a look at Inquiries in the Last 6 Months.
Sorting by pROI, we can quickly see that it is probably a good idea to filter loans with less than 5 inquiries. Using this strategy we can develop a base filter using historical evidence. For sake of brevity here is the final filter in text format:
- grade=’a’ or grade=’b’ or grade=’c’ or grade=’d’ or grade=’e’ or grade=’f”
- purpose=’car’ or purpose=’credit card’ or purpose=’debt consolidation’ or purpose=’home improvement’ or purpose=’house’ or purpose=’major purchase’ or purpose=’wedding’
- home_ownership=’mortgage’ or home_ownership=’none’ or home_ownership=’own’ or home_ownership=’rent’
- annual_inc>=30000 and (emp_length=’1 year’ or emp_length=’10+ years’ or emp_length=’2 years’ or emp_length=’3 years’ or emp_length=’4 years’ or emp_length=’5 years’ or emp_length=’6 years’ or emp_length=’7 years’ or emp_length=’8 years’ or emp_length=’9 years’ or emp_length='< 1 year’
With this base filter we are closer to our target of 8% target with a iROI of 9.26% (increase from 8.3%) and projected ROI of 6.02% (increase from 5.1%) with an average age of 10 months. Also, notice the amount of loans dropped from over 127,000 loans to just over 100,000. Now, lets complement our filter with machine learning. Lets only invest in loans that have at least a 90% predicted probability of being fully paid.
Simply adding this minimum probability threshold improved our instant ROI to 11.22% and pROI to 8.55%! Also, notice the loan pool shrunk from 100,000 to just over 31,000. This illustrates the much more dynamic and selective nature of artificial intelligence, while still providing a large loan selection pool for investing in current notes.
The above example literally takes a few minutes in PLS compared to the exhaustive amount of time required to create manual filters. There are significant benefits to using proven machine learning technology combined with a base filter to obtain increased alpha. In fact, many current PLS users use their already existing filters they have been using in the past, and simply add a minimum probability threshold to improve performance and confidence.