As regular readers are aware, I’m gradually increasing my portfolio allocation of Peer To Peer Loans (P2P) to 5%, or about 20% of my fixed income allocation.  I rolled over Roth IRA money to Lending Club this fall and have been selecting loans to invest in.  The biggest downside to P2P has reared its ugly head- the time issue.  I’ve been investing $25 per loan to maximize diversification.  With the $10K I’ve rolled over thus far, that’s 400 separate loans (plus the 50 I’ve got in a taxable account).  So I’ve had to log-in frequently to select loans.  Now a fair amount of loans that I select, perhaps 1/3-1/2 of them, aren’t ever finalized because the person doesn’t meet final approval with Lending Club, probably due to lack of income verification or something similar. So that means to get 400 notes, I might have to select 600.

Why Does This Matter?

This didn’t seem like a big deal initially, I mean there are always hundreds of loans available on the platform to choose from.  But the longer I do this, the more hassle it becomes.  The main reason is because I’m not willing to settle for average returns.  It’s pretty easy to go to a site like and find out which loans gave better returns in the past.  For example, loaning to mortgage holders, doing only 5 year loans, and ensuring a low number of credit inquiries in the last 6 months all increase returns on the notes.  The problem is that every time you use one of these filters to try to increase your return, you decrease the number of loans available on the platform.  This becomes a huge deal if you want to invest a lot of money, want to maximize diversification, and don’t want to have to log in to the website every day for months to get a lump sum invested.

Balancing Return With Availability

I thought I’d take a look at how these two factors interact.  So I took a look at all of the loans made at Lending Club from August 2009 to August 2012.  It turns out there were 65,704 loans, with an average return of 7.1%.  This “index return” is what I’d like to beat.  I’d like to have long-term returns of 8-12%.  The only way to do that is to use multiple “filters” to select the better performing loans.  I went through all the filters that offers, especially those which can be automatically filtered on the Lending Club site.  I discovered that if you only purchase lower quality 5 year notes (D-G) from mortgage holders with no inquiries in the last 6 months, with a credit score less than 678, with a credit history of at least 10 years, employed for at least 2 years, without any public records, and only from the best 22 states, that you can increase your return from 7.1% to 17.80%!  That sounds fantastic!  What’s the catch?  Well, over those 3 years, only 117 loans met all those criteria, or about 0.2% of all loans.  As I write this, there are 862 loans available on the platform.  862* 0.2%= less than 2 loans at any given time.  And remember that loans stay on the platform for up to 14 days before closing.  So that could be as little as 3 loans PER MONTH that meet my criteria.  Buying 400 loans at the rate of 3 per month will take over a decade.  That’s obviously not a viable strategy.

So the investor is left with a choice.  He can either decrease diversification and put more money into the good loans, or he can become less selective in choosing notes (or some combination of the two.)  Since I couldn’t use all the filters, I wanted to know which ones increased return the most while decreasing the number of notes to invest in the least.  This is what I found:


Filter Return Additional Return # Loans % of Total Additional Return X % Total X 1000 Screenable


B-G 7.58% 0.48% 49776 75.8% 3.64 Yes
D-G 8.43% 1.33% 15587 23.7% 3.16 Yes
Mortgage 7.60% 0.50% 29535 45.0% 2.25 Yes
CreditCard/Debt Cons 7.85% 0.75% 34421 52.4% 3.93 Yes
Inquires 0-2 7.54% 0.44% 59242 90.2% 3.97 Yes
Inquiries 0-1 7.63% 0.53% 50070 76.2% 4.04 Yes
Inquiries 0 7.81% 0.71% 32074 48.8% 3.47 Yes
60 month term 7.80% 0.70% 16361 24.9% 1.74 Yes
DTI > 25% 9.29% 2.19% 2833 4.3% 0.94 No
CS <678 8.50% 1.40% 12778 19.4% 2.72 Yes
CS<713 8.16% 1.06% 37895 57.7% 6.11 Yes
Max Loan 20-35K 8.52% 1.42% 11585 17.6% 2.50 No
Earliest Credit > 5Y 7.16% 0.06% 62848 95.7% 0.57 Yes
Earliest Credit >10Y 7.27% 0.17% 47041 71.6% 1.22 Yes
Earliest Credit > 15Y 7.68% 0.58% 23761 36.2% 2.10 No
>8 Open Credit Lines 7.64% 0.54% 36697 55.9% 3.02 No
>15 Total Credit Lines 7.39% 0.29% 46230 70.4% 2.04 No
Balance > $25K 8.25% 1.15% 8871 13.5% 1.55 No
Balance > $15K 8.10% 1.00% 21196 32.3% 3.23 No
Balance > $10K 7.94% 0.84% 32986 50.2% 4.22 No
Balance > $5K 7.61% 0.51% 48647 74.0% 3.78 No
Utilization > 30% 7.64% 0.54% 50166 76.4% 4.12 No
Utilization > 40% 7.78% 0.68% 44003 67.0% 4.55 No
Utilization > 50% 8.00% 0.90% 36766 56.0% 5.04 No
Utilization > 60% 8.15% 1.05% 28969 44.1% 4.63 No
Utilization > 75% 8.12% 1.02% 16513 25.1% 2.56 No
Utilization 60-80% 8.27% 1.17% 16518 25.1% 2.94 No
Months Since Delinquency -7.10% 0.0% 0.00
Income > $5000 7.92% 0.82% 33044 50.3% 4.12 No
Income > $6000 8.38% 1.28% 23158 35.2% 4.51 No
Income > $7000 8.77% 1.67% 16230 24.7% 4.13 No
2 Year DQ 1-3 7.66% 0.56% 6834 10.4% 0.58 Yes
2 Year DQ 4+ 7.93% 0.83% 222 0.3% 0.03 Yes
No Public Records -7.10% 62821 95.6% High
Location 8.07% 0.97% 28339 43.1% 4.18 Yes
Employed > 2Y 7.30% 0.20% 52421 79.8% 1.60 Yes
Combo 1 11.37% 4.27% 1473 2.2% 0.96
Combo 2 10.49% 3.39% 8649 13.2% 4.46
Combo 3 12.08% 4.98% 2737 4.2% 2.07
Combo 4 17.80% 10.70% 117 0.2% 0.19
Combo 5 12.76% 5.66% 1750 2.7% 1.51
Combo 6 12.28% 5.18% 3797 5.8% 2.99


Let me explain a little bit.  Looking at the first row in the table, you see that if you just get rid of all the AA and A loans, that you increase your return from 7.1% to 7.58%, an increase of 0.48%.  Unfortunately, that also eliminates nearly 25% of the loans on the platform.  In an effort to relate those numbers to each other, I multiplied them together, and then multipled by 1000 to make it an easy number to work with.  This gives me 3.64, which is pretty high as these filters go.  The last column indicates that this process is automatable using the filters on the Lending Club site.  Unfortunately, many of these filters aren’t automatable on the site, so that requires more time for you to look at the loans individually and be your own filter.

Now, compare buying only B-G notes with selecting only 60 month notes.  The average 60 month note returns 7.8%, an increase of 0.7% over the “index return” of 7.1%.  That’s a higher return than you’ll get by using B-G notes.  Unfortunately, that filter removes about 75% of the notes on the platform.  Once you multiply the numbers together (3.64 vs 1.74), you see that you’re probably better off using the B-G criteria than the 60 month loan criteria, especially when combining multiple filters together as most investors are apt to do.

By doing this with all of the possible filters about which I could find return data over this time period, I started looking into various combinations of factors.  I first looked at what I had been doing- basically D to G loans, mortgage holders, 60 month loans, >5 year credit history, >$6K per month income, no public records, employed at least 2 years, credit card/debt consolidation only, and discovered why it was taking forever to invest my money.  Only 2.2% of the loans on the platform met criteria!  864 loans * 2.2% = 19 loans every couple of weeks.  At that rate it could take me well more than 6 months to invest $10K into $25 notes.  There’s an additional problem with this- that cash drag on having money sitting in the Lending Club account not earning money can be significant over long periods of time.

At that point I started playing with the filters trying to only include those with a relatively high multiple, 3.5-4 or higher.  I wanted to get higher returns, but I also needed notes to invest in.   I eventually settled on the combination above entitled “Combination 6.” Over the three years I studied, this combination increased returns from 7.1% to 12.28%, which is actually higher than what I was doing, which only increased returns to 11.37%.  The icing on the cake is that instead of only having 2.2% of the loans on the platform meeting my criteria, now 5.8% do.

My New Criteria

So what is Combination 6?  It mostly includes the filters that gave me the highest multiple of the increased return by the % of loans remaining after the filter.  I just needed some type of equation that related the two items to one another.  So I included the following filters:

Automatable Filters

  • Grades B to G
  • Purpose: Credit Card and Debt Consolidation only
  • Inquiries: 0-1 only
  • Credit Score: < 714

Non-automatable Filters

  • Credit Card Balance: > $10K
  • Credit Utilization: >50%
  • Monthly Income: >$6000

When To Be Picky

When you’re just reinvesting the proceeds of your notes, and thus only need to choose one or two notes, you can afford to be pretty darn picky, especially if you haven’t been on the site for a couple of weeks.  If you just rolled over $5K into a Roth IRA, you’ll have to be less picky.  So perhaps I ought to use “Combination 6” when I need a lot of notes, and add a few more filters when I don’t need very many.  For example, eliminating B and C grade loans from Combination 6 increases returns to 12.45%, although it decreases available loans to 3.4%.

A Moving Target

Remember, of course, that nothing is guaranteed.  This is data-mining at its finest, and from a pretty limited data set to boot.  For example, I found if you only lend to people in 22 states, that you can increase returns by 0.97% and only eliminate 57% of the loans.  It turns out that loaning to Alaskans was far smarter than loaning to Utahns over this three year period.  But there seems little rhyme or reason to the list of the good states versus the bad states.  (In case you’re curious, the good states were AK, AL, AR, CT, DC, DE, HI, IL, KS, LA, MA, MD, MT, NH, NY, OK, PA, RI, TX, VA, VT, and WY.)  So some of my results are probably similar to the nonsensical predictions people use for the stock market, you know, the price of butter in Bangladesh or the conference of the Super Bowl winner.  If you look at enough data, you’ll see patterns that aren’t really there.

To make matters worse, Lending Club changes their methods from time to time, which makes our limited sample of past data even less useful than it is.  There are loans in the database that Lending Club wouldn’t make at all now.  The reason some of these filters work isn’t that the notes filtered out are necessarily better credit risks, but that the yield given to you for taking that risk is higher than it should be.  Basically, Lending Club penalized people too much for that low credit score, for instance.  As Lending Club tweaks its model, some of those factors are bound to change.

Also keep in mind that many of these loans I’m looking at haven’t yet run their course.  Returns are almost surely bound to be lower than projected here since some of the loans are only 3 months old.  In fact, only

New Tools

invest, investor, investing, lending

While I was writing this post, I was alerted to a new tool called Nickel Steamroller. has been having a few bugs lately, so I tried out Nickel Steamroller.  It actually has a feature called “Current Listings” which lists the loans currently available to purchase on Lending Club and assigns them an expected return based on loan length, FICO range, loan purpose, total funded amount +/- $2K, delinquencies, and inquiries (all from data-mined past loan data). also hasn’t been keeping up the Prosper side of its site.  You can get Prosper data at

Two Other Possible Solutions

There are two other ways to deal with my dilemma of trading return for diversification.  The first is to keep half your P2P investments at and half at  This way you can select the highest returning notes from each.  It obviously increases your portfolio complexity, but might be worth it for you.  The second is to purposely decrease diversification.  Is there really a significant benefit to having a thousand $25 loans instead of five hundred $50 loans?  Probably not.  In fact, when deploying a significant sum of money, you could put $75-100 into loans that met your strictest criteria, and then $25-50 into loans that met most of your criteria.  This would likely boost returns even further as the portfolio is tilted toward the loans with the highest projected returns.

I’m obviously learning more about P2P as I go.  It’s fun to have an investment to tinker with where the tinkering seems to actually improve returns (unlike stock picking.)  I’m perfectly happy with my returns so far at Lending Club (around 10%).  But the more I can automate the process the better.  Do you have any tips for making your P2P investing more automatic without sacrificing returns?  Sound off below!