The reason these numbers don't agree is that they are different things. Much like your local temperature and barometric pressure.

"Note investors generally pay us a servicing fee equal to 1% of payment amounts received from the borrower.

First we need to clarify vocabulary. That's "note investors" which INCLUDES both fractional-loan notes and whole-loan notes. Anyone who buys notes pays this 1% OF AMOUNTS RECEIVED.

Next, we need to talk about what percentages mean. It is critically important that you take great care to understand what a percentage is OF. If you compare my salary as a percentage of GDP vs your salary as a percentage of your rent, the two percentages may both represent something about salary, but you can't compare the two numbers.

Note payers pay a percentage

of AMOUNTS RECEIVED.

Whole loan purchasers pay a monthly servicing fee of up to 1.3% per annum of the month-end principal balance of loans serviced.

As I mentioned above, these whole loan purchasers are NOT the guys who buy notes representing whole loans, on the marketplace. These are BANKS who have the regulatory systems and whatnot to allow them to buy the actual loans without the intervening note. The loans these guys buy are not visible to you on the API at all.

Loan buyers (banks) are charged a fee which is a percentage

OF MONTH END BALANCE.

Certificate holders do not pay a servicing fee, but pay a monthly management fee of up to 1.5% per annum of the month-end balance of assets under management."

Similarly, folks who buy via certificates rather than notes are charged a fee which is a percentage OF MONTH END BALANCE.

So the fees paid by retail investors (ie note investors) are computed an entirely different way than the fees paid by these institutional guys.

Therefore, you can't compare the numbers directly.

Here's how you can compare. I'll show you a way that avoids the complicated math associated with declining balance loans and payments. Look at your monthly statement. Look at total fees paid, and look at your month end balance. Divide fees by month end balance. This will give you the monthly fees pad as a fraction of month end balance. Now multiply by 12, because you pay this every month. Now you have your annual fees paid as a fraction of end balance.

You will be shocked. If you are pretty much fully invested comes out around 0.35% . We've calculated this exactly like the fees are calculated for instutional guys, and we've come out with a smaller number.

Retail customers pay much lower fees than mid-sized institutional guys!

Now you're probably confused and disbelieving that the number came out so small. Here's a way to get your head around why it comes out so small. I don't want to do the math for actual loans and actual interest rates, so lets consider a 36 month loan at 0% interest. The 0% interest just makes the math real simple. A different interest rate will make the number come out slightly different, but not much. So you invest $1 in this loan, and you get paid back OVER THREE YEARS. The first year you get paid $0.33 ... Now when LC takes their 1% of those payments, that's 1% of 33 cents, ie 0.33 cents of fees in the first year. $0.0033 is what percentage of your original $1 investment? It's 0.33% The 2nd and 3rd year will come out exactly the same. About a third of a percent.

Retail investors pay lower rates than most institutions.

I've been collecting data from the API, and the actual service fee rates (averages) look like this:

To understand this we first have to understand what that number provided by the API is exactly. It is not well documented. They used to display this number on the web site, but I think they dropped it awhile back, as most people didn't understand what it was. It is the impact that the fees have on YTM of this loan. A loan yields less when some guy is taking a little bit out of every payment.

Not surprisingly, because these numbers represent something different than the numbers I described above, they aren't equal to those other numbers, just like your local temperature is not equal to your zip code or barometric pressure.

You can reproduce these numbers by making a spreadsheet for a loan, and then removing 1% of each payment, and then computing the IRR (YTM is the IRR of a loan's cash flows). Taking out the 1% (of payments) fees will make the IRR drop by this amount. The numbers you get match the numbers from the API.

These numbers vary depending on the maturity of the loan, and the interest rate. Nothing else. It just flows from the math.

This number is worthless.

I'm puzzled about how this corresponds to the written policy. Fees on F look to be significantly lower than 1%, and W fees are even lower. Can someone explain?

I think the big thing is that you took three different numbers, which will never be equal, and didn't realize that they are three different things.