OT - Upstart (UPST)

Stocks A to Z
21

“I bought some more today at $29.8-30.2.”

If you are planning these buys with long-time horizon, then there is no need to rush wait for the dust to settle. If you think sell off is excessive, then be prepared to immediately book your gains.

I am buying with a long time horizon, and I think the sell off was excessive. :slight_smile:

I am aware that my investment thesis could be wrong, but I haven’t bet the house, or anything close to it.

You told me the other day that if I thought Upstart would survive the recession, it’s a buy at $95 just as much as $75. I deferred on that thinking, having a comfort level at $73-75 where I’d purchased it that I didn’t have at $95. Obviously that was before Upstart lowered expectations. With the newly published expectations, I’m comfortable buying at $30 to hold and of course have some regret about the earlier higher purchases, but we make decisions with the info we have on hand, so not really any real regrets.

Thank goodness I’m not someone who had 30% of my portfolio in it, lol.

1 Like
22

I haven’t looked at UPST close enough to understand this. One of the risks, I am seeing or banks/ Insurance companies (to some extend) is mark to market of their bond portfolio. Bonds are going down heavily. Unlike equities they are not expected to move so much and also heavily leveraged. I am expecting many banks are going to take a hit on their TBV when they mark to market their bond portfolio (at least the ones they marked for sale).

This hit is going to flow through the balance sheet of banks, anyone who does ABS (like carmax), or MBS like UPST etc. Have you looked into it?

23

Ben, if you haven’t read and reflected upon this post, you may want to do so:

https://discussion.fool.com/upst-not-hypergrowth-anymore-3510712…

1 Like
24

Ben, remember reading quarterly reports is the over-the-top intense theme of Saul’s board. Saul just again posted that valuation is absolutely not allowed on his turf. Fair game. But quarterly report investing is librarians trade, it should reward in the librarian’s pay scale.

What my friend to these numbers mean? Tell me precisely the intrinsic value of UPST based on quarterly numbers, the obsession of them and the fact they must go up exponentially.

Ben, the post sited above is good…but I tend to do more global thinking as to models. Here it goes and if you can’t read between the lines or it is very likely the patsy has been identified. Oh, one last thing, UPST stock price will gyrate with incredible intensity so it is a game you may win with big time in the short run.

Here we go, here’s the game:

Right now old man dealraker is quite certain he (that’s me) can sell his waterfront home for well over $1 million. Yep, within a week it would be gone.

And, food for thought, in the last 20 years there has been two times when I’m pretty sure I’d have had trouble giving my home away. We had an 80 acre tract here on my lake sell for $5 million in 2006, then sell for $800 thousand in 2019.

UPST, the business, will have bizarre ups and downs. UPST is a business, a vary rare exception, where the business is very likely to be far MORE volotile than the stock price.

That is a rare business right there. If you want excitement? Buy UPST!

I just looked at my Toosie Roll (bought in the 1980’s) and Mondelez (bought Cadbury in the year 2000) holdings. Oh lord those two have some history, the stocks stagnate seemingly forever at times. Both get overvalued and only fall to fair value.

The profits I have in those are damn rediculous. The business of both is steady Eddie. Boring, boring, boring.

2 Likes
25

Stock is down 21% today. I accumulated some more, averaging down.

The investment thesis is simple. Do ML models work and perform better than FICO or not ?

Long discussion of Revenues, Balance Sheet, Loans on books, Risks, Lending Partners, Income is important but it all sorts out if the models work and get better over time.
If ML models don’t work then the business collapses.

1 Like
26

I was a director of two banks in my time. Here’s my interview:

Upstart founder: “We’ve got this AI, this deep information package, that we only charge an upfront 5-8 percent of the loan for…that going to cut your loan losses big time based on a short period of history. It is THE best.”

Me: "Somebody bring me a 5 gallon bucket please!

Upstart founder: “What do you need that for?”

Me: “I’m going to fill the damn thing up with tears I’m laughing so hard.”

In summary:

What percentage of Upstart’s loans will sour in a downturn? My guess is the percentage will be beyond hilariously high and the AI program they sold is absolutely worthless.

4 Likes
27

What percentage of Upstart’s loans will sour in a downturn? My guess is the percentage will be beyond hilariously high and the AI program they sold is absolutely worthless.

Upstart customers are liking the ML results as it is outperforming the FICO scores model.
11 out of 57 have abandoned the old school FICO score model due to its underperformance.

This is all that matters. Other opinions are irrelevant.

3 Likes
28

Nah divvy, not even close.

Next chase please! We need some new blood; SAAS stocks are pathetic.

29

Quick, why don’t you bring out the Cisco example…again.

1 Like
30

<Upstart customers are liking the ML results as it is outperforming the FICO scores model.
11 out of 57 have abandoned the old school FICO score model due to its underperformance.>

How long has Upstart been in loan business?

1 Like
31

divvy the Upstart model is much more simplistic, yet exponentially more sensational, than the Cisco one. Remember the experts over on Saul’s were euphoric about Upstart and now it is down over 90% since their big party and celabration…

…which of course should have been sell-a-bration to buy oil stocks.

But what the hell, you’re chasing Upstart and I’ve been allocating capital over to oil and gas for a couple of years now. Let’s be honest with one another big boy: THE GAME???

IT AIN’T OVER YET! By damn though…it sure is fun!

I’m king of guessing that those SAASSY stocks gunna have some sales disappointments upcoming. We shall see!

32

I’ve been allocating capital over to oil and gas for a couple of years now

I think you sit in your homeless shelter and make up stories.

8 Likes
33

11 out of 57 have abandoned the old school FICO score model due to its underperformance.

You are assuming somehow ML can make bad credit good credit. No. Sub-prime is sub-prime. At the end of the day, you can wrap ML/AI/SaaS model/cloud online, etc, but the underlying business is making sub-prime loan and convince rating agency they are prime and get a BBB rating and bundle the sub-prime loans and sell it to investors.

It works great when investors are starved for yield, when they chase yield, they are ready to buy any story. When you get prime MBS (mortgage based securities) at 5%, and sub-prime with make-up at 7%, anyone managing risk knows they are taking way too much risk at 7%.

UPST is still peddling sub-prime.

2 Likes
34

People get sold over and over again on the idea that AI or some other tech-driven innovation will alter the fundamentals of lending. Even worse: masquerading as something other than credit.

Maybe I don’t get it, but to me, credit is credit. The fundamentals are 4000 years old.

3 Likes
35

You guys are talking like FICO is the be-all and end-all of credit.

So you’re not creditworthy unless FICO says so?

Have you not considered the possibility that FICO is a blunt instrument that unfairly scores some customers as sub-prime?

You say Upstart is loaning to sub-prime customers, but that’s as per the FICO definition of sub-prime, which Upstart is in the process of disrupting. It’s like criticising Netflix for not having an over-the-air TV channel (I know, poor example).

No position

4 Likes
36

“Maybe I don’t get it, but to me, credit is credit. The fundamentals are 4000 years old.”

Do we really need AI or is it actually a simple issue?

It reminds me of an oil company I worked for early in the implementation of computers. They came up with 20 variables to consider, and fed them into a computer, thinking it would spit out the answer of where to drill and allow computers to replace the simpleton geologists.

It turned out that 90% of the correlation came from just two important variables, and those were the first two a geologist would look at when making a recommendation. Common sense gave a quicker answer, just as accurate, without the arduous task of loading 20 variables into the software.

1 Like
37

I would forget the AI buzzword. It’s overused. As is machine learning. There’s still lots of people making decisions and creating new models.

I would also suggest that geological formations are much less complex than human behaviour.

1 Like
38

Perhaps, though humans are better at imagining how another human will behave over the course of a few years than they are at imagining how the earth has worked over a few hundred million years.

39

It reminds me of an oil company I worked for early in the implementation of computers. They came up with 20 variables to consider, and fed them into a computer, thinking it would spit out the answer of where to drill and allow computers to replace the simpleton geologists.

That brings to mind a comment Dick Hamming once wrote at the beginning of a Numerical Analysis book he wrote long ago.

Numerical Methods for Scientists and Engineers

Richard W. Hamming (1915–1998) was first a programmer of one of the earliest digital computers while assigned to the Manhattan Project in 1945, then for many years he worked at Bell Labs, and later at the Naval Postgraduate School in Monterey, California. He was a witty and iconoclastic mathematician and computer scientist whose work and influence still reverberates through the areas he was interested in and passionate about. Three of his long-lived books have been reprinted by Dover: Numerical Methods for Scientists and Engineers, 1987; Digital Filters, 1997; and Methods of Mathematics Applied to Calculus, Probability and Statistics, 2004.

In the Author’s Own Words:
“The purpose of computing is insight, not numbers.”

“There are wavelengths that people cannot see, there are sounds that people cannot hear, and maybe computers have thoughts that people cannot think.”

“Whereas Newton could say, 'If I have seen a little farther than others, it is because I have stood on the shoulders of giants, I am forced to say, ‘Today we stand on each other’s feet.’ Perhaps the central problem we face in all of computer science is how we are to get to the situation where we build on top of the work of others rather than redoing so much of it in a trivially different way.”

“If you don’t work on important problems, it’s not likely that you’ll do important work.” — Richard W. Hamming

The purpose of computing is insight, not numbers."

4 Likes
40

Maybe I don’t get it, but to me, credit is credit. The fundamentals are 4000 years old.

Nonsense.

News is news ? Newspapers or Google
Transport is transport Tesla
Retail is retail ? Sears or Amazon

Disruption is never smooth.

3 Likes