WSM’s portfolio end of May 2021


May was an up and down month for my portfolio, and I guess I’m not alone in that. In the end more up than down.

I ended the month up ytd 6.6%.

2021 ytd results were as follows (with lots of volatility in between):

End of Jan	+	7.6%
End of Feb 	+	7.7%
End of Mar 	-	6.0%
End of Apr 	+	4.6%
End of May	+	6.6%


During May I decided to sell all of my Bandwidth shares after reviewing their Q1 results. I also exited my small ETSY, Docusign and Chainlink positions.

I redeployed the cash into Upstart. I also trimmed Crowdstrike a little and bought more Snowflake and Upstart with the proceeds.

So the big changes for me were sellling Bandwidth, which was a 8% position for me prior to exiting, and buying Upstart.


I waited until they reported Q1 results to see if there was evidence of my investment thesis coming to fruition - revenue growth acceleration and GM% improvement. - and it didn’t

In a nutshell I believed that the Voxbone acquisition would result in broader success with international customers and Zoom, a customer’s Zoom Phone product and call centre customer success would drive exceptional usage and be an inflection point for growth.

Also I thought that management’s path for gross margin improvement was plausible - with a target GM% of around 60%.

I pencilled in $120m in revenue as quite possible, using not outlandish assumptions for Q1.

Instead the following happened:

Revenue growth did not accelerate and came in at $113.4m for Q1 - flat qoq. vs the acceleration which I was expecting.

Gross margins contracted from 49% in Q4 and the Q1 2020, to 46% in Q1 2021. This was equal to the level in Q1 2019 which was the lowest it was in the last 4 years. For the first quarter of the last years, GM% looked like this:

2018 52%
2019 46%
2020 49%
2021 46%

Customer growth for Q1 decelerated vs prior years. Q1 qoq customer growth was not impacted by acquisitions and looks like this for Q1 compared to Q4 of the following years:

2019: 10%
2020: 5%
2021: 4%

ARPU continued on a slow decelerating path. Here is Q1 ARPU for the last 4 years in $:

2018 54,922
2019 39,452
2020 37,887
2021 38,324

So, all in all a rather disappointing Q1 in light of the expectations of acceleration that I had. With all of the macro tailwinds in Bandwidth’s favour, I was expecting more. In stead every metric above disappointed. In the end it was an easy decision to sell.


There has been so much written on this company on the board recently that it may seem obvious to many. But let me try to pull the various bits and bobs and my own thinking together.

First off, some links to read and watch:

IPO Letter:

“By way of example, four in five Americans have never defaulted on a loan, yet less than half have a credit score that would qualify them for the low rates that banks offer.2 The implication is eye-opening: with a smarter credit model, lenders could approve almost twice as many borrowers, with fewer defaults.”?
“While many Fintech companies aspire to become the bank of the future, we’re not one of them. Instead of competing with banks, we decided to partner with them, helping banks navigate and thrive through one of the most foundational technology transformations in history.
While a handful of our country’s largest banks may aspire to build an AI-powered lending platform, thousands of others will look to form strong and reliable technology partnerships in order to secure their future. We aim to partner with them.”

17 Mar PR on Prodigy acquisition:…

“In September 2020, the first AI-enabled auto loan was originated on Upstart’s platform. In this initial phase, Upstart is enabling consumers to refinance expensive and mispriced auto loans, saving borrowers an average of $72 per month. Following the initial launch, Upstart continues to roll this program out in states across the country.
With the acquisition of Prodigy, Upstart will accelerate its efforts to offer AI-enabled auto loans through the tens of thousands of auto dealers nationwide where the majority of auto loans are originated.”

Saul’s month-end review with details on Upstart (before Q1 results):…

“And they are guiding to $115 million this quarter, up 80% yoy and up 32% sequentially!!! And they certainly expect to beat guidance. Guidance for the full year 2021 is $500 million. That guidance is up 115% from 2020’s $233 million, and they certainly expect to raise that several times as well. Look, if they beat their first quarter guidance of $115 by $10 million, they will come in at $125 million and they will have a run rate of $500 million already! [WM: Q1 came in at $121m and they upped their guidance to $600m for the year] They imply that they are not counting growth in Prodigy in guidance”

Q1 2021 Results and prezzo:…

“Our Q1 2021 revenues were up 90%. Our profits were up by a factor of seven over the first quarter of 2020. Despite softer loan demand from consumers [WM: they are saying they did this with headwinds in the macro environment] due to government stimulus programs, almost 170,000 loans were transacted by our bank partners in Q1, more than double the volume from just two quarters ago.”

Q1 Results discussion:…

I liked this quote from WillO2028:
“The more I read about Upstart the more I see that that Upstart is Creating a vertically focused SaaS Market. They’re providing the bases on which others take financial risk.”

Thread on customer concentration risk:…

Dave Girouard on the outlook for the rest of the year May 13th 2021 (7mins):

The bull and bear case:

If I were to try to summarise the two camps’ views on this stock, it goes something like this.

On the one hand you have a PROFITABLE company with 85% gross margins growing at 100%+, led by a very capable and proven CEO who ran Google’s Enterprise Business for years, together with his two co-founders who are still at the business. They are a first-mover with an AI-enabled solution which will only get better over time and recently moved into a new market which will act as a second engine of growth. And they have a forward revenue multiple under 20 which is low for this level of growth.

And on the other hand there were two main concerns raised in the various discussions: customer concentration, and the company apparently having little competitive moat vs, say the likes of Crowdstrike (and a lockup expiration and high short interest, but that’s got nothing to do with underlying performance so I’m excluding that). I personally have a third, which is around what they do - originate relatively risky unsecured loans.

Now let me try to unpack these concerns:

1. Customer concentration

At first glance it would indeed seem that there is concentration risk as 63% of fees on loans originated in 2020 were from one bank: CRB. However if you look at the mechanics, this bank originates the loan and then most of the loans are on-sold to institutional investors.

So if one looks at CRB as part of the process/plumbing of writing the loan, then this concern becomes less imo. Many companies have a partner or multiple partners that they rely on in the value chain to deliver the end product to customers - think AWS / Google as providers of compute to most SaaS companies for example. And, yes, there is a form of risk there because of potential lock-in, but it is supplier concentration, not customer concentration risk.

In addition, this type of arrangement is in line with CRB’s mission; it seems like this is what they do. They are a partner to fintech companies like Upstart:

“Our Mission is to enable our clients to deliver responsible financial solutions that empower consumers anytime, everywhere.”

Here is a critical Dec 2019 Forbes article discussing what CRB is and what they do. In the article they highlight a couple of big risks as well, so well worth a read:…

“LendingClub went public in 2014 with a valuation of $5.6 billion. Today it is worth $1.2 billion. On Deck Capital, a New York City–based fintech that makes superfast small business loans, is worth $290 million today, down from $1.9 billion the day it IPO’d in late 2014. It’s a similar story for other fintech IPOs like Funding Circle and GreenSky.

“[These] companies positioned themselves as tech companies, [but] in reality [they] are just leveraging tech to further an old-school business solution like consumer lending,” says Andrew Marquardt of Middlemarch Partners and formerly of the New York Fed and BlackRock. “You have investors looking at it and saying, ‘This is a bank, it is not a tech company.’?”

“Today Cross River continues to expand, seemingly oblivious to the looming risks. Just as banks competed in a frenzy to issue “low doc” and low-rate mortgages while the housing bubble inflated, some fintechs have begun making riskier loans.
Last year, one of Cross River’s biggest fintech partners, Freedom Financial, agreed to a $20 million settlement with the FDIC after the regulator determined Cross River used “unfair and deceptive” practices by failing to effectively oversee its partner during the origination of over 24,000 loans. Cross River was forced to pay a $641,750 fine.

An even bigger threat to fintechs is an economic downturn.”

So how does the arrangement work?

Best I could figure it out, Upstart acts as the agent for CRB for originating the loans on the Upstart platform, in terms of a loan platform agreement - because CRB is a bank, and Upstart isn’t. CRB pays Upstart for the loans originated.

After CRB has originated the loans, CRB keeps a portion of them on their balance sheet, and Upstart buys the rest back from CRB in terms of a loan sale agreement, and pays CRB for the loan plus a fee on top:…

Following this, the loans are on-sold to institutional investors, who are the final customers buying/funding the loans.

There are two big numbers in the cash flow statement that shows this dynamic. In Q1 2021 “Purchase of loans for immediate resale to investors” and “Proceeds from immediate resale of loans to investors” - both numbers $1.3bn.

In 2020 the split (for all banks, not just CRB) was 21% were kept by the banks, 77% were sold back to Upstart and then on-sold to institutional investors, and Upstart retained the remainder, 2%. So they don’t build a book, which means they don’t sit with the credit risk.

Zoro made the point that revenue concentration risk was overblown in a thread on Upstart on the board where he quoted the CEO as saying in the Q4 CC:

”…fortunately, we’ve built our model such that we are not dependent on how many banks and how quickly these banks adopt our technology. We’ve developed a platform that flows loans through to capital markets and institutional buyers if banks are not ready and willing to fund them and hold them. And that’s why our business can grow as quickly as it does”

However, the 10-K does state that:

“CRB, originates a large fraction of the whole loans sold to institutional investors under our loan funding programs. For the years ended December31, 2018, 2019 and 2020, fees received from CRB accounted for 81%, 80% and 63%, of our total revenue, respectively.”

So CRB is making good money out of their arrangement with Upstart, as they get to keep the loans that they want, make a cut on each loan they originate, and offload what they don’t want to institutional investors via Upstart. So I’m in the Zoro camp wrt customer concentration. I do not see the CRB relationship as having the level of risk that the Fool, for example attributes to this, as CRB is not the end customer for the majority of loans that they originate but rather more of a key supplier, with the institutions the end customers.

2. Limited competitive moat

This monster thread questioning the moat of Upstart’s AI is worth reading:…

In it, Saul made the excellent point that the revenue and operational growth numbers speak for themselves - loudly - and that they therefore must be doing something right.

In addition draj pointed out that:

  1. Existing methods for evaluating loan risk are worse than poor.

  2. In general the banking industry can’t seem to deploy sufficient talent to devise a better system.

→ WM: Banks often have tons of disparate, complex and interrelated legacy systems; often acquired via M&A and therefore a mix of technologies and architectures - some very old, customised at great cost to talk to and work with one another. Changing one thing often leads to a cascading of knock-on impacts on other systems which require custom changes, consultants, unforeseen effort, cross-departmental changes and trade-offs. The issue is not so much talent (although draj’s point is probably true to some extent for that too) but the inertia inherent in the underlying systems and processes. So I agree wholeheartedly on this one. MoneyTree and mostlylong made similar points.

  1. Upstart’s AI models are barely 10% of the way on the path to “the best” predictive model.

→ WM: And they’ve been working at the problem for 8 years now with new cloud-native tech…

And Oracleoo made this point:

  1. Along with the competitive moat of AI algorithm, what is also important is the time required to build the credibility and the relationships.

3. Unsecured loan origination business

As I said, what they do - loan origination and pricing for small unsecured loans, democratizing finance for the underserved - is something that makes me pause due to how many times I’ve seen it go wrong. Even with the best of intentions. The cycle that I’ve seen play out when banks have tried to target the “bottom of the pyramid” has been one of initial success, followed by bust - when the economy turned. Something that happened while I was living in South Africa was the failure of African Bank. They did lots of unsecured loans to poorer people with non-existent/poor credit scores and claimed much lower loss ratio’s while writing lots of loans at relatively low interest rates. The traditional big banks in SA wouldn’t lend to this group, so African Bank was hailed as a model for the big banks of what financial inclusion could look like. The eventual collapse was therefore especially painful. From a HBS case study of the events, here is an excerpt from the Abstract:

“[…]African Bank, a provider of unsecured loans to help individuals pay for expenses such as vehicle repairs, home renovations, funeral costs, medical bills, and education. As ABIL’s profits rose and share price soared, it became the darling of analysts. However, when the South African economy turned sour, customers started defaulting on their loans, and ABIL and African Bank began to struggle. As ABIL’s stock price dropped, the management team faced difficult questions about the company’s lending practices, risk management processes, and plans to turn the business around.”

So that’s one example of many of what happens if you get it wrong.

However unlike the example above, Upstart does not keep the loans on balance sheet, so offloads the credit risk. And the approach is different. They use AI. And as long as the AI knows what it’s doing and does not make bad underwriting decisions, all will be well.

And that’s the crux of the matter for me. Why this time could be different to previous attempts to lend profitably to the lower end of the market without eventually burning your fingers: AI has really come of age in recent years. Having seen a lot of failures like African Bank over the past decades, it must also mean that getting it right is exceptionally hard. And therefore exceptionally valuable. So I guess I need to respectfully disagree with the arguments made on the board about Upstart’s AI constituting less of a moat as the problem is relatively easier to crack. I think it is very hard to get right at scale and therefore very valuable and confers a big competitive moat.

They have a very strong position in a wide open market if the loans do not end up eventually going bad - i.e. if the secret AI sauce is, in fact, that good. And I think it just might be.

That’s my view but the risk remains though: should a larger part than expected of loans go bad, Upstart will find themselves without buyers for the loans and will then need to seriously curtail origination, which will cut their revenues rapidly. That’s probably what happened in Q2 of 2020. I guess only time will tell on this score.


2017: $57m
2018: $99m up 73%
2019: $164m up 65%
2020: $233m (pandemic) up 42%
2021: $600m+ (at least) up 157%

In the table above I pencilled in their current guide for the full year, which they increased by $100m in Q1. The growth rate is up by a lot vs prior years. Q2 numbers will be up >800% yoy due to the easy comp in the COVID-impacted Q2 2020. Saul expects them to increase that $600m guide multiple times during the year, so the short-term potential is huge.

in terms of the drivers of that revenue, things also look good: They grew the number of bank partners from 10 in Q3 2020 to 15 in Q4 and 18 in Q1 2021. That’s huge acceleration given that the first 10 bank partners probably took them their entire 8 year existence to get, and they almost doubled it in 6 months.

The number of loans originated in the last 3 quarters were 81k, 123k and 170k - 53% and 38% qoq growth and the average revenue per loan has remained relatively constant, going from $808 to $703 and $712 in Q1 2021, meaning that the key metric driving revenue growth is the number of loans originated, which should increase given the increased number of banking partners.

And they’ve just started moving into auto loans with the recent Prodigy acquisition, opening a whole new, and huge market for them :…

Upstart feels like a company with multiple tailwinds coming together at the same time: lots of consumer demand for personal lending at a time of massive stimulus, a re-opening of the global economy after a pandemic and AI coming of age, with a head-start in an area where the market is ripe for disruption. And it’s well led with a top-class CEO and founder, profitable and growing exceptionally fast. So even after the incredible recent run-up in the stock I remain bullish.


Crowdstrike	23%
Datadog		23%

Snowflake	14%
Cloudflare	12%
Upstart		10%
Zoominfo	10%
Inari		8%


Crowdstrike (CRWD) - 23%

Crowdstrike remains one of my two top conviction stocks. I did a deepdive on the stock in January:…

And Saul’s monthly review covers the Q1 results best - it was another stellar performance.

During the month of May they updated the Falcon platform and extended a partnership with E&Y.

They will report on 3 June, and Zscaler probably gives an idea of where they may land. Zscaler had a great quarter, with accelerating revenue and billings growth.

I think Crowdstrike will again report excellent numbers. They guided for 63% yoy for Q1 and I think they may just come in above 70% which will be fantastic.

Datadog (DDOG) - 23%

Datadog has become a top conviction stock for me. They’ve seen some valuation compression recently despite accelerating underlying performance. They will have a Q2 revenue comp which was negatively impacted by Covid a year ago, which means that Q2 revenue growth will almost certainly accelerate a lot vs Q1 which was 52% yoy.

They guided for 51% yoy for Q2 and I think they’ll come in above 60%. GauchoRico seems to agree - he’s pencilled in 63%. Let’s see.

Snowflake (SNOW) - 14%

This post summarises my thoughts on the latest results, which were impressive:…

They have an exceptionally high NRR at 168% and a whole lot of signed up customers from the last 2-3 quarters who still need to start generating revenue. They have a war-chest of $5.1bn on hand for smart acquisitions. Their Product GM% is trending up from the already high 72% and their operaitng margin is improving a lot as the business scales. Their Sales force verticalisation and data sharing opportunity still needs to bear fruit, and international growth is very high, and accelerating, with the CEO, who is originally Dutch, committing to giving it his personal attention. So all in all a great set of results.

Cloudflare (NET) - 12%

Cloudflare continues innovating, but the growth rate did not materially tick up in Q1 2021, so I’m not yet convinced that it will in future quarters. It is a fantastic company for reasons that have been discussed a lot on this board, so I won’t repeat them. The relatively lower growth rate is why it is above 15% allocation.

Upstart (UPST) - 10%

Covered above.

Zoominfo (ZI) - 10%

In this thread I gave a fairly detailed overview of the company and results:

It looks to me like the company is experiencing tremendous acceleration.

In Q1, new customer growth doubled yoy, and they are executing very well on upselling to their existing base - ARPU grew to the highest yet, and they added 100 customers to the cohort spending more than $100k - from 850 to 950. Their International growth is really starting to hum. International revenue grew 14% qoq (69% annualised), and within the Q they said the month of March (the last month of the quarter) was the best ever which means that Q2, if that momentum holds, will be even better than Q1. New product adoption is going well with the following 3 products growing a lot: Salesforce sync - 8x increase, Engage ACV up 100% qoq, Intent ACV doubling yoy. And their customers are truly engaged with the platform as seen in the 12% sequential increase (57% yoy) in both MaU and DaU. And lastly they have great gross margins (89%), operating margins (43%), free cash flow margins (64%). They guided for 48% yoy growth for next Q, and I think they may breach 50% yoy for the first time next Q.

Inari (NARI) - 8%

I liked the latest reported numbers but had a nagging feeling after the conference call about the future. What worried me is the path to keep on hyper-scaling as there did not seem to be a lot of optionality and the additional products that they were developing, they were stressing were very cheap relative to the existing ones at around $200 vs the ±$10k for Flowtriever and Clottriever.

I also just read Saul’s thoughts and why he sold, and based on that and the above, I will probably reduce my position in Inari.


It’s been a lot of work the past couple of months for a 6.6% return…but I’ll take it gladly. I made a couple of sub-optimal bets but luckily a couple others that worked out too. Last year everything worked. I guess that this year is more normal.

All of the best, and be careful out there!

  • WSM

Previous reviews:
April 2021:…
March 2021 Q1 ytd:
Dec 2020 full year:


Excellent write-up on Upstart. The reality is no one knows how good their AI is and it might not matter long term. For me, the bottom line is they are lending money to people that are limited in their capacity to repay the loans if the economy declines and we experience another wave of job losses. Yes, that scenario looks unlikely at this time.

I would imagine that their system works well at a certain scale. But there is a point where they have exhausted the pool of low-risk, low credit score borrowers. Growth would then only come from lending to higher-risk individuals. But where is the tipping point for that? I don’t know if anyone knows.

I like this stock now, but I don’t think it’s a long-term buy-and-hold, 3+ year investment for me. The risk will catch up to them eventually as they grow. The bigger they get, well…


Perhaps the AI approach of UPST is not limited to people with lower conventional credit ratings. But will apply to those with higher credit ratings too, resulting in lower interest rates paid by borrowers. Assuming it is better than FICO.
Starting low and working up fits classic innovator’s dilemma theory. But of course you are right ,it has not been tested in a deep recession

My investment is mostly based on the idea that there must be something better than FICO.

what FICO looks at…

factors ignored in FICO…