New Constructs' Friday Webinar: Intelligent Capital Allocation 14 April 2023

This is our Friday webinar. We're going to have Garrett and Lee join us, and we're going to move a little bit faster, and you can get the slides in Circle. What is this about? What is New Constructs about? What are we looking to accomplish? We want to build a community of people who understand profitability, valuation, intelligent capital allocation, not just because we think it's fun and interesting— we do— but because we think it's crucial for the success of our capital markets and for our society. If we have a lot of people out there just destroying capital, ultimately that's going to make us all less happy. Disclosures and disclaimers. Please take a look at those. We are not an advisory firm. And then here's the agenda. So we're going to go over some of the top research that came out this week. We're going to touch on some of our macro fundamental research. Garrett's going to go over our new feature. Lee's going to talk about our RoboAnalyst. And with that, I will kick it over to you, Kyle, to highlight our beat and miss reports that we just published.

Great.

We put out a lot of stuff this week, so definitely go and check out the research tab of the site. We're right in the middle of the updated quarterly sector ratings and everything, so David will touch on that a bit. But these 2 reports are earnings previews. What stocks are most likely to beat, what stocks are most likely to miss EPS in the upcoming quarter or the upcoming earnings season. And it looks at core earnings and Street earnings. If core earnings are higher historically, we think that's a company that looks like it's likely to beat. Core earnings are lower than Street earnings, it's something likely to miss. A lot of times the Street earnings numbers Historical project into the future. If they're missing something, they might continue to miss it. And so we look at where are companies where those differences are the biggest. And you can see in these two charts here, and there's more details in each of the reports of what companies are again most likely to beat, which would indicate that our core EPS number is higher than Street EPS. And there you got Constellation Energy, American Tower Corp, Synopsys, T. Rowe Price, and Charter Communications. All of them have more profit, more profits than you might realize if you're just looking at Street EPS. And it's important to note these are not necessarily long-term ratings. Some of these companies might have neutral, unattractive stock ratings, but they've got high earnings distort— beat or strong beat earnings distortion scores. So it's just looking at kind of the next quarter. On the miss side, you've got First Solar. Intel, Dish, American Water Works, Atmos Energy. You can see the percentages can be large, 200% overstated, 204%, 210% understated in some of those. And so it's really a good look at where there's big disconnects in the market between core earnings and that Street EPS number. With that, we'll jump into individual company, another zombie stock. Do you wanna switch one slide there?

Yeah.

Lucid, zombie stock 32. We've talked about the EV market in general. We've been bearish on Tesla largely based on valuation, the competition in the market as well. Rivian has been a previous zombie stock and now we're adding Lucid to that group. It's a similar situation to Rivian. These companies don't have the scale to ramp up and build out production that would be required of their valuation. It's not to say that they might not make a great car or something people are interested in, but it's to say that their production does not match the valuation. The fundamentals don't match the valuation. When we look at Lucid specifically, we've seen a company with $9.8 billion in free cash flow burned in just the past 2 years. They want to ramp up production and they're plant in the US. They're trying to build a new plant over in Saudi Arabia. So it's got huge capital expenditures coming up on top of huge free cash flow burn in the past. So that cash burn's probably not going to slow down at any point. Our estimate, optimistic scenario, gives them 23 months till bankruptcy. Management's actually come out and said that they have enough capital to the first quarter of 2024, so it could be even shorter than that. based on what management's looking at as sort of investments. And then it's a similar story that we've pointed out with Rivian, with, uh, profitability well below. Rivian's actually a little worse than Lucid, but Lucid had 331% margins in 2022, ROIC of -29%. And then we look at the actual valuation and what Those, the $8 a share implies on revenue, profits, production. It's numbers that would seem to stretch the limits of reality there. When we look at the current share price, $8 a share, hey, it's fallen from $55. It seems like it could be a deal, but not at all. We looked at different scenarios at different selling prices. So if we keep the selling price that they're currently selling, $139,000 to justify that current price, they'd have to sell 409,000 vehicles compared to just over 7,000 last year. We've got huge expectations even in a scenario where they're keeping that sales price way high. If it falls to $49,500, which is the average car price, new car price in the US at the end of last year, we've got to sell 1.2 million vehicles to justify the current price. And again, that's just below what Tesla sold, and it's compared to Again, 7,000 cars in 2022. They actually just put out Q1 production last night slash this morning, and it's below expectations. It's, they're producing more than they're actually delivering. So it could be the demand problems. So it's everything has to go right to just get to that $8 a share, which is why we think that valuation here is stretched again beyond the limits of reality, especially with competition for Lucid to grow into a 1.2 million vehicles produced, they're going to be taking significant market share, which is something they're not seeing right now.

On the cash flow front, right? So we're giving them— we're showing 23 months if they maintain their current burn. What do you think it is about their own internal cash flow suggestions or forecasts that suggests they might even have less runway than we're giving them credit for?

Probably. Well, you would think that They're expecting to ramp up production or they've gotta invest to build out that scale for future years. But the problem is they're investing for the future and that future might not actually be around when they get to it if they don't have a business model that's actually making money. But they've already laid out how they want to increase production, again, at home and abroad. And so that free cash flow burn of $9.8 billion in the past 2 years may accelerate even further, which would give them less of a runway. If we were to plug in, hey, cash burn is actually going to increase, that 23 months would fall significantly.

Yeah.

I mean, I think that's an often overlooked point that in order to remain competitive, and this also applies to Tesla that's drastically reduced their R&D and has not really spent the CapEx needed to come anywhere close to justifying the expectations in the stock price, is that Manufacturing cars is not cheap. It takes a lot of money. That's why there aren't that many companies in the world that do it. And we've probably, in terms of the electric vehicle market, I mean, you went from 1 to, I don't know, half a dozen, plus the legacy automakers, at least half a dozen, right? That have jumped into the business. The point being, there weren't a whole bunch of large legacy automakers out there. Why? Because it's very capital intensive. You need scale, and it doesn't make sense to have that many companies competing Than what there were. And yet we have almost double that amount already in electric vehicles. So there's going to be a lot of companies that go out of business. That's the takeaway. And it's a very cash-intense business and all the more reason why the incumbents I think are in a good position.

Yeah.

And to sort of round out that valuation, that chart that we had at the bottom here, actually this one here shows the implied production at different selling prices.

I love this analysis you guys do. This is such great analysis, understanding the implied production. Right. By 2032, in 10 years, where they got to be in terms of production to justify the stock price, and then what that means in terms of market share. And that's just a, that's a lot of market share for that expensive a car.

And it flies in the face of what they were projecting to begin with. And there's some of the SPAC, when they were doing the SPAC process, some of the projections there were market share's going to peak and actually start declining. through 2026 and beyond. And that's largely because the market's growing. So yes, their market share might fall, but those market share numbers implied by the current stock price were multiples higher than what even the company was projecting, which is why it lands on that zombie stock list.

Yeah.

Okay. Time to skip over here to some of our macro work. And I've been talking about this a little bit in the last couple of weeks. I want to focus on this stuff some more because this aggregate analysis is really unique. Nobody else in the world is able to put together numbers like this. We can look at all companies across all sectors on an apples-to-apples basis because of the in-depth analysis that we do and the ability to understand the underlying economics of businesses and transcend accounting. That's just the New Constructs thing. So I wanted to make sure that We had a second to talk about some really cool posts that Garrett put up a week or so ago on how the different performances by sector are so different from the overall. What I'm saying is the S&P or the New Constructs 2000, which I was showing on the prior slide, that's our all-cap index. It's sort of our proxy for the entire market. It represents like 98% of the market cap. So A lot of the small companies that we don't always cover, they don't stay around for long. That's why we don't spend the resources to go cover something that's just going to die in the next couple of weeks. But anyway, it's our all-cap index. That's the NC 2000. And what Garrett pointed out here, there's some really interesting disconnects. He did this for the S&P 500 last week, or we covered it last week. This is for the NC 2000 disconnects between core earnings in the healthcare sector And gap earnings, right? And then a very different disconnect on economic earnings, right? Going down and gap earnings going up. This is the overall ROIC. You can see a huge increase in WACC for the NC 2000. Free cash flow yield at the basic materials sector, right? You got enterprise value going up, free cash flow going down. Not a good situation. And then a big disconnect on economic book value at the technology level or at the technology sector level. You can see market cap going up, economic book value going down. Again, not a good thing. I added a post yesterday that's a nice summary of all of these NC 2000 reports. Click here, it'll take you to that all-cap index. GAAP earnings versus core earnings. And what I put it— put, put in here for each one of these reports is a quick summary of where the sectors were so— were most different from the overall. So, or showed the biggest contradictory performance. Like, so for example, consumer cyclicals, GAAP earnings were down big where core was flat to up. Consumer non-cyclicals, I meant to say non-cyclicals here, GAAP down big. And the same is true, flat to down for core healthcare. GAAP earnings down big. And core was up. And so I did that for all of these so that if you, if you want to get a quick Cliff Notes on what's going on, you can get that from this, this post right here. Quick question from Kent. Is it market cap or equal weighted? Is market cap weighted, right? So that's how we're trying to keep it balanced. You can see here a bunch of churn going on underneath the surface. As well, on a sector-by-sector level where we're detailing that in in in the overall report also. Then I want to talk about our best and worst. This is a quarterly product that highlights all the best and worst ETFs and mutual funds by sector, and nobody else. I don't think anyone else does this. We do it also by style, but these are. These have been published. And so if you wanted to get a sense of how many ETFs get a very attractive rating, well, there's 112 of them and they represent 24% of all ETF assets out there, equity assets. And then attractive, you get 12 and 116. The most, is it neutral? Right. Which is a bummer. You'd hope that more people would allocate more money to the left side of the picture here. As opposed to neutral and unfortunately very unattractive. 21% of all equity ETF assets that we cover are in unattractive funds. And surprise, surprise, guess how— what class of or what type of sector ETF are there the most of? Very unattractive. 287. So anyway, that's part of why diligence matters, right? Do the work in the underlying fundamentals. Don't just buy it because it— you think it looks good or it's from a fancy issuer. Do the diligence because there are more bad ETFs out there than there are even neutral, and they're more than double the bad, double the attractive and very attractive. So swimming with sharks. The other thing I want to highlight is our sector and industry fundamentals update. Kyle just put this out. It's up on our website. I'm going to click to it here. Every quarter we'll just go through and we share with our institutional clients just a ton of great stats on a, on a sub-industry level and a sector level. So you just get a ton of detail at this aggregate level for net income, core earnings, NOPAT, invested capital. It's all there and you can see, and we're just— it's a big breakdown, tons of stats, very similar to what Aswath Damodaran does and his— some of his stats update and updates. If you're looking to know where to find the best and worst ETFs and mutual funds, you would do that. You go to this, go to the research section of the site, and we've done so many of them recently, you'll find them in the main area. But if you want to go get those on a regular basis, you come down here to model portfolios, you go to best and worst ETFs and mutual funds by sector, click here and you'll get them all.

All right.

Here's a recap on how— where all the individual reports are so you can in one spot quickly get to them. And then in the beginning, we'll give you the ratings overall so you can see which sectors have the most capital allocated to very attractive versus very unattractive. And then when it comes to the sector stats that you can get clicking here, sector and industry fundamentals, and here's our 4Q update and you click here and that's where they all are. So if you want to find those on your own, that's where you can find them. All right. I don't have any— no more questions on that. So I think what we can do here is pass it over to Garrett. He can get started a little bit early here to talk to us about his work that he did to improve our DCF models, particularly the Excel version. Garrett's a longtime member of our analyst staff. one of the best, fastest analyzers of filings in the entire world. I would say that out of the top 20 people in the world that know how to read annual reports, 25 of them are at New Constructs because these guys have technology and capabilities and training to do this better than anybody. And I'm proud to hand it over here to Garrett to talk about some really cool feature enhancements that we made to the Excel version of our models. Take it away, Garrett.

I appreciate it, David. You are way too kind, so thank you. You can now reverse DCF in Excel. Extremely exciting update recently. Let's just take a big picture here. Download xOS if you've never used it before. Brings our web-based company models into Excel. Just makes it easier for clients that enjoy using Excel. Some people don't like using the web-based version, so we have the Excel version for people who like that. 20 tabs of data on the Excel version. It mirrors the website, so return on invested capital, free cash flow, everything you need directly in Excel. And the thing that I like about the Excel XLS more than the web-based version is we have formulas scattered throughout the entire XOS, so it's auditable. There's no black boxes anywhere throughout it. It's educational. I'm constantly using it to calculate invested capital. Sometimes I forget, do we add that? Do we subtract that out of invested capital? So I'm constantly in the XOS using it every day. And then now it's extremely applicable. Expectations investing is directly in Excel. In the past, you would just look at the DCF page and just look at it, but now you can change the revenue assumptions, profit margins, invested capital assumptions. So it's great, it's usable. And lastly, a lot of busy professionals Use our data. So once you download it on your laptop or computer, if you're in the airport, if you're anywhere that doesn't have Wi-Fi, if you're on the beach, you can use the XLS. You can reverse DCF from anywhere now. So super exciting. But yeah, here are the, the changes that we actually made to the DCF tab. It's now 100% dynamic. Like I said, the DCF tab in the past, you would just be able to look at it and not interact with it. Now you just go in there, you find the red font and yellow highlight where the revenue assumptions are, profit margins. You can change the cash tax rate, weighted average cost of capital. It's all there. Your expectations of the future, you insert them and shareholder value dynamically changes for you. So super usable. We added 70 formulas on that page as well. So anything on that page, if you don't know how we calculate it, it's right there for you. So super educational.

Yeah. I hesitated to put this out there in Excel because it's like, it shows how we do all these calculations. That stuff's in the public domain. So it's not really a big deal, but it is really cool if you But for people who would like to learn how to do a DCF right, people who would like to execute expectations investing modeling, there's no better way to do it. I mean, this is, this is, this is one of the most sophisticated, if not the most sophisticated discounted cash flow models in the world, which of course driven by the best data, which matters as much as the formulas, if not more.

Yes. Okay. Locating the download XOS. How do you actually find this? I'm just gonna go ahead and share my screen and take you guys to the website. Give me 1 second. All right. So there's basically 3 ways to access the XOS. When you log in, you get directly taken to the portfolios page. So if you want to get the XOS from here, the portfolios page right here, You just expand one of these guys right here, and it's right below the chicklet. So you would just click on download XOS next to our company model marked up filing. So just go directly here. If you're on the ratings page, it's basically the same thing. Chicklet, download XOS, come here. Look for any ticker. You got a new XLS at your disposal. And lastly, the screeners page. Let's say you want to look at the energy sector. You're curious about that. Filter that. Then same thing as the portfolios page. You want to see Consol Energy. Expand that out and boom, you have the download XLS. Right there. With that said, I'm going to take you to the actual Excel now. So here we go. Once you actually open up the XLS, you get take— you get taken to the stock ratings, credit ratings, everything on the website is here, return on invested capital, everything. And like I mentioned before, these blue Highlighted tabs or rows. They have formulas in here, so you just go in there and you see how everything is calculated. Can educate yourself. But all right, here's the new DCF. So when you first get in here, it's key to read this: the red font and yellow highlight. These are input estimate drivers of the DCF. So as you can see. We have red highlights scattered all throughout here. I'll just go through each one, one by one. So yeah, total revenue growth. If you want to change our assumptions that we originally have to say 10%, 7.5%, 5%, you just go in there, you enter what you want, and shareholder value down here will dynamically change using Profit margins are going to increase. Change that, and you can do that with the cash tax rate, the invested capital assumptions, and that gets you all to corporate value on the way to shareholder value. But you can also change any of our valuation adjustments. So this is McDonald's we're looking at, $46 billion in debt. Maybe you think in the future they'll have $30 billion in debt. You just enter that, and then from EY2 to EY100, everything will dynamically change for that. Also, you can change shares outstanding. Maybe they buy back shares. Basically, you can change everything on here, and probably the most important one that you want to look at is the weighted average cost of capital. This is really sensitive to the model. So if you disagree with our 6.6%, you can come in there, add 5, whatever you want. So however you view the future, you enter it how you want to enter it, and shareholder value and the DCF will change for you flawlessly. To connect this to the website, I can share you guys with a company model. Basically, we took the forecast page. This is where you can create your optimistic pessimistic scenarios. So yeah, if I'm optimistic about the future, you change your revenue assumptions here on the website, profit margins. Tax rate, invested capital. Once that's done, you save that just like you do with the Excel version now. And the other thing that you can change on the web-based version is the weighted average cost of capital down here. You want to enter it into TTM for the DCF valuation adjustments. So Basically, we took the forecast and overrides page and converted that directly into the Excel version. So for example, I put my optimistic scenario in for McDonald's, and I wanted to look at a 25-year forecast horizon. You just come here, you can go to 25. And I'll show you guys on the Excel version how it all matches up. So we have $395.39 as our shareholder value per share there. And then here's one of my DCF optimistic scenarios I made. I entered the same exact assumptions as I did on the web-based version. You come here, it might be a little bit small for you to see. But there you go, $395. So whatever you do on the web-based version, you can do it in Excel now.

Hey, real fast, a couple things I want to ask you about. So how did you get multiple scenarios, or how did you get an additional DCF tab in the Excel file so that you could have multiple scenarios in Excel like you do on the web?

Yeah, super easy. This is the original DCF that— this one was completely unchanged. So I'll just right-click on the DCF, and then I'll click on Move or Copy. And then this is all the tabs here. I'll just find the original DCF, and I will click on Create Copy here. Click OK, give it a second, and then this DCF2 pops up. So an exact copy of this guy, and then I can just come in here and rename it anything I want. I already have an optimistic and a bear case, so maybe I'll just call it DCF best case scenario of all time. Everything The brightest of the bright for the future comes up on this one, but you get the point. You can name it anything you want. With my DCF optimistic, you just come in here. I thought 6% revenue growth for McDonald's might be pretty optimistic if you look at their past history. And then up here, I just added some rows and I just took some notes. And I jotted it all down here on my reverse DCF. I think McDonald's is leveraging technology effectively. Post-COVID ordering through the mobile app, delivery growth, digital kiosks— they're using technology efficiently. Economies of scale through their superior distribution network and high free cash flow. Just some general notes. I wrote, I wrote up there. And then of course I put the New Constructs research article. So, but yeah, not too much more to say here.

Garrett, real fast, let's take a question.

Yes.

From John. Do you generally use company forecasts for DCF inputs when available? The answer is no. We will use consensus estimates where available, and those tend to be available only at the revenue level. As far as we know, there are no other firms calculating net operating profit before tax margins and cash tax rates. So there is no consensus, and there are no companies that really usually like to talk about that. Same is true for working capital and fixed assets. So in the absence of any consensus values, what we will tend to do is intelligently extrapolate what the company has done in the past. So we've got quite a few algorithms and intelligence around the best way to conservatively extrapolate what's happened in the past. Like with biotech companies, you've got super negative and up and down, and so you've gotta, gotta deal with a lot of noise. But the general thesis around what it is we're doing and why is we want to, in our default scenario, project what we think is most reasonable for the company to perform in the future and let the valuation flush out how unreasonable or unreasonable its expectations are. So if we're sort of extrapolating middle of the road, and the valuation, the stock price makes it— it looks as if the stock price is going to— it's going to take forever for the DCF to produce a value equal to the current stock price. Well, then it's expensive, right? Or if the DCF values are way above the current stock price, well, then it's cheap. The idea is we don't want to put in extremely positive or positive at all, or extremely negative, or negative of all— negative at all forecasts in our default, because we want the valuation of the stock to be flushed out based on a conservative extrapolation of what's happened in the past where we don't have consensus expectations, right? We are not here to try to color your thinking. We're not here to try to predict the future. We're here to be a critic of the market as a fortune teller. The market's giving us a stock price every day and therefore providing a forecast for future cash flows every day. And we don't want to try to be in the business of forecasting cash flows. We want to be in the business of being critical of what the implied forecast is. That's just a better way to go about valuation. We don't have crystal balls, but we are going to be critical of people who do use a crystal ball, Mr. Market uses one every day to assign a stock price. And so we will flush out exactly what those cash flows are. That's the idea. So yeah, in terms of estimates, that's the rationale.

Yeah, that's spot on, David. Yeah. Predicting the future's tough, so we let other people do that.

Did I miss you show the notepad analysis and the applied share prices tabs?

Yeah, I can show that right now. I guess another thing about the XLS, it just makes analysis Super easy. Like, here's my optimistic case, my bear case. I just created this little DCF scenario analysis between the optimistic and the bear. I was specifically looking at 25 years into the future as my forecast horizon. McDonald's is currently trading at $285 per share, so If my optimistic case, if that all hits, then maybe I think McDonald's is worth $395 per share today. So maybe they're a buy. But on the other hand, my bear case, 25 years into the future, says McDonald's might be worth $225 today. So with those 2 things in mind, trading at $285, it might be a tough call for me to say maybe McDonald's is fairly priced. So maybe I don't do anything here, but, but like I said, you can create simple analysis with a chart like this. Notepad analysis. This is also super easy. I grabbed all of these from the Notepad page and then just copy and pasted them in here. So from '98 to today, you can see how Notepad has grown. So right up to this point, so pretty steady growth, I'd say. Um, but this was a nice spot check for me personally, because look at my optimistic case. I have Notepad growing from about $10 billion to, let's see, $35 billion. That might be a little unrealistic. So sometimes making graphs like this of your optimistic bear cases Spot check you here. Maybe my bear case makes a little bit more sense. You see, in my bear case, I have NOPAT not even doubling. So maybe maybe that's a little bit more conservative. Maybe that makes sense.

That's great.

That's a cool chart we use in in a lot of our reports, right? The past versus present NOPAT to get a sense of just how. big a difference there is. So that's good stuff. Go do the ROIC. That's a great one too.

Yeah, and then ROIC, another good spot check. Uh, McDonald's current ROIC is about 18%. So as you can see, my optimistic case, I have it growing from 18% to about 60% 25 years out. So maybe I was a little bit too optimistic on this Friday. So maybe, maybe, maybe my bear case makes a little bit more sense here, growing from 18 to the mid-20s. So, but like I said, uh, with all the— with the DCF, you can create as many different scenarios as you want, and then from there you can create any sort of analysis too. So like in my case, it was a great spot check for me. So my optimistic case was probably Too optimistic.

That's awesome. Thank you, Garrett. Good stuff. Again, these slides will be posted in Circle, so if you got more questions, let us know. But now it is time to pass it over to the one and only Lee Mineta Kohler. Let me see if I can get back to sharing the screen.

Here we go. Cool. Thank you, David.

You're welcome.

Great job, everyone.

Yeah, great job. Yeah, Lee Manetta Kohler, our COO. And look, Lee and I've been working together for like almost 20 years. He was around when we built this first technology. He is in many ways the guy in Frankenstein that creates the monster, and the monster's the roboanalyst, except our roboanalyst is much nicer than Frankenstein. And he didn't ever die. So maybe it's a really bad analogy. But anyway, Lee knows our technology and our filings and our models as well as anybody in the world. And I'm really excited for him to come in and spend a few minutes talking a little bit about how the magic works here at New Constructs for getting the data and how much better we are at it. Take it away, Lee.

Thanks, David. Yeah, and the magic's getting, getting better every day. I'm really happy to say in the first quarter of 2023, we updated more than 2,500 models with filings from the SEC. David talked earlier about how great Garrett is as an analyst. He updated almost 400 of those himself, and Garrett is, is a superstar. But updating 400 models in a quarter, I think, is something that requires a little bit of technology to do well. It's, it's, it's hard work. I think we've talked before about how annual filings on average are around 250 pages, and those 250 pages are not written to be Easily consumed. Most companies have fiscal years that end on 12/31. I think most people know that. So that means their annual reports are due in late February and early March. That keeps us pretty busy. Annual filings are like the best guide to the financial performance of companies because we get more disclosure in those filings than we do in quarterly filings.

What's the average length of a 10-K, Lee?

About 250 pages. It's up from about 210 almost 5 years ago. Not quite 5 years ago. They continue to grow.

Yeah, I remember they were like 15 or 30 pages. Is it like, what's it like? Is it like reading a comic book or is it harder than that?

It's definitely not like a comic book. I think generally speaking, these things are written by accounting firms or accounting, CFO offices at big firms. It's written by a bunch of different people and sort of all aggregated together at the end. It's built with a bunch of software. I'm not quite sure that anybody actually reads them afterwards except the parts that they want to make sure that they're controlling the language in. So they're pretty disjointed, pretty hard to follow.

And not just disjointed, but there's a lot of repeated information as well too.

So there is.

Yeah, no, that's right. And that you get the same information twice. Often you get it in the MD&A at the top of the filing and then again in the notes. And companies will sometimes give you pretty different stories. The notes have clear— clearer disclosure rules. They can say pretty much whatever they want in the MD&A as long as it's not fraudulent. And they do say whatever they want. And those 2,500 filings that we got this first quarter, those really aren't spread out over the quarter. We get them sort of all at once. We get about 250 of them a day during filing season. And this is the first year ever where we were able to process every filing we got that day, the day that we received it, or the day after we received it. That's the first time we've ever done that. We've— David's got this chart up on the screen of our annual filing season backlog days. It's a KPI we've measured for many years. It shows the number of days that we've not been able to process all the documents we got that day. And we got zero of those days this year. I think that's an important milestone for us and for the clients. I know that you all want our data and ratings as quickly as possible. We don't want you to have to wait. And we're really happy that we were able to get this, this done. And we did this primarily through the use of technology. Analysts like Garrett are very diligent and work really hard and put in long hours for us during filing season. But the real driver of our increased productivity, our increased capacity, is the improvements that we make to the machine learning system. David, if you want to swap over to that next tab.

Yeah, quick question, Lee, about how many ADRs do we cover?

It's a little around 200. I'm not sure exactly offhand. ADR do have slightly different reporting requirements than domestic companies. The SEC over the last decade has continued to shrink the amount of time that companies have to report their 10-Ks and 10-Qs. When they classify companies in, in 3 ways, the, the first ones are large accelerated filers, and then they have accelerated filers, and then they have regular reporting, and they have a separate set of rules for international firms. So for companies that have fiscal year ends at the end of December, we— the deadline is generally a couple of days into March to receive their annual filing. So most companies are reporting usually on a Thursday or Friday. I think the accountants want to get those filings in and go home for the weekend is usually when we see the most filings. Whatever the Thursday or Friday is before the deadline is the most popular reporting date.

Michelle Leader calls that the footnoted lady. Michelle Leader, she calls it the Friday night dump.

Yeah.

Yes, that's right.

She's totally right.

The Friday, the deadline for submitting those is something like 4:30 Eastern to the SEC. Then they shut off access to their, to their system. And you will see at 4:20, like a huge— just people piling in. Sometimes we will see between 4, 4 o'clock and 4:30, we'll see 75% of the filings come in that were going to come in that day. So they like to wait till the last minute.

Yeah, yeah, yeah.

So for, for international companies, for ADRs, they have different reporting requirements.

We're—

we still haven't seen their Ks or their 20Fs or 40Fs. We're likely to see those here at the end of April. That makes for an unusual situation where we have an extra long wait from their Q3 filing until their annual filing for companies with a regular fiscal year end. The quarterly reporting requirements for international companies are much more similar to domestic companies. The annual ones, they have much, much more time. So we do mark those companies, we suspend their ratings and mark those companies with an analyst note. It'll say foreign filing delay. And I think there's, I think around 60 companies that we've marked with that delay recently when we feel like they're Quarterly data is too old. So if you see that on an analyst note, that's what that means. So yeah, all of these improvements are really driven by machine learning improvements and machine learning technology. We've got an awesome software development staff here at New Constructs who work hard to help us improve our software to automatically collect and classify data. What, what the system does is automatically collects data from tables and from paragraphs throughout the filing, from the statements to the notes to the MD&A. And it, generally speaking, gets that data right. Makes Garrett's life a lot easier. He doesn't spend a lot of time, or any time really, looking at income statements, balance sheets, cash flow statements, pension tables. Well, those ones are sometimes more complicated. ESOs, operating leases, All the kinds of tabular data that exists. So we do a great job of getting that stuff right. The other thing that the system does is it directs analysts through to sections of the filings where we think there's more complicated disclosure that the machine can't figure out on its own. And that saves Garrett time too. That way he's only looking at maybe less than 5% of the pages in the filing instead of all 250. And he's not just looking at those pages. He's looking at specific paragraphs or specific items that we think are interesting and that we need to be able to model correctly. The software is also always watching what Garrett does. We're always watching Garrett and every other analyst, and we learn from what Garrett does. So if this machine makes a mistake and Garrett corrects it, we track all that information and improve our model as we go. You can see this chart on the screen. We've improved the of the data that's collected with our machine learning system from the mid-50s to over 70% this year. This is for annual filings. In quarterly filings, these numbers are even higher, primarily because companies just disclose a lot less in their quarterly filings than they do in their annual filings. We're talking about anything from 400 or 500-page-long documents for a complicated utility company, for example. Or a company that engages in segments like an auto manufacturer that also has a financing division. Those are pretty— those are basically separate businesses. The machine learning system is able to distinguish all between all those kinds of companies and filings and automatically collect this data for us. So it's, it's, it's really impressive to watch. Like David said, we've been doing this together for a long time. When I started working with David at New Constructs, we were doing some stuff in Excel and building the early versions of the way that the system worked. And to see what is capable— we're capable of doing today is absolutely amazing. And the reason that it's possible is because we have expert analysts who understand the finance, and we have excellent software developers who understand how to use the technology. That's great. I've worked with and talked with other firms who have both of those things as well. But what they don't have the ability to do is communicate between those people. We spend a lot of time making sure that our analysts and engineers, our software developers, can communicate and speak to each other. You can have awesome software developers who can build cool technology that does something that doesn't matter. And you can have analysts who, I think the traditional Wall Street analyst thinks pretty highly of themselves and they don't want to believe that technology can replace what they can do and their special insights. So I think that much of the magic at New Constructs is our ability to talk with each other and to hear each other and to listen to each other, for analysts to teach software developers why the finance is the way that it is and what they're thinking and why. And for software developers to explain to analysts how the technology works and why it works that way so that we understand what's possible. That's the big driver of these improvements.

To jump in, sorry, Lee. I mean, the original design for this technology was always to enhance the expert's ability to go through a filing faster, better, smarter. And the idea was always to track what an expert would do. There is no such thing as, as an algorithm, a formula, technology, machine learning, artificial intelligence that's going to figure out what's going on inside of a filing without an expert helping it navigate the nuances of how things are disclosed. There are terms that say the same thing or spell the same way and mean different things, and vice versa. And without the intelligence of our experts to go through and make those subtle distinctions, The technology is never going to go anywhere. Right. And that's what we've been diligently building for so long. The original tools and literally just a better way for us to accelerate the collection of expert instructions about how to go through a filing in a format that a machine can follow and read. And as we've gotten faster, the database of instructions for the machines to follow has gotten bigger, faster and bigger and faster. And so we really have a robust set of training data And so you see here in this chart, in terms of the, the direction of the technology, that hockey stick is going to continue way up, and it maybe even gets steeper because we have built so much underlying intelligence infrastructure, etc., around collecting the specific instructions about how to analyze the filing that we're ready for the technology to just start mowing that stuff down and, and We think we're going to get a lot faster really soon.

We have technology that's not ready to be pushed out into the world yet that's all— that's already meaningfully faster and better than what we have here. So we're getting faster at it and more successful at it already.

And, and that's a, that's a compounding effect, right? Because the faster we get, the more intelligence we can add to the system faster, the more training data it has, the more filings we parse.

Yeah, machine learning doesn't work without, without consistent and accurate data. It's built on top of what you tell it. You've got to have a disciplined process to create a dataset in the first place. New Constructs has always been good at that. We've always been people who are paying attention to the details and making sure that we get every part right. And those are— that attitude has been what's allowed us to be successful at using machine learning more than others in the marketplace.

Lee, let's, let's close it out with one, one common question we get a lot. We get this all— I get this all the time in meetings, whether we're talking to audit firms or talking institutional investors. I always, I always want to know who are the— and I'll quote Kent here— who are the serial less-than-honest filers, or which companies produce the most difficult-to-analyze Ks? Because honest is a sort of a a word, I don't know how well we can use it, but which Ks cause us the most trouble?

Sure.

I think that the most difficult Ks to read are generally from large-cap companies with complicated businesses that generate their own reports. So we see big companies like Leucadia and Liberty Media in particular that have large numbers of segments and divisions that all operate independently, and those companies produce their own reporting. So a mid-cap company or a small-cap company, generally speaking, going to rely on somebody like KPMG or PwC to generate their reports for them, and they're more likely to be a little more templated. They're like more likely to be a little more straightforward. That doesn't mean that they're not bending the rules here and there, but I think the difficulty in understanding filings are ones that are written sort of won off by larger firms that have their own accounting staff and rely less on their auditors or less on software and that have complicated businesses. I think Liberty Media and Leucadia, Garrett may have other ones that remind him, but auto manufacturers have been difficult in the past. Ford and GM have always been difficult. GE and its complicated financing division has always been very difficult. Those are the hard companies to just read the filings. That doesn't mean that their economics are good or bad necessarily, but just understanding what's happening is complicated in those companies.

Thanks, Lee. I'm going to hit one more question and then we're going to close out. The question was, are there any REITs that we like? So let's hit the screener. And look for good REITs. We got 2 different types here. We'll just look for— start with residential commercial. There's 1, 2, 3, 4, 5, 6, 7, 8, 9 that get an attractive or very attractive rating. That's where we'd start. If we're talking about real estate ops, we've got 6. So that's how we would answer your question, Mark, on the REIT front. Otherwise, thanks every— very much, everybody. It's been a great time. We hope you have a great weekend. Thank you for being with us, and we'll see you here next week.