Hi everyone, My friend @LibertyRPF, who writes the eclectic newsletter, Liberty’s Highlights) recently interviewed me. Check out the podcast! Also, a reminder that MBI and I recently launched a podcast (Spotify, Apple, RSS feed). The second episode, a conversation about Xpel, drops soon. At a basic level, biologic drugs bind to specific proteins embedded in cells, like […]

(Note: Several weeks ago, MBI and I recorded a podcast conversation about AppFolio. You can listen to it on Spotify, Apple, YouTube, or via this RSS feed). While narrowly geared toward property managers today, AppFollio was founded in 2006 with the more expansive aim of providing business management software small and mid-sized businesses in general. Its horizontal orientation […]

Happy holidays everyone,

This post will kick off an annual tradition where I provide brief summaries of blog posts published during the year. As always, I’ll discuss my thoughts on the business of scuttleblurb in my annual interview with LibertyRPF ^[4]. Most of the text in the summaries below is extracted directly from the write-ups and all commentary reflects my views at the time they were written. These summaries in no way do justice to the actual posts. They are meant to be teasers that I hope will convert some free subs into paying subs and encourage paying subs to take a second look posts they may have passed over for lack of immediate interest. This will be the last post of 2024. I will begin the new year with writeups on AppFolio and Maravai.

A few other announcements:

First, on a personal note, I recently launched a gofundme for Boulder Crest Foundation ^[5] in honor of my brother, Rich, a veteran of the Iraq and Afghanistan wars who struggled with combat stress for many years before taking his own life this past March. Boulder Crest has helped more than 100,000 veterans, servicemembers, and first responders transform their struggle with trauma into strength. Their Warrior PATHH ^[6] program includes a 7-day in-person initiation retreat, followed by 90 days of dedicated support and accountability. It is offered to participants at no cost. Please join me in supporting the brave men and women who have made enormous sacrifices on our behalf. To those of you who have already donated, I am incredibly grateful. So far, we’ve raised more than $36k (including my match contribution), blowing past the original $10k goal.

Second, in 2024, I removed the paywall off two posts from prior years: Class 1 freight rails: part 3 – Hunter Harrison, PSR, and investment implications ^[7] and some thoughts on Adyen ^[8]. If you’re thinking about becoming a paid subscriber, consider reading those and other posts I’ve made freely available.

Third, I’m excited to announce that MBI ^[9] and I launched a podcast, where we casually chat about companies we’ve written-up once a month. The first episode is a discussion on AppFolio (Spotify ^[10], Apple ^[11], RSS feed ^[12]). MBI writes a terrific blog ^[13] and if you aren’t already a subscriber, well, you should consider becoming one!

Finally, thank you all so much for the support and feedback you’ve given me over the years. It is a joy and privilege to write this blog for you. See you in the new year!

On the differences in serial acquisition programs, from [AME] AMETEK ^[14]

“As a serial acquirer grows, so too does the amount of M&A required to move the needle. But because, as a general rule, the larger the deal size the loftier the valuation, it’s not just how much capital gets deployed that matters but over how many acquisitions.

….Vertical market software disaggregates across countless pockets, but software is alike enough that you might treat VMS as a giant, homogenous mass. Constellation can do 100+ deals a year because the playbook applied to running a software business that sells to restaurants also mostly applies to running a software business that sells to spas or auto repair shops. Constellation Software is less a bet on technology or product or even end market than it is a bet on a system for efficiently researching and acquiring small software companies at scale, and the more companies they buy the larger the database against which to hone base rates on future acquisitions.

So my point is: when the angle to M&A is less about technical sophistication than it is offering a permanent home for sellers plus superior execution against a sticky customer base, the particulars of product and end market seem almost beside the point. Competent managers with pride of P&L ownership, supported by the right incentives to drive value, are the critical drivers.

Teledyne is governed by a different dynamic. They do in fact acquire for differentiated technology, technology that tends to be concentrated in just a handful of companies, companies that become increasingly scarce as they become big enough to absorb the growing sums of capital that Teledyne needs to deploy every year. Also, Teledyne doesn’t want technology that is too far outside their existing competencies, which further limits where they can hunt. These constraints force them to accept lower returns as they put more capital to work. So unlike Constellation, Teledyne has deployed larger amounts of capital by buying bigger, not more.”

On the extreme pro-cyclicality of BFS’ buybacks, from Building materials: part 3a (BLDR, AEP.V) ^[15]

“…while BFS’ share repurchases are optically accretive (”share count go down”), they are extremely concentrated during periods when share prices are elevated. Why? Because BFS returns cash to shareholders when they have lots of it to spare, which happens to coincide with cyclical peaks, which also happens to be when the stock is soaring.

Buybacks = dividends. That’s Finance 101. Management can repurchase shares or pay a dividend to shareholders, who use that cash to buy shares themselves. No value is lost or gained. Same-same. But that equivalency assumes that the company is no better at repurchasing its own shares than shareholders themselves.

Buybacks generally fall into two buckets:

1. opportunistic: management has a better sense than shareholders do of whether their company trades at less than intrinsic value per share. They can buy their stock better than shareholders can. In this case, repurchases are preferable to special dividends paid to shareholders.
2. programmatic: management uses cash flow to buy shares, without any assessment of intrinsic value. The cash comes in, they turn around and buy shares with it. Assuming cash flows are reasonably stable from one year to the next, management will find itself sometimes buying stock when it is cheap (below IV) and sometimes when it is not. The company has no better or worse sense than shareholders of when the stock is cheap or expensive, so it makes no difference whether they repurchase shares or shareholders buy shares with dividends they receive from the company.

An investor would prefer 1) over 2), but 2) isn’t necessarily value destructive so long as cash flows are somewhat consistent from one year to the next and intrinsic value grows at a reasonably measured pace. An analogy here would be an investor who auto-invests her steady paycheck in an equity ETF without any regard to valuation or market timing. Some of those purchases will turn out poorly, others great, but on balance things should turn out all right so long as the diverse collection of businesses constituting the ETF create value over time.

But now imagine a business that implements a programmatic buyback but also generates the vast majority of its cash at cyclical peaks. A special dividend would be better in those cases because even an investor who uses that dividend to buy more shares but does so at random will outperform management, who systematically invests at the top and at no other point.”

SUMMARIES

GFL Update (1/3/24) (long) ^[16]

Garbage collection and disposal is one of the few businesses that you can be sure will still be around in 20 years. Its recurring necessity gives rise to predictable and stable cash flows. Residential collections are locked in under 3 to 10 year exclusive agreements that are very hard to lose, as city officials don’t want to risk pissing off voters by re-bidding contracts, a lengthy and onerous process that could disrupt garbage pickup. Commercial (retailers, restaurants) and Industrial (manufacturers, contractors) customers are secured under 3-5 year contracts that are tough to profitably steal away from incumbents who have established route density (the more customers a collector has along its route, the more cost effectively it can serve an incremental customer). They will accept hike prices without much resistance because trash collection is a necessary but small enough ^[17] component of their overall operating costs.

In its May ‘22 Investor Day, GFL projected C$1bn of free cash flow by 2024, a target that they were soon forced to walk back after being hit with soaring fuel costs and interest expense (~28% of its debt floats). These headwinds would prove manageable. Throughout 2023, management recovered underlying margins by pushing through fuel surcharges and higher base rates, and improved the balance sheet by using proceeds from the sale of certain non-core assets. But even absent the divestiture, with GFL pricing 100bps+ above cost inflation, I can see EBITDA growing by 8%/year, outpacing the 6% growth in cash interest expense and bringing leverage in line with a best in-class peer like WCN. Including incremental profits from renewable natural gas and Extended Producer Responsibility (EPR) fees takes free cash flow to C$1bn by 2025, growing to C$1.3bn by 2027, which capitalized at 25x implied a mid-teens return. In short, I see this as a low brain damage stock with a credible path to mid/high-teens returns so long as GFL manages the business responsibly and pushes prices above cost inflation. The harrowing economic climate of the last few years offered sound evidence that they can do just that.

why Cintas wins (1/18/24) ^[18]

Uniforming employees is an important but non-core activity. For safety and presentation reasons, healthcare professionals need to be outfitted with clean scrubs, electricians with flame-resistant gear, used car salesmen with matching polo shirts, etc. Companies could manage garment programs internally, but this entails various complexities they’d rather not spend time on. You could order a dozen maroon polo shirts from Amazon, but there’s no guarantee that Amazon will have in stock the same shade of maroon in a given size to outfit new employees as your business grows. If a plus-sized employee is replaced by someone with smaller dimensions, you’ll need to buy a brand new uniform while the XXL shirt and pants sit on the shelf. Over time, you will find yourself tying up more and more cash in garment inventory and volatizing cash flows with unpredictable garment purchases. By assuming inventory on their own balance sheet, a uniform rental provider can help dampen the swings and ensure you get exactly the right uniform when it is needed.

The business of buying and renting uniforms is operationally complex (not only are rental providers collecting shirts, they are cleaning and returning them to customers as well), vulnerable to alternatives (customers can always just buy shirts and pants for their employees) and, outside of enterprise accounts, isn’t protected by high switching costs, which limits pricing power. And yet, Cintas has created enormous value for shareholders, its stock compounding by 15% over the last 20 years and 27% over the last 10 (before dividends). Over the last decade, they’ve outgrown primary competitors Unifirst and Vestis on an organic basis and realize far more revenue per employee and facility than either. They’ve done so through a combination of better route density and superior execution, the latter evidenced through ERP modernization, vertical-specific offerings, and smart M&A.

Vestis recently spun-off from Aramark, an inattentive owner who mismanaged the company for many years. The turnaround story involves margin expansion and accelerating organic growth, powered by cross-selling and route optimization. I have concerns about the practical difficulties of dual-purposing unionized truck drivers as salesmen and question whether Vestis’ antiquated systems can support the logistical complexity required to compete head on with Cintas, who spent a decade modernizing its ERP.

[TDY] Teledyne (2/7/24) ^[19]

At the time it was spun-off of Allegheny in 1991, Teledyne was primarily in the business of selling electrical components and subsystems – devices that amplified microwave signals, transmitters and receivers that send signals from aircraft to ground equipment, sensors and software that stored telematics data, etc. Starting in the early 2000s, they made a series of margin accretive acquisitions that expanded their presence in Instrumentation, which you can think of as things that measure hard-to-see physical phenomena (sonar systems that map ocean floors; gyroscopes and accelerometers that measure location and speed; instruments that precisely measure gas flows and monitor air quality); and Digital Imaging, sensors that convert light waves into digital representations. These acquisitions transformed Teledyne. In 2004, the company got 80% its revenue from selling aerospace electronics and serving as a prime contractor for the US government. Ten years later, those businesses were whittled down to just 35%. The balance came from Digital Imaging and Instrumentation products used mostly in commercial applications, which has remained Teledyne’s primary focus since.

The next 7 years were characterized by progressively larger deal sizes in instrumentation and digital imaging that ultimately culminated in the whopping $8bn purchase of FLIR (short for “forward-looking infrared”), which made sensors to identify heat emissions from industrial gases, greenhouse gases, enemy combatants (night vision goggles), and overheating factory machines. At the time it was announced, the deal represented 54% of Teledyne’s total enterprise value and more than twice the $3.7bn spent on all previous acquisitions combined.

The FLIR acquisition was consistent with Teledyne’s pattern of buying adjacent technology. But the valuation was surprising. As late as 2019, they expressed bewilderment that anyone would pay 17x. Then, just a few years later, Teledyne announced they would be buying FLIR for…17x EBITDA! (more like 20x excluding the one-time COVID sales). And it’s not like FLIR was a fast growing gem of business. It had been mismanaged for some time and was growing by just 1%-2% when FLIR bought them. Even with optimistic growth assumptions and full credit for cost synergies, Teledyne paid 11x year-5 EBITDA, still a far cry from the 10% after-tax returns within 2-3 years that they’ve historically targeted.

But the collection of businesses Teledyne has rolled up over the years generally seem like good ones. Most of the high-end imaging and instrumentation niches where Teledyne plays are rational oligopolies. Their products are designed into systems with long development cycles and lengthy qualification processes, making them tough to displace. They are sold to customers who care far more about performance than price, insulating them from low-end Chinse competition.

The businesses don’t generate much growth, though. On a consolidated basis, organic revenue has grown by just 1.5% over the last decade. Adjusting for the mix shift toward Digital Imaging, forward growth is maybe more like 4%, but that does get you much more than 5%-6% organic earnings growth. You probably aren’t paying any more than 16x-17x earnings for something like that. If they can re-invest all their earnings into acquisitions at 10% after-tax returns, as they’ve done historically, you might justify the low-20s multiple that they trade at today. But can they? Will they?

[AME] AMETEK (2/26/24) ^[20]

AMETEK reminds me of Teledyne in several ways. It began its life in industrial machinery before acquiring its way to higher margin, more technically differentiated and less capital intensive areas. This includes instruments that measure and test things that are very close and very far away, stuff like: mass spectrometers and electron microscopes used by scientists to understand the structure of compounds and by semi manufacturers to detect impurities in silicon wafers; optical assemblies (lenses and mirrors that manipulate light) used by astronomers to study the planets; rheometers used by consumer goods manufacturers and coatings companies to assess how fluid-like materials behave under different stresses.

Alongside that is a smaller division that sells instruments that capture aircraft telematics, ruggedizes off-the-shelf components from other tech vendors (FPGAs from Xilinx, routers from Juniper) for aerospace and defense platforms exposed to extreme environmental conditions, supplies utilities with simulation tools to test the integrity of their grids. More recently, they’ve made a big splash in healthcare, now their single largest end market at 20% of revenue, with the acquisitions of Rauland-Borg (call stations, mandated by regulation, that patients use to call for help and that nurses use to receive those calls) and Paragon Medical (surgical instruments, bone screws, hip implants, rods, springs, wires, etc.). The acquisitions seem all over the place when viewed through the lens of end market exposure. But what they generally share in common are #1 or #2 positions in $200mn-$500mn oligopolistic niches, products with high switching costs, and, under AMETEK’s ownership, the rigorous application of kaizen values (AMETEK once referred to itself as a “mini-Danaher”). Revenue is wildly cyclical, geared as it is to a long tail of industrial markets, and organic growth through the cycle looks similar to Teledyne’s.

AMETEK has paid progressively higher multiples for higher quality, faster growing businesses (7x-8x in the early/mid-2000s to 15x for Paragon). The days of buying undermanaged companies for high-single digit multiples and doubling margins over 3-5 years are over. Compared to 10-20 years ago, they are leaning more so on growth than profitability improvements to hit their 10% after-tax return bogey within 3 years.

[ROP] Roper Technologies (3/18/24) ^[21]

Of the 3 serial acquirers I covered this year, Roper has undergone the most extreme business transformation. From its starting point in capital intensive industrial equipment, Roper acquired its way into test and measurement instruments during the ‘90s, then software throughout the ‘00s. Management was intentionally targeting capital light businesses with high margins, recurring revenue, and modest macro sensitivity, with the directive that each new acquisition should be better on those dimensions than the portfolio average. If the bar is ratcheted higher every year, then whether you originally meant to or not, you will eventually end up buying software companies.

Roper’s software acquisitions covers a wide range of verticals and are grouped into 2 divisions: Network Software that, for instance, matches shippers with truck drivers (DAT), pairs general contractors with subcontractors (ConstructConnect), and consolidates purchase activity across long-term care facilities (MHA); and Application Software, which you can think of as industry-specific ERP software that customers use to run their operations. A prominent example of the latter is Vertafore, which P&C insurance agencies and brokers use to manage sales pipelines, renew state licenses, and track agent commissions, among many other things. Given how mission critical these products are to day-to-day operations, it comes as not surprise that Roper’s software divisions retain 95%+ of customers and consistently push through price hikes.

Roper is organized in a decentralized fashion, with 28 business units, each run by a manager with near full discretion to make strategic and operational decisions, though they do so within tight governance constraints that prevent financial metrics from veering off track. Capital allocation, meanwhile, falls in the hands of a handful of executives at headquarters, who work with third party professional services firms to diligence transactions. Nearly everything Roper buys comes from a private equity, Thoma Bravo and Vista in particular. Whereas a private equity firm will cost cut their way greater profitability over a 5-year time horizon, Roper, as a “forever” home, has the luxury of optimizing for growth and creating value over long time horizons. And thanks to their investment grade rating, they can borrow a much lower rates than a private equity sponsor.

Owners in both parts of the lifecycle can realize respectable returns. A PE sponsor who pays 27x depressed EBITDA for a 5% grower, funds the purchase with 8 turns of leverage, doubles margins over 5 years, and flips to Roper at 18x, will generate a 19% levered IRR. By taking organic growth up to 7%, funding with 4 turns, and expanding margins by ~30bps a year, Roper in theory might expect ~12%-13% levered IRRs, assuming a 15-year time horizon and a terminal EBITDA multiple of 13x

This is not to say that Roper has a dominant advantage when it comes to winning deals. They can’t outbid a strategic for a target with easy synergies, or even a private equity firm for a mismanaged target with a broken cost structure. But for well-run companies with full margins competing in modestly sized markets, a long time horizon, low funding costs, and competency at accelerating organic growth allow Roper to create value to an extent that neither strategics nor private equity can.

[XPEL] Xpel Inc. (4/1/24) ^[22] (long)

Most of Xpel’s revenue comes from selling a product that you may not even know exists. And that’s kind of the point. When installed right, paint protection film is hard to see. But its invisibility belies its utility. A thin and transparent wrap, comprised mostly of a versatile material called thermoplastic polyurethane, PPF protects a vehicle’s paint from scratches, bird droppings, acid rain, bug splatter, UV rays, and gravel strikes, with “self-healing” properties that allow the film to reconstitute itself after minor assaults.

In 2023, a widely circulated short report on Xpel ^[23] made two central claims: first, that Xpel gets much more of its revenue than it disclosed from Tesla, who was bringing PPF installation in house; and second, that PPG Advanced Surface Technologies (PPG AT), a joint venture between PPG and entrotech, “have developed a way to integrate entrotech’s paint protection technology directly into PPG’s paints” and that “the JV will set to integrate this technology directly with OEMs, straight from the production line, hence virtually eliminating the need for Xpel’s clunky aftermarket wraps”, obviating XPEL’s entire business.

As my conversations with PPG and Ford (the OEM supposedly integrating this supposed PPF-paint hybrid) made absolutely clear in no uncertain terms, the existential claim about PPF technology being integrated into PPG’s paint was fabricated. The Telsa concern had a bit more teeth to it. I called a bunch of dealers and, indeed, a disproportionate number of vehicles they wrapped were Teslas. But on the whole, they were not seeing much, if any, impact from Tesla in-sourcing. Moreover, a week prior to the short report’s publication, Xpel explicitly disclosed that its “paint protection film-related revenue (including product and service) with respect to Tesla vehicles is approximately 5% of its total year to date revenues”. The short seller interpreted this announcement as a desperate attempt to support the stock, which doesn’t make a whole lot of sense.

Consider what it means for the short seller’s claim to be right. Xpel dealers must enter into DAP the car model whose templates they want to cut from film, meaning management almost certainly knows how much film revenue is tied to each car model with a very high degree of accuracy. So by privileging survey results over Xpel’s disclosure what you’re essentially saying is that management, knowing the real percent of revenue coming from PPF installation on Teslas, deliberately published a fake number instead. So, after rescuing Xpel from near ruin and methodically building it over 15 years to a profitable market leader that is firing on all cylinders, Ryan Pape just decides “screw it, time to commit blatant securities fraud and risk career, personal reputation, and financial consequences that are bound to be far more disastrous than just saying the real Tesla number?” That’s a pretty bold claim, in my opinion.

Revisiting Burford (5/2/24) ^[24]

Imagine you are General Counsel of a F1000 company and want to sue a competitor for violating your company’s intellectual property. While you have strong case, litigation is an inherently speculative undertaking. You don’t know whether the defendant will settle or fight and, in the latter case, whether a judge or jury will rule in your favor or even how long it will take for them to do so. Rather than run ongoing litigation expenses through the income statement, crimping the reported earnings that investors capitalize, in return for an uncertain payoff, which investors dismiss as a one-time event, you would much prefer the law firm you’ve hired to work on contingency, bearing the expense of litigation and sharing the spoils of a successful outcome. But the law firm, culturally and financially ill-suited to assume such risk, prefers the assurance of hourly fees.

This is where a commercial litigation funder like Burford steps in. Burford will commit to paying the law firm’s fees up to a certain amount and in turn take a percentage of any resulting settlement or damages paid. The client gets a free option on litigation; the law firm gets fixed hourly fees to pursue damages even as it serves the client on a contingent fee basis; and Burford, who pays fixed hourly fees to the law firm or corporate client, gets a piece of any resulting payoff.

In what seems like ages ago now Fintwit darling, Burford Capital, was flamboyantly attacked by Muddy Waters. The prominent short seller claimed that Burford “misleadingly boosted its IRR numbers”, rendering ROIC and IRR metrics “meaningless”; misled investors through the “egregiousness of its fair value accounting”; and was “arguably insolvent”, raising outside capital as a matter of survival rather than growth. These charges followed on the heels of a Canaccord Genuity analysis from Apr ‘19 that found Burford’s reported ROIC “confusing and significantly above reality”.

For most of this post, I once again provided a long-winded rebuttal of that bear case. My argument largely resolves to:

contrary to common perception that Burford’s historical returns are the result of one-off windfalls that we can’t bank on going forward, I would argue that settlements provide an engine for repeatable outcomes – outcomes that deliver, if not spectacular, at least solid returns – for approximately 70% of Burford’s invested capital. Moreover, Burford boasts an enviable track record outside of settlements. Just over 70% of capital deployed on adjudications from inception to 2023 has generated gains, up from 65% from inception to 2019.

and…

it appears much harder to compete away returns than many bears imagine. There is only so much capital that can be put to work in any given deal, the underwriting process doesn’t scale well, and its accuracy is aided by proprietary data that takes years to acquire. Given the resources, time, and lack of scalability, not to mention the added complication of suing enterprises that you’d like to maybe have business relationships with, I somehow doubt that someone like Blackstone is taking this market all that seriously.

But the lack of scale also pushes back against the bull case. Sure, Burford might generate strong asset level returns. But lawyer salaries (litigation finance deals don’t scale), finance costs, and taxes consume a lot of this. Consider that Burford’s tangible book value per share ex. YPF (a one-off windfall that we can’t really expect to repeat) has only grown by between 8% and 11% over various 7, 10, and 12 years. And from 2010 to 2015, before YPF was even in the picture, TBV/share grew by just 5%/year. While the YPF proceeds alone, should they be paid (a big “if”), account for more than all of Burford’s enterprise value, the most committed bulls seem to believe the stock is only pricing in the value of Burford’s non-YPF business (i.e. “you are getting YPF for free”). I don’t agree. Ex. YPF, Burford’s stock trades at 4x TBV. Having grown TBV by just 8%-10% historically, why should they be valued at such a lofty multiple?

[LW] Lamb Weston (5/28/24) ^[25]

Lamb Weston is the largest frozen potato processor in North America, with 40% share of a market dominated by 3 players, all of whom have origins dating back to the early 1900s. More than 80% of its production is concentrated in Oregon, Washington, and Idaho, giving it more exposure to the highest quality crops than Simplot and especially McCain, who compared to Lamb are more heavily skewed to the Midwest and East Coast.

Over the last decade, Lamb Weston has generated ~high-20s pre-tax returns on capital. Why couldn’t a new entrant compete these mammoth returns away? Well, what management and industry folks will often say is that even assuming a new entrant had access to the ~$500mn+ required to build a processing plant, they’d have to somehow dislodge the decades-long relationships that the big 3 have cultivated with farmers and customers. In 2016, Lamb disclosed the average duration of its customer relationship to be 28 years, with its longest going back 45 years. Meanwhile, Lamb’s average farmer had been growing for them for 15 years, 35% of them for 20 years. Also, new plant builds are usually anchored by volumes from a major QSR. For McDonald’s, french fries are a menu item they absolutely can’t screw up, so why roll the dice on a new processor? In securing QSR volumes, the big 3 enjoy scale production economies and procurement savings on key processing inputs that a newcomer does not.

Of course, this doesn’t immunize Lamb’s earnings from volatility. The abnormally hot summer in fy21 produced potatoes with unusually low starch content, which on a largely fixed cost manufacturing base, resulted in higher per pound processing costs, and forced Lamb to buy potatoes in open market at huge premiums to meet volume commitments to customers. Rising potato prices, along with labor shortages and double-digit inflation in transportation and oil costs, caused Lamb’s margins to contract significantly in fy22.

Lamb successfully pushed price through to its customers, recouping incremental costs from the heat wave and supply chain shortages. It was able to do so because frozen potato processing capacity in North America has been very tight over the last 7 years, well above the historical range of 93% to 97%. The 5.6bn incremental capacity that the industry is expected to add through 2027 supports strong utilization assuming demand grows between 2% and 4%. But that’s hardly a given. USDA data shows several 5-year stretches where per capita demand was flat-to-down 1%. In that scenario, capacity utilization would fall all the way down to 89%. By my estimate, it appears we are looking at ~10% more capacity in North America, outpacing lsd demand growth. Lamb Weston would be more exposed than others because its idiosyncratic volume challenges – the purposeful shedding of 4 low-margin contracts and a botched ERP rollout resulting in unfulfilled orders, which exacerbate headwinds from deteriorating restaurant traffic – suggest its capacity utilization is likely starting below that of peers.

quick Xpel update (5/28/24) ^[26]

I bought some XPEL shares in March and wrote some nice things about them. Naturally, soon after, the company reported an awful quarter and the stock cratered. Management cited broad aftermarket weakness in the US – “it was not uncommon to see dealers who were down 10%, 15% in the first quarter from the prior year period” – and a horrendous 78% revenue collapse in China.

But there were some silver linings: 1) port delays in the US caused a ~20% reduction in sales of Porsche’s and Audi’s, two of Xpel’s top brands for film coverage, and this disruption was resolved toward the end of the quarter, perhaps explaining Xpel’s atypically strong m/m growth in April; 2) outside of the US and China, Xpel continued to grow at an impressive clip, as did 3) revenue from Installation Services, where Xpel realizes revenue from selling and installing film itself.

IS growing 35% despite double-digit declines in retail implies really strong growth from the new car dealerships and OEMs that make up the rest of this segment. I think this is an important point, one that is easy to overlook. A key unresolved matter for the bull case is whether PPF adoption can spread beyond enthusiasts. That Xpel continues to report strong growth at auto dealerships and OEMs, two critical channels for reaching mainstream car buyers, is evidence that it might.

CEO Ryan Pape again directly addressed the allegation that a disproportionate amount of revenue comes from Tesla:

“There’s this idea that all of the revenue is concentrated in one brand or 2 brands or something. And that’s really not the case. I mean, I think we talked about the type of make related concentration last year, like 5% or less…I really just can’t stress enough that there isn’t a single point concentration risk into any one vehicle in this business”

I later followed up with Ryan. One key point from my conversation with him: about 30% of Xpel’s revenue comes directly from dealerships or from aftermarket installers who work only with new car dealerships. None of these installers or dealerships work on Tesla’s and none of them will be found on Xpel’s “Find An Authorized XPEL Installer” tool that both the short seller and I used to source leads. So, if your diligence consists of calling aftermarket installers listed by Xpel and asking “what percent of your installations are Tesla’s”, you are missing 30% of revenue for which the answer is “0”.

[MKTX] MarketAxess (6/19/24) ^[27]

The market for fixed income trading platforms has consolidated into an oligopoly, with MarketAxess, Tradeweb, and Bloomberg at its center. A number of other competitors have emerged over the years, some are still around (Trumid), others have failed (Bondcube, Algomi). But, for the most part, when an asset manager like PIMCO buys Treasuries or corporate bonds from a sell-side dealer or even another buyside participant without picking up the phone, more likely than not they are doing so through one of those 3 platforms

When I first wrote up MTKX and TW in November ‘20, both companies looked like inevitable winners as bond traders flocked to their electronic platforms in search of liquidity. Since then, their results have diverged starkly. Through net spotting, Tradeweb successfully leveraged its Treasury franchise to claim significant volume in high-grade credit. It was also early to market with Portfolio Trading, a protocol that allows a large collection of bonds to be bought and sold at a single net portfolio-level price. PT’s growing popularity has coincided with the ascent of fixed income ETFs, as the ability to efficiently trade large bundles of securities in one go is especially useful for market makers involved in the creation/redemption process, where ETF shares are exchanged for the cash bonds underlying them.

MarketAxess also began offering its own version of Portfolio Trading but never took it as seriously as Tradeweb, fixated as it’s long been on pushing Open Trading, which allows all market participants to anonymously trade with one another. Since launching in 2013, OT has become the central point of differentiation for MarketAxess, intermediating ~34% of its volumes. It is their ambitious attempt to create a new market structure, one that they hope will come subsume all major protocols and fixed income asset categories. OT tends to thrive in high vol environments, when liquidity is scarce. But if the last decade is any indication, low volatility appears more the norm than management would like to admit.

That MarketAxess has fallen so far behind Tradeweb in PT and so aggressively pushes OT also points to a philosophical difference between how the two platforms approach the broader ecosystem. Perhaps as a result of its origins as a dealer-backed consortium and its substantial dealer-to-dealer activities, Tradeweb has always worked more in harmony with the existing, dealer and phone-dominated ecosystem. MarketAxess, on the other hand, has never never been shy about placing themselves at the center of the ecosystem, with dealers and investors serving as spokes around the Open Trading hub. I suspect that by pushing OT so doggedly, MarketAxess alienated broker-dealers and neglected protocols, like Portfolio Trading, that depend on them, creating an opening for Tradeweb, who has always maintained a less revolutionary, more cooperative posture.

Building materials: part 1 (7/11/24) ^[28]

This multi-part series concerns the building materials (BM) industry, specifically the layer downstream of timber growers and upstream of home builders, home centers, and remodelers. Investors applaud canonical compounders like NVR and Home Depot and have in recent years been losing their minds over Floor & Decor. But beneath that strata of the value chain is a thriving ecosystem of manufacturers and distributors that is rarely discussed by generalists or even dismissed as unworthy of study.

With lumber prices soaring during COVID, the last few years have been pretty wild for those – like BLDR, BXC, BCC, and LPX – most geared to the vicissitudes of commodity wood pricing. BLDR and BXC free cash flow’ed 1.6x and 3.8x their 2019 year-end market caps, respectively, in just 2 years!

A big chunk of the post explores the causes of soaring lumber prices during COVID and quantifies the extent to which the US housing market remains in deficit. Towards the end, I discuss BlueLinx, a “two-step” distributor that sources wood products from mills and distributes them to lumber yards and home centers, who in turn re-sell to builders and contractors. BlueLinx has had a long and troubled history since it was taken public by Cerberus in 2004. With 4 turns of leverage on peak earnings, the company spiraled into near oblivion when the housing crisis hit.

Fortunately, BlueLinx owned real estate whose value nearly covered its enterprise value and over the next decade, they sold and leased back properties to pay down debt. Along the way, they tripled their share count with two massive rights offerings and downsized operations. Nevertheless, BlueLinx spiraled in the shitter for a while. In 2016, even with the worst of housing crisis in the rear view, BlueLinx’s market cap plumbed new lows and its stock, trading below $1, was on the verge of being delisted. The first break in this dolorous state of affairs came in the form of Cedar Creek, another two-step distributor that BlueLinx acquired in 2018. But the integration was shaky, one of Cedar Creek’s major siding programs was lost, housing starts were weak, and commodity deflation offset much of the realized cost synergies. By the end of 2019, before COVID was on the radar, the stock had collapsed ~80% from its 2018 peak. At ~$14, shares traded for less than they did before the transformational, synergy-rich merger was announced.

Then COVID hit, lumber prices went ape shit for a few years, and BlueLinx reported gross and operating margins that it never came even remotely close to achieving in its history as a public company. Management thinks that after the COVID craze, things have now settled to a sustainable place. But gross and EBITDA margins are still well above historical norms and I’m skeptical that whatever operational gains they’ve made are enough to counter mean reversion, especially considering the structural challenge of major home centers and lumberyards sourcing directly from mills.

Building materials: part 2a (LPX, JHX) (8/5/24) ^[29]

By the mid-90s, Vinyl overtook wood to become the most common siding material in the US. Unlike wood, vinyl siding can be mass produced in a factory. It isn’t susceptible to insect infestation or water rot, and endures for more than 20 years without maintenance and repaint. It is far lighter and easier for contractors to handle and install. To a homeowner, the cost of a vinyl siding project is ~1/10 that of hardwood. But Vinyl has important downsides too. It warps in extreme heat, cracks in extreme cold, and damages more easily when exposed to strong winds and hail. The wood patterns that manufacturers emboss onto it fade over time. Vinyl is used because it is reliable and cheap, not because it is pleasing to look at.

Homeowners willing to yield on price can get the durability of vinyl and the aesthetics of wood or stucco through two other options: fiber cement (a hardened slurry of cement and wood fiber) and engineered wood (an enhanced formulation of OSB). Of the dozen or so exterior cladding options commonly available, these two have gained the most prominence over the last 20 years. Their popularity has come at the expense of vinyl, whose share of siding has declined from 35% in 2010 to around 20% today. The fiber cement and engineered wood siding categories are near monopolies, with James Hardie claiming 90% of fiber cement and LP’s SmartSide brand claiming ~85% of engineered wood. Both are vice-gripped into the value chain by the pull of homeowners and the push of contractors and dealers, making them difficult for competitors to dislodge.

That LP has come to command a meaningful presence in a branded specialty category like siding is surprising given how rooted it is in OSB, a classic commodity. Whereas OSB is a commodity whose profitability is largely governed by pricing and single family housing starts, which are impossible to predict and fluctuate wildly, Siding is a branded business with steady pricing power that derives 60% of revenue from relatively more stable repair & remodeling demand. Siding margins are higher and far less volatile too, with profitability far more a function of production efficiencies and capacity utilization than commodity price trends. So, over the years LP has been converting legacy OSB plants to siding plants and within siding, mixing volume toward still higher value ExpertFinish lines, creating a visible path to higher margins and returns on capital.

Building materials: part 2b (TREX, AZEK) (8/25/24) ^[30]

The themes in residential decking are similar narrative to those animating residential siding. Wood is still used in 75% of decking projects, but that’s down from 95% in 1999 and probably on its way to 40%-50% as share continues to shift to composite materials (wood fiber + plastic), as composite has consistently taken anywhere between 1 to 2 points of share from wood per year over the last decade, with gains fueled by expanded selection and enhancements in durability and appearance. Every point of share donated by wood translates to 3%-4% growth of the composite category as a whole. The premium one pays for composite relative to wood is more than made up for by lower maintenance and superior longevity. Over a span of a quarter century, the total cost of ownership of a wood deck is more than twice that of one made of composite.

Within the composite category, Trex claims about 40% share and Azek has another ~30%. As is the case in residential siding, Trex is “vice gripped” into place by unrivaled brand awareness among homeowners (90% of surveyed homeowners are aware of the Trex brand and more than 60% of web traffic in the decking category is claimed by Trex-owned sites, an important source of leads for qualified remodelers who are fortunate enough to be listed) and exclusive relationships with a handful of two-step distributors in each of their markets. Trex is the only composite brand stocked on shelf at both Lowe’s and Home Depot, reinforcing its position as a trusted consumer brand.

Trex resembles James Hardie in several respects. James Hardie commands 90% share in fiber cement category, which itself comprises 20% of the overall siding and trim market. Trex has ~40%-50% share of composite decking, itself just 25% of the overall decking. Both James Hardie and Trex are established brands with robust channel support. Both expect ~10%-12% organic growth, supported by remodeling spend and category penetration against vinyl (in James Hardie’s case) and wood (in Trex’s). Both generate 30s EBITDA margins with ~35%-40% incrementals, translating to ~35%-40% pre-tax returns on gross capital (gross PP&E + working capital) at scale. And both strike me as appropriately priced for the high quality businesses that they are.

Building materials: part 3a (BLDR, AEP.V) (9/16/24) ^[15]

Builders FirstSource (BFS), the main subject of this post, is a pro dealer that distributes the structural components that comprise the skeleton of a house….framing lumber, OSB, engineered wood products, and floor and roof trusses, as well as complements like wall panels, siding, doors, and window frames. They sell hundreds of thousands of SKUs across ~570 locations in 43 states, boasting #1 or #2 share in 89 of the top 100 MSAs. Around 2/3 of revenue is tied to single-family home sales, another 20% and 13% to repair & remodel and multi-family structures, respectively.

BFS has been transformed through so much M&A that I think of it less as a consistent, singular identity than a corporate name affixed to ever more grandiose expressions of the idea that pro dealers can better and more profitably serve builders by moving the production of structural components from jobsites to manufacturing plants and by offering a broad assortment of SKUs across a dense, nationwide network of branches. The Builders FirstSource that we know today is a nested collection of rollups, each migrating up the value stack and reaching new markets in parallel before fusing together under a single corporate owner:

Like BlueLinx, BFS had a tough go of it during and in the years following the GFC. But by 2014, years into a tepid housing recovery, Builders began to flicker back to life. Losses inflected to profits and free cash flow dribbled out. Over the next decade, BFS didn’t merely resume its small-ball M&A program; they rage-acquired their way to the top, as if making up for lost time. Since 2004, Builder’s has spent $7.9bn on acquisitions, $7.8bn of which has been concentrated in just the last decade. Two acquisitions – ProBuild (July ‘15; $1.8bn; 5.5x post-synergy EBITDA) and BMC (Jan ‘21; $3.7bn; 8.4x post-synergy EBITDA) – account for around 2/3 of that spend.

As BFS amassed scale and influence, they increasingly sourced lumber directly from mills and value-added SKUs directly from manufacturers, keeping more margin for themselves. The more significant way Builders has evolved, though, is in thinking about the homebuilder’s P&L rather than their own. BFS could have limited their activities to buying mixed lengths of lumber from mills; chopping them up to lengths required by builders; ensuring the right planks are delivered to the jobsite on time for the next build phase; and always staying in stock. But they extended to other facets of a homebuilder’s job: hammering lumber planks into frames and trusses; streamlining the build process to hasten turnaround times; providing installation services through third party subcontractors; and even managing a digital platform through which builders can not only order products and pay bills, but even 3D model projects and schedule workflows.

Still, the emphasis on prefab components should be appreciated, not as an isolated goal, but as central to the broader strategic aim of running an “everything store” for homebuilders’ lumber-based material needs. BFS will bundle relatively stickier and relationship-driven value-added SKUs with commodity lumber as part of an overall package, yielding what they must to maintain share in the latter while re-capturing lost economics through the former.

Like BlueLinx, BFS’ gross and EBITDA margins exploded higher during the COVID years and while they have come down a lot from the peak, they still remain well above historical averages. Management believes this a new normal but I am skeptical.

[AEP.V] Atlas Engineered Products (10/3/24) ^[31]

Atlas Engineered Products is a thinly-traded nanocap based in Canada and the only publicly traded pureplay truss manufacturer I’m aware of. It was founded by its current CEO, Hadi Abassi, who in 1999 paid C$50k for a small truss plant in Nanaimo (Vancouver Island), taking it from less than C$100k in revenue to nearly C$6mn over the next 16 years. He observed that in Canada, truss manufacturing was scattered across mom-and-pops and took Atlas public in 2017 through a reverse merger with the aim of rolling them up. Seven years later, Atlas owns 8 plants, the 1 plant in Nanaimo that Hadi acquired in 1999 plus 7 more picked up through acquisition.

There are 3 key pillars to the bull case. 1) the Canadian housing market is very tight. 2) Atlas pays low-single digit EBITDA multiples for acquisitions and realizes huge improvements in profitability by modernizing those acquired plants, sourcing lumber directly from mills, and cross-selling joists. 3) by outfitting plants with robotics, they will be positioned to “double output and reduce labor costs by 50%, and that is very, very conservative”.

Some pushback:

First, the implied multiple that Atlas paid for its largest acquisition, LCF, was very much deflated by an EBITDA denominator that surged during the COVID lumber boom. Second, I am having trouble reconciling the reported financial results with Atlas’s bullish qualitative claims. To put it succinctly: if I sum the trailing revenue of all of Atlas’ acquisitions at the time they were announced (adjusting LCF to a lower, more normalized level) and add to that Atlas’ fy17 (pre-acquisition) revenue of C$8mn, I get to C$56mn. Atlas reported C$53mn of revenue in the 12 months through June ’24. If you do the same exercise with EBITDA, you get C$10mn compared to the reported figure of C$9mn. So…where are the organic growth and cost synergies? And given that Atlas paid just ~4x+ EBITDA for most of its plants and maybe more like 7x-8x mid-cycle for Hi-Tec and LCF, you’d expect something greater than 11% pre-tax ROIC.

Finally, I spoke with an executive who observed robotics in use at Trussway and then at BFS. He was outright skeptical of Atlas’ theoretical return assumptions. Differences in labor costs, plant setup, and other implementation details could account for some of the gap. But the productivity gains that Atlas is proposing – triple digit leaps in output with a fraction of the labor cost – would be impossible to overlook if they were, in fact, being realized by others in practice today.

I know many of you are bananas about Atlas, so don’t take this the wrong way. I think there a several interesting aspects to the Atlas thesis, including gusty tailwinds from a housing-constrained market and the leadership of a committed and dynamic founder with considerable skin in the game. Still, given the absence of proprietary IP and learning curves, as well as the commodity nature of trusses generally, I think bold proclamations about a robotics-driven structural ROI reset should be received with some skepticism.

Wise and the business of cross-border transfers: part 1 (10/16/24) ^[32]

Wise is a cross-border money transmitter that bypasses the labyrinthine SWIFT network by taking on laborious work of establishing direct connections with local banks and, increasingly, direct connections to a country’s payment rails. In doing so, they are able to offer their customers significantly cheaper and faster cross-border transfers than conventional banks. On both dimensions, speed and price, Wise outcompetes traditional banks, who take around 2 to 5 business days and charge anywhere between 2% and 7% to move money cross-border.

Nor do banks offer an experience that is anywhere near as clean and transparent. The process of sending a transfer through the Wise app is about as straightforward as it gets. You pay-in funds to Wise through a domestic transfer or card payment and enter the foreign recipient’s account details. After running KYC and fraud checks in the background, Wise will convert currency at a mid-market rate and clearly disclose the transfer fee, FX rate, and estimated time it will take for the recipient to receive their money, with live updates and timestamps at each step of the transaction. This feels like magic to anyone accustomed to sending international wire transfers through bank ^[33]s, who usher users through clunky displays, hide fees through marked up exchange rates, provide no guidance on when funds will reach recipients nor any visibility on which checkpoints have been reached in the transfer chain.

Wise operates according to a cost-plus model, where any cost reductions beyond those required to sustain a target margin are recycled into lower take rates for users. This policy is enforced at a granular level, such that no corridor, product, or customer segment is allowed to cross-subsidize another. Wise is fanatical about being the cheapest and fastest money transmitter. Its maxim, “Mission Zero”, encapsulates the continuous erosion of fees up until their ultimate elimination. The dual mandate of lowest price and sustainable profits requires an intense focus on whittling down unit costs, which it does not only by establishing connections to bank partners and local rails, but by running scalable cloud infrastructure and exercising frugality in aspects of the business that do not directly tie to customer benefit. Conventional banks have neither the bilateral connections, the culture, nor the modern technology stack to compete.

Flywheel: more volumes means lower transfer fees and more leverage against fixed costs of engineering and product development, with cost savings plowed right back into lower transactions fees for customers, who in turn promote the service to friends and family.

Wise and the business of cross-border transfers: part 2 (11/4/24) ^[34]

From its origins in cross-border transfers, Wise expanded into a number of other products, the most significant of these being Wise Account, which allows small businesses and consumers to store and receive money denominated in more than 20 different currencies, in much the same way a local bank account holder would. Wise Account then sets the foundation for two other product extensions: Wise Debit card and Wise Assets. With a Wise Debit card, customers can spend in as many currencies as Wise supports. With Wise Assets, customers earn a return by investing their Wise Account balances in low-risk interest-bearing assets or in the iShares World Equity Index Fund.

I admire the careful and methodical way Wise has expanded its footprint. A lesser fintech would have taken a top-down approach to product development, placing bets on huge TAMs as they lurched toward a super app that cross-sells every conceivable financial service a consumer might need. A fintech that commands a desired end state into existence rather than allows its perimeter to be organically stretched is like a city, erected overnight, that lacks a certain cultural middleware tying things together.

Wise doesn’t work down from a sprawling, grandiose vision. Instead, product ideas bubble up from user demand and are only pursued if they directly reinforce the cross-border core. By this, I don’t just mean using an existing product as an on-ramp to another one (like PayPal offering crypto trades to P2P customers because it can) but ensuring the new thing directly leverages the thing it is best known for and even strengthens it. Wise doesn’t underwrite personal loans to consumers because, well, that’s got nothing to do with cross-border transfers. But it did launch a multicurrency wallet because: 1) that product takes advantage of the licenses and liquidity pools it had spent the prior 6 years building for international money transfers and 2) compared to drawing from an external bank account, doing so from funds already parked at a local Wise account accelerates transfer speeds, which reduces the time Wise is exposed to FX movements, lowering FX hedging costs. And because money is already in Wise’s account on the sender side before funds are sent to recipients, Wise Account transfers carry lower fraud risk. The resulting savings can then be plowed right back into lower transfer prices. Customers who park funds in Wise Accounts then need a medium through which to spend those funds and a way to generate returns in the meantime, hence debit cards and Wise Assets, respectively.

Some have speculated that stablecoins may come to represent an existential threat to Wise, but I don’t see it. A world where stablecoins are ubiquitously adopted for payments is one where, by extension, US dollars are too and, again, there are valid reasons why non-US countries would be wary of accepting dollar hegemony. Assuming local payments remain denominated in local currencies rather than in digital currencies backed by US dollars, stablecoins on the blockchain will eventually need to off-ramped.

That’s not to say stablecoins don’t have a place in cross border transactions. The US dollar is already the dominant vehicle currency for invoicing and settling global trade outside of Europe, even when US parties aren’t involved. Here, I can imagine stablecoins, which have the same value as US dollars but are cheaper and faster to transact with, replacing actual dollars and trade partners keeping those stablecoins in perpetual circulation. But Wise is not just in the business of transferring value across borders. It is in the business of transferring value across borders and converting that value to usable form in the recipient’s country. There is a big difference between the two.

The bigger concern is far more pedestrian: Wise might be a low-cost provider in cross-border transfers, but a neobank with a full array of services could be a low-cost provider of banking, generally, by virtue of how they bundle and cross-subsidize lending, asset management, primary checking, and other products. A neobank like Revolut white-labels rails from CurrencyCloud and could offer cross-border transfers to its customers at cost or, up to a certain limit, even subsidize losses through the subscription fees they charge to customers.

what did Warren Buffett see in Simpson Manufacturing? (11/25/24) ^[35]

The next time you find yourself at the site of a half-built home, direct your attention to the joints of its wooden frame. You’ll notice a variety of nondescript steel components tying the planks together. Those are called wooden connectors and there’s a better chance than not that are made by Simpson Manufacturing. The company claims around 75% share of the North American market.

Punched out of high grade American steel, Simpson’s connectors are stress-testing in company-owned labs. Using massive hydraulic presses, Simpson can simulate the forces of a 7.0 magnitude earthquake or Category 5 hurricane on a four story building and evaluate how well their components hold up on various species of wood. The rigorous testing regiment establishes credibility with officials who are responsible for specifying minimum design and construction standards. Simpson’s products were referenced in more code reports than those of any other competitor. And not only does Simpson meet code requirements but, as the industry leader with the most rigorous testing requirements, they work closely with standards bodies and code officials to develop them as well.

To be clear, Simpson’s connectors aren’t literally written into building codes. It’s not like builders are mandated by regulation to buy Simpson-branded products. Codes will specify, for instance, that a wall withstand forces of a certain magnitude, but architects are given free reign to meet that specification however they want. It’s just that, if an architect wants to include an airy wall with lots of windows while still meeting code, the easiest way to do that is with connectors…and so, blueprints will often specify something to the effect of “Simpson connector or equivalent”. The builder decides which connectors to use but they will more often than not go with Simpson because the Simpson brand is specifically called out as a reference in code reports and has over the last 70 years established itself as the industry standard. So while Simpson connectors are not “spec’ed” in by regulation, for all practical purposes they may as well be.

This has the makings of a good business. A manufacturer of engineered components, de-facto spec’ed into design, essential to the structural integrity of a building but comprising a small fraction of overall costs, Simpson brings to mind widely celebrated compounders like Texas Instruments and HEICO. You can understand why Warren Buffett once reached out to Barclay about buying this company.

Having said that, over the years management has made a bunch of acquisitions in Europe that, on the whole, have been value destructive. Frankly, I do not understand Simpson’s preoccupation with Europe and see their continued involvement there, even after decades of subpar returns, as an irksome red flag in an otherwise accomplished track record.

[SNA – Snap-On] when brand meets distribution (12/12/24) ^[36]

Snap-On is best known for its Tools segment (40% of revenue), through which it sells tools, tool storage boxes, and handheld diagnostic instruments through 4,800 vans worldwide, 3,400 in the US. Those vans are owned or leased by a Snap-On franchisee and stocked with about 4k SKUs (out of a catalog of 40k), which are sold to mechanics along a designated route. The average franchisee tenure is ~15 years and annual turnover (including retirement) amounts to ~10%.

Why do more franchisees sell Snap-On over other brands? Because mechanics go nuts over Snap-On tools. When Frost & Sullivan asks technicians who makes the best hand tools, nearly 60% answer Snap-On. The second highest ranking brand garners less than 20% of the votes. The fandom accrued over its storied 100+ year history manifests in pictures of newborns with Snap-On wrenches in their hands and small Snap-On tool boxes repurposed as sacred urns containing burial ash.

Doesn’t the long-lived nature of Snap-On’s premium tools limited replacement demand? How many different tools does a mechanic really need? A lot, actually, and that number grows with the variety of car models. When auto OEMs design a new car, their attention is solely dialed in to reliability, safety, appearance, and features. The last thing they are thinking about is whether that new car model’s design is backward compatible with tools that were custom made to repair prior models. Instead, OEMs give Snap-On a heads up on a new model’s unique design features, so that Snap-On can design custom tools and distribute them to the OEM’s dealers. Through its 3.4k franchisees who make contact with ~850k technicians every week, Snap-On also has a recurring presence in repair shops, where it gathers intelligence about novel repair issues that its customers encounter.

The Repair Systems & Information (RS&I) segment, 30% of total revenue, sells capex-like equipment – tire aligners, car lifts, electronic parts catalogs, shop management software, and advanced diagnostic systems – to the independent garages and OEM dealerships that employ those mechanics. The most proprietary and compelling aspect of RS&I is SureTrack, its repair database. With 2.7bn repair records and 412bn records, much of which is proprietary, SureTrack boasts the most comprehensive repair database in the industry, providing time constrained technicians with a short cut to what is likely the highest probability fix for any given trouble code. This proprietary data is also inaccessible to third-party diagnostic instruments, giving Snap-On’s own tools a competitive leg up.

The Commercial & Industrial segment, 25% of revenue, gets around half of C&I consists of Bahco, which sells hand tools in Europe to the trades through third party distributors. The other, more strategic half is what management often describes are “the Snap-On brand rolling out the vehicle repair garage to other industries which are critical”, industries like military (the segment’s largest exposure), aviation, mining, and oil & gas, where the cost of failure is unacceptable.

Tools, Repair Systems & Information, and Commercial & Industrial are Snap-On’s operating divisions, with Tools being the most competitively differentiated of the three. Together, they grow topline about ~3%-4% organically and bottom line a few points faster than that. Running alongside those is the financing division, Snap-On Credit (SOC), which mostly serves the tools segment. Seeing as Snap-On’s customer base is predominantly made up of technicians with sub-prime credit, one might expect the SOC portfolio to be prone to blow-ups. But that doesn’t appear to be the case at all. Mechanics depend on tools and diagnostic systems to make money. Only in dire circumstances will they discontinue payments on products essential to their livelihoods. Those payments are collected by franchisees who have developed an understanding of the creditworthiness of the mechanics they serve after calling on them week after week. After backing out the value of SOC, the operating businesses trade at about 18x net earnings, which seems about fair.

Disclosure: At the time this report was published, accounts I manage owned shares of $GFL and $XPEL. This may have changed at any time since

Snap-On’s origin coincides with that of the US auto industry. In 1920, an engineer by the name of Joseph Johnson created a set of 10 sockets of varying dimensions that interchangeably “snapped on” to 5 different handles, allowing mechanics to rationalize their toolset (“5 handles could do the work of 50”). When presenting to garage […]

The next time you find yourself at the site of a half-built home, direct your attention to the joints of its wooden frame. You’ll notice a variety of nondescript steel components tying the planks together and anchoring them to the ground. There’s a better chance than not that those are made by Simpson Manufacturing, a […]

Related post: Wise and the business of cross-border transfers: part 1 I concluded part 1 describing how, by looping unit cost advantages back into price, Wise scaled to become a leading player in cross-border payments. From that starting point, it launched several other notable products. The most significant of these is Wise Account, which allows […]

Every year, consumers the world over move ~£2tn across borders, with roughly 2/3s of that funneled through a labyrinthine network of banks^{1 [37]}. A typical cross-border bank transaction might look something like this: Amber, a US resident with a checking account at PNC, places an order to send Rupees to her mother’s account at Seylan Bank in Sri Lanka. After ensuring there are sufficient funds in Amber’s account and running proper KYC checks, PNC directs that order to SWIFT, a communications protocol connecting more than 11k financial institutions. The PNC-originated order is routed to a US correspondent bank like Citi, which then passes it an Indian correspondent like HDFC Bank, ultimately reaching its final destination via HDFC’s direct connection with Seylan Bank in Sri Lanka.

Every bank involved in intermediating this transaction assesses a transfer fee, the sum of which rolls up to Amber as an Overseas Delivery Charge that might run anywhere between, say, ~$20 and $70 depending on the number of hops and which borders are crossed. Those fees are sometimes waived for appearances sake but smuggled in through inflated exchange rates. Banks also impose cutoff times for processing transfers. Amber might place an order at 2PM Pacific, but Citi may not receive the funds from PNC for several hours, at which point HDFC’s daily cutoff time may have already passed, leaving Citi to hold the funds overnight before sending payment instructions through. Depending on the number of time zones crossed, delays at one stage of the transfer chain can cascade downstream, compounding settlement lags. In the end, the sender typically pays between 2% and 7% of the principal in fees, while the recipient waits anywhere from 2 to 5 days to receive funds.

For all its faults, SWIFT is a useful innovation. It provides a consistent way to transmit payment details from one bank account to any other. Without it, any bank hoping to offer global money transfers would need to establish direct bilateral relationships with tens of thousands of other banks. SWIFT eliminates that upfront burden, but raises the ongoing cost and complexity of transmitting money from one country to another. You can imagine the opposite tradeoff, though, where an entity bears the significant upfront burden of connecting to many individual banks, but in return avoids the cumbersome process of hopping money across several nodes. That’s what Wise set out to do.

Source: Wise prospectus

In each country, Wise integrates its systems with those of local bank partners to exchange, in some standardized format, files containing payment and accountholder data. Bank partners earn fees for processing transactions on Wise’s behalf. Although these fees are significant, accounting for 27% of Wise’s total costs, they shrink as a share of rising transaction volumes, making scale essential to achieving a cost advantage.

Securing the proper licenses – which cover everything from transfers in, transfers out, different product lines (debit cards, asset management) and customer segments (personal vs. business) – to send and receive funds through partner banks is typically a multi-year diligence process in which regulators review a transmitter’s business plan, volume expectations, sanctions screening practices, and collateral holdings. Wise was finally granted a “Type 1” business license in 2023, allowing per transfer amounts greater than JPY 1mn, a full 7 years after it first launched payment services there. That said, credibility earned in one major developed market makes it easier to gain a foothold in others, and experience navigating one regulatory swamp provides a playbook for the next. It took five years for UK regulators to approve Wise’s connection to FPS; in Singapore, the process took just six months. A fintech starting from scratch could acquire the same licenses and recreate the web of relationships that Wise has weaved with banks and regulators since 2011. Still, doing so would take a hell of a long time and by that point Wise’s ever mounting volumes will have pushed it still further down the cost curve.

For senders, a series of point-to-point relationships is a considerable improvement over SWIFT. By cutting out correspondent banks and connecting directly to local financial institutions, Wise reduces intermediary transfer fees and settlement delays. But even bank integrations come with complications – for instance, Wise can’t control how quickly a partner bank responds to a transfer request – and transfers through partner banks are much more expensive than interfacing directly with a country’s central bank.

So then the next natural step is to bypass local bank partners and access a country’s payment rails directly. This is an even heavier upfront lift than before. Regulators need to be doubly sure that Wise, which is now part of the country’s underlying debit and credit system, has redundant collateral, that its processes and infrastructure are robust enough to meet strict availability requirements and withstand various exception scenarios, like reversing fraudulent cross-border transfers, and will typically require a physical presence on the ground. Basically, all the nitty operational details that a bank partner would otherwise handle on Wise’s behalf, Wise must tend to itself.

But in exchange, Wise gains near-total control over the payment flow and can move money at the same speed and cost as a local bank. In the UK, it was the first non-bank to plug into the Faster Payments Service (FPS), the country’s real-time settlement network, cutting out roughly 90% of partner bank fees and reducing transfer times to under 20 seconds. In Hungary, a direct link to the central bank trimmed costs by 14% and pushed the share of instant transfers (< 20 seconds) from 17% to 82%. Since then, Wise has integrated with central banks in Australia (again, as the first non-bank), Singapore, the Eurozone, and most recently Brazil, where it also secured a second license as a payment institution. So in addition to tapping local rails through its 90+ banking partners, Wise has (or will soon have) direct access in six countries.

Whether through bank partners or direct connections to local rails, Wise relies on domestic infrastructure to drop funds into local bank accounts, raising the question of whether money is really crossing borders at all. On a per-transfer basis, it’s not. Wise maintains local liquidity pools in each country it serves, so what looks like an international transfer is really just a coordinated pair of domestic ones. This setup harkens back to Wise’s founding myth, as recounted in the company’s prospectus:

Kristo Käärmann and Taavet Hinrikus (Wise’s co-founders) were two Estonian professionals living in London. Taavet worked for Skype in Estonia, so was paid in Estonian kroon. At Deloitte in the UK, Kristo was paid in pounds but needed Estonian kroon to pay his mortgage back in Estonia. They devised a simple plan. Each month the pair checked that day’s mid-market rate on Reuters to find a fair exchange rate. Kristo put pounds into Taavet’s UK bank account, and Taavet topped up his friend’s Estonian account. Both got the currency they needed, and neither paid a cent in hidden bank charges.

Today, Wise is that same money-saving arrangement between two friends on a massive scale. When Amber uses Wise to send 300k Rupees to her mom in Sri Lanka, Wise isn’t literally converting Amber’s 1k US dollars to Rupees at that point in time. Instead what’s happening is Amber pays 1k US dollars to Wise’s bank account in the US and Wise pays Amber’s mom 300k Rupees from its bank account in Sri Lanka. No money crosses borders and no currencies are exchanged. One international transfer is really the product of two domestic ones. That is how Wise is able to send money “instantly”.

In theory, the money Amber sends to her mom in Sri Lanka could be offset by an equal amount sent by a Sri Lankan resident to a relative in the US (just as Kristo’s pound payments to Taaet’s UK bank account were exactly matched by Taavet’s kroon payments to Kristo’s Estonia bank account), such that no international transfer fees and FX markups ever need to be paid. Wise benefits from this “netting” effect to some degree. But, in practice, transfers bias heavily in one direction. Money transfers from the US to Sri Lanka will never be entirely offset by transfers from Sri Lanka to the US. And so, over time, the Rupees that Wise has parked in Sri Lanka will be drained to some critical threshold, at which point Wise’s in-house Treasury team will survey account imbalances and FX rates across corridors to identify which currencies it should exchange for Sri Lankan Rupees. In short, money eventually crosses borders, but it does so on an aggregated, rather than a transaction-by-transaction, basis, and after some offsetting has occurred.

After 13 years building relationships with banks and connections to local rails in over 160 countries, Wise is positioned to transfer funds denominated in more than 40 currencies, at an average cost of just ~0.66%^{2 [38]}, with 95% of transfers completed in less than a day and 65% in less than 20 seconds.

Source: fy24 Investor Presentation

(That Wise’s ~0.7% take rate (which is really more like ~0.35% if you exclude the fixed fee component) isn’t much lower than it was 5 years ago does not necessarily reflect a failure to scale, as that figure is a blended average across corridors that vary significantly in cost: mature routes like the UK to the Eurozone may carry fees as low as 0.35–0.40%, while transfers to developing markets with more arcane currencies can approach 1%. It also takes time for volumes to build to a point where Wise can negotiate lower fees with bank partners and scale other fixed costs, so a recently launched corridor will carry higher costs than a mature one even if its terminal take rate is the same.)

The industry for cross-border transfers is rather fragmented. Even a prominent global bank like Citi moved just $358bn across borders in 2023^{3 [39]}, a figure that includes flows from enterprises (~half of global cross-border payments volume in 2020), which Wise doesn’t serve. JPMorgan might be moving around 3x that amount, again including B2B flows. Wise, with ~$164bn of volumes, is among the world’s largest consumer money transmitters for consumers and SMBs and even then its market share barely cracks 5%. The set of competitors also varies a lot by corridor. Flows to developing countries, where take rates can sometimes run in the double-digits, are facilitated not only by banks but by Money Transfer Operators (MTOs) like Western Union and MoneyGram, whose customers are heavily comprised of unbanked immigrants. Reflecting its origin as a cost-saving scheme between two two salaried professionals, one of whom was paying off his mortgage, Wise caters more to the white collar crowd and over-indexes to flows between developed economies, where the per-transaction sums are larger (Wise’s customers transfer 6x-7x more money per year on average than Western Union’s) and rates are closer to low-single digits. Here, banks are a far more significant presence.

It’s easy to understand why ~80% of Wise’s customers come from traditional banks. Besides outcompeting them on speed and price, Wise also offers a far cleaner and more transparent user experience. The process of sending a transfer through the Wise app is as straightforward as it gets. When I lived in Germany some years ago and used Wise for the first time, I was stunned by how easy, cheap, and transparent the experience was. You simply fund a Wise transfer through a domestic bank or card payment and enter the recipient’s foreign account details. After running KYC and fraud checks in the background, Wise will convert currency at a mid-market rate and provide a clear breakdown of the transfer fee, FX rate, and estimated delivery time, with live updates and timestamps at each step of the transaction. This feels like magic to anyone accustomed to sending international wire transfers through bank ^[33]s, who usher users through clunky displays, hide fees through marked up exchange rates, provide no guidance on when funds will reach recipients nor any visibility on which checkpoints have been reached in the transfer chain.

Even if a bank could provide international transfers to consumers for less, would they ever voluntarily do so? Traditional banks are not in habit of passing lower unit costs to customers, absent competitive pressures. They are institutionally geared to hoard savings. Wise does the reverse. It abides by a cost-plus model, where any cost reductions beyond those required to sustain a target margin are recycled into lower take rates for users. This policy is enforced at a granular level, such that no corridor, product, or customer segment is allowed to cross-subsidize another. Wise is fanatical in its pursuit of being the cheapest and fastest way transmit money. Its maxim, “Mission Zero”, encapsulates the continuous erosion of fees up until their ultimate elimination. Even if aspirational in reality (you can’t actually charge 0 fees and turn a profit, right?), the credo is treated internally as though it were literal. Peruse the vast compendium of former employee interviews on Tegus and AlphaSense if you dare. That no, really, everyone at this company does in fact have immediate recall of this central tenant and strives toward its actualization is echoed again and again.

In addition to the analog moat of licenses, bank partnerships, and rail connections – assembled piecemeal over more than a decade – Wise operates scalable cloud infrastructure to manage liquidity and onboard customers efficiently. Machine learning models forecast which liquidity pools are likely to be in surplus or deficit based on expected fund flows, enabling Wise to shift capital across countries ahead of partner bank cutoff times and ensure sufficient balances to support instant transfers. The same technology also accelerates customer onboarding and verification. Manual checks that used to consume thirteen agent hours now only take up two. In fy21 (ending March 2021), Wise automated 91% of document reviews for consumers with greater than 90% accuracy and resolved 80% of cases in just one interaction. One can only assume these metrics have improved since then.

Wise’s cost obsession shows up in its frugality across areas of the business that don’t directly impact the customer experience. As a former Wise employee put it in this Tegus interview ^[40]: “You have a stipend for travel. You can barely buy any food with it. You’re always on overnight flights, it’s something which he generates consistently across every person in the business”. Another elaborates ^[41]: “There’s also a very strong internal culture that everything we spend in marketing is customers’ money….There’s this very strong sense that all the dollars that we spend are customers’ money, and they’re trusting us to make smart decisions with it”.

Granted, it’s a huge cliche to say that traditional banks are stodgy, bloated institutions running on antiquated tech who have lost the script on innovation and customer service. But does anyone doubt this is true? Wise has more than 800 engineers split across ~120 autonomous teams – split by region, then by product, backend service, or feature – who are encouraged to arrive at solutions on their own, fostering a rapid problem-solving cadence. It pushes more than 120 updates a day while a large banks, according to Wise, might release more like one every 3 to 6 months. CEO Kristo Käärmann, who owns 18% of the company and retains the same number of shares today as he did at the time of IPO three years ago, is still very much in founder mode. He is known to work directly with engineers closest to a problem, a scenario impossible to imagine at a large European bank.

I can’t think of a significant digital banking initiative launched by a traditional bank that succeeded on its own two legs, even as several failed efforts come to mind: Finn ^[42] (JPMorgan), Greenhouse ^[43] (Wells Fargo), Simple ^[44] (BBVA), and Bo ^[45] (RBS). Notably, in 2020 Santander launched a cross-border transfer app, PagoFX, that was intended to compete with Wise and could lean on the advantages of a parent with banking licenses South America and Europe. It was ultimately forced to shut down less than two years later. Earlier this year, HSBC launched a competing app, Zing, in the UK. In assessing 0.2% of funds on conversion and no transfer fees, it charges lower fees than Wise, Revolut, and other money transmitter of note in nearly all of the dozen or so corridors I checked. I doubt this is sustainable. There’s almost no way that Zing enjoys lower unit costs than Wise, as it has nowhere near the volumes. And to power its flows, the company relies on a third party, CurrencyCloud (owned by Visa), a euro-centric service provider with no direct connections to any central banks.

Anyway, most traditional banks are simply not set up to facilitate or prioritize consumer and SME money transfers. With maybe the exception of Citi and JPMorgan, global banks are better described as a collection of local banks, with no central treasury management system coordinating flows across them. They offer competitive exchange rates for cross-border flows only if the currencies in question happen to align with those the bank happens to be transacting in that day for its own security-hedging purposes. Also, to minimize risk, banks are reluctant to enter into any more relationships than they absolutely must, hence their continued reliance on SWIFT even as bilateral connections present a faster, lower-cost path for customers.

Adyen had the right idea. The best way forward is to “start over again”. And much as Adyen, through the rapid product innovation enabled by a single unified platform, stole gobs of share from legacy merchant acquirers who were hobbled by a patchwork of poorly integrated acquisitions, so too has Wise, through its streamlined infrastructure of direct bank relationships and user-friendly products, overseen by empowered teams and coded on modern cloud infrastructure, siphoned away cross-country flows from legacy banks who transact through SWIFT and build on ancient tech.

Growing volumes drive lower unit costs through lower bank transfer fees and greater leverage on engineering and product development, which in turn produces more gross profits that, beyond a sustainable margin target, are reinvested in product, marketing, and price, which in turn draws more volume (this is akin to Ryanair using cheaper tickets to attract more passengers, who are then leveraged to negotiate lower landing fees and aircraft prices, with the resulting savings then feeding back into cheaper tickets). Enchanted by low fees and fast transfer speeds, customers promote the service to friends and family – around 2/3 of new customers come through word-of-mouth – reducing the Wise’s reliance on paid marketing to drive growth. In its most recent fiscal year (ending March 2024), Wise spent just £37mn of marketing against £119bn of transfer volume while LTM through June ‘24, Remitly spent $282mn against $46bn).

The flywheel looks something like this:

Source: fy24 presentation

But to what extent can Wise leverage its low cost advantage in cross-border transfers to other product domains? And what are the soft spots in its business model? Stay tuned for part 2!

Wise and the business of cross-border transfers: part 2 ^[34]

Disclosure: At the time this report was published, accounts I manage owned shares of $V, which owns CurrencyCloud. This may have changed at any time since

(taking a break from building products after this. Up next: a 2-parter on Wise) In part 3a, I provided an overview of pre-fabricated trusses, namely the value they offer homebuilders over traditional stick building and what’s involved in the transition from Phase 1 manufacturing (automated saws and lasers) to Phase 2 (robotic systems). You’ll want […]