I was going to write up Topicus but realized that it wouldn’t work without writing up Constellation Software first. So I changed the write-up to Constellation and will write up one on Topicus next. (Read it here).
I recently read Michael Lewis’ book, The New New Thing: A Silicon Valley Story and a thing that stuck with me was his reference to the phrase “business model” as “a term of art”. That’s what Constellation is: a piece of art on Mark Leonard’s canvas. His annual shareholder letters (which started as quarterly after Constellation listed in 2006, turned into an annual occurrence in 2009, and now arrive whenever Leonard has something to say) possess humility, competency, and a nail-to-the-head type of writing that makes understanding the business the easiest thing in the world. Many investors are better off reading his letters back to 2006 to arrive at a fair assessment of the business rather than reading my, or any other’s, explanation of it. Find them here.
An example of his humility was on display in the 2015-2016 letters in which he set out to study other “high-performance conglomerates” (HPCs) such as Constellation, even as his company by many measures was the most successful of the rare bundle of HPCs. Over the past decade up until 2015 when Mark did the study, Constellation had compounded revenue/sh and CFO/sh by 24% and 30%, respectively. The share price had gone from CAD24/sh to CAD510/sh, a 36% CAGR over the 10- year period.
The study found patterns in the HPC lifecycle of three main phases: 1) the HPC starts out earning extraordinary returns through operations in asset-light businesses, 2) then embarks on attractively-priced acquisitions to which it applies its refined operating practices learned from phase one to increase ROI, and 3) it then drifts towards paying higher multiples for larger acquisitions as size became a constraint and declining ROIs across the conglomerate turns inevitable.
Constellation hasn’t shown reversion to the mean in a meaningful way yet but likely found itself transitioned to the third phase recently for reasons I’ll discuss.
Throughout its history, Constellation has earned the HPC badge largely from intelligent capital allocation in VMS businesses rather than through operational excellence. Not that Constellation’s VMS operations haven’t been excellent—obviously, that goes without saying—but that’s largely been a result of earning a high return on existing capital rather than reinvesting new. By existing capital, I mean essentially no capital.
Due to an advantageous cash conversion cycle and minuscule fixed assets, Constellation has historically been a negative-tangible-net-assets business. Goodwill and intangibles comprise most of the balance sheet, but these are amortized every year, even as the assets don’t lose value. In such a business ROIC can fly into the sky, and other times, as in the case of negative invested capital, into infinity. So you need another measure for capital invested in the business. Since debt has historically been absent from the capital structure (as have buybacks) Constellation has always used share capital + retained earnings + accumulated amortization as its definition of invested capital, which makes sense since the business is capital-light and has consistently reinvested more than its FCF.
Adding the return (net income + amortization) made on this capital to organic growth has been its historical proxy for intrinsic value growth to shareholders. While it’s been remarkable, you can see the evidence of how the factor propelling intrinsic value forward lies in excellent capital allocation rather than organic growth:
The historical ROIC+OGr performance shows that as consistent Constellation’s excellent returns on capital have been, equally consistent have been the disappointments in organic growth. It’s not that Constellation is striving for high growth at the BU level anyway. It used to do it, but that was until it learned more about the business it was in. Constellation used to target annual organic growth in the 5-10% range until 2009 when the goal was adjusted down to a little above the growth in GNP, or msd.
The organic growth constraint inherent in Constellation’s VMS businesses originates from the same source that makes up their moat. VMS is far from your hot software business. Where the mantra in horizontal software—AI, data storage, cybersecurity—is bigger and more powerful, the VMS mantra is smaller and iterative. Whereas horizontal software companies are trying to prove a product to a huge TAM with long runway, VMS businesses are already proven in their own, oft-saturated niche. Where horizontal software companies try to create something like full social networks, VMS businesses make something like the underlying software system of a local fire station. VMS businesses are not disruptive but they are not subject to disruption either.
You can imagine the moaty, once-monopoly-like local newspaper business that Buffett and Munger were into. The similarities are numerous: scarce competition, differentiation, high gross margins on incremental sales, minimal churn from high switching costs (financial for Constellation’s customers; informational for the newspaper customer).
The latter, high switching costs, come from iterative feedback loops with VMS customers. To the customers, the highly customized software delivered by a Constellation VMS is mission-critical and a need–to-have. Constellation’s businesses help their customers run their businesses efficiently, adopt industry best practices, and adapt to changing times. The longer the VMS works with the customers to fine-tune their needs, the wider the moat becomes. For new entrants to come close to competing with that, they’d have to go through the trial and error themselves. But even so, the annual cost rarely exceeds 1%/customer’s revenue so a customer wouldn’t even bother trying a new vendor unless they can provide something revolutionary. As a result, Constellation consistently retains +90% customers across all VMSs.
So rather than expecting a sudden uptick in organic growth, the more pressing question is whether the existing VMS portfolio will continue to generate a predictable level of cash flows for decades into the future which is needed to continue running Constellation’s M&A machine. This is where the similarity to the local newspaper business halts. Ironically, the one factor that crept in to disrupt the declining newspaper business was…technology. Constellation is not prone to technology disruption to the same degree. Rather, it’s in a technological sweet spot as evidenced by its consistently high returns and low churn. Constellation is one of the few technology businesses in which you would feel safe adding back all amortization expenses every year to adjust net income and feel content with that since you know that the economic value of the goodwill and intangibles remain.
To elaborate, Mark has many times raised the importance of looking at the company’s organic maintenance revenue growth as an indicator of whether the VMS businesses lose intangible value. As long as maintenance revenue (recurring revenue primarily consisting of fees charged for customer support on software products post-delivery) grows organically, you can make the case that the value of the company’s intangible assets hasn’t declined but instead appreciated. Up until 2016, the decomposition of maintenance was broken out into the following, but since SaaS revenue has begun to comprise a larger share of maintenance (as seen below starting in 2013), the company stopped showing the decomposition, likely due to historical incomparability (SaaS revenue involves higher churn, but it doesn’t necessarily mean it’s a worse business).
The flip side of many software companies’ capital-light operations is that they need to spend a commensurate amount on R&D and S&M to stay much in the same place. At Constellation, this doesn’t weigh to the same degree:
Our favorite businesses are those that are growing just slightly faster than their markets, gradually adding market share and customer share (i.e. “share of wallet”), while generating a good return on the capital that they have invested to produce organic growth. Small market share gains are much less likely to trigger a scorched earth competitive response that erodes pricing and triggers wildly unproductive R&D and S&M binges. We believe that we have struck that balance at many of our businesses.
R&D spend at BUs is largely attributable to growth investments, not maintenance. Constellation can afford to slow down discretionary costs and investments when few Initiatives (significant long-term investments required to create new products, enter new markets, etc.) have commenced, increasing FCF. That it did after 2005 when the company found Initiatives to account for >50%/total RDSM expenditures until it was rationalized and new Initiatives plummeted. Constellation kept the early burn rate of Initiatives down until it had proof of concept, sometimes getting clients to pay for the early development.
These points are all you need to know about the VMS businesses. These are the factors that allow Constellation to generate nice FCF from its VMS portfolio on miniscule tangible assets which it uses to pursue growth by M&A. The majority of this M&A responsibility lies with the company’s six Operating Groups, each responsible for >200 BUs in different industries operating across a wide range of competitive environments.
The BUs within the Operating Groups share accounting, acquisition, IT functions, and varying degrees of HR, tax, R&D, legal aspects, as well as senior staff who’ll be parachuted into large new acquisitions or turnarounds. Also within these Groups sit Portfolio Managers who are usually promoted VMS managers who have shown excellence in operations and then in capital allocation and are now responsible for a concentrated number of 2-12 BUs. Together with another double-digit number of M&A specialists spread across the organization, these Portfolio Managers spend >50%/their time on M&A and the rest advising their BUs. If an excellent Portfolio Manager proves themselves through multiple successful deals, they might be given the “Compounder” designation.
Below the Operating Groups sit the BUs where each usually contains a single VMS but sometimes a few. The BUs are treated as the designation implies: they’re separate entities. They possess full operational autonomy, are held accountable for their results, don’t cross-sell products, and they sometimes even compete with one another for the same customer.
The Manager (or “Craftsman”) of the BU is nearly always from the acquisition itself, having expertise in the vertical. Once in Constellation’s fold, the BUs are kept deliberately small to remain agile and prevent overhead creep (which is a function of Mark’s experience with small high-performance teams from his time as a venture capitalist, akin to the “day 1” mentality embraced by Jeff Bezos). Because the BU managers know this, they are incentivized to groom other functional managers beneath them, enthused about running and growing a spun-off BU so that they increase their track record as leaders and capital allocators within the organization. As the BU Manager’s ambition drives them to acquire another VMS, their role is usually changed to Player/Coach. And if things go well, and the Player/Coach finds a second or third VMS to acquire, they eventually have to give up the day-to-day responsibilities and become a full-time Portfolio Manager by moving up in the Operating Group level. Once a Portfolio Manager, the Manager’s job changes to advising aside from capital allocation. They fly around to their BUs, gather information, share ideas and problems with others within Constellation, fix the problems with the BU Manager, and then move on to the next company. If the new BU Manager disappoints, it’s up to the Portfolio Manager to find a replacement.
It’s here the difference lies between Constellation Software and the biggest HPC in the world, Berkshire Hathaway. Like Berkshire, Constellation’s organizational structure is decentralized (out of a total of ~25k employees, only 10 sit at HQ in Toronto, a ratio of 1:2.5k). But while Berkshire’s capital allocation responsibilities are solely held at HQ with Warren Buffett and, to a lesser extend, the two portfolio managers, this job is spread widely throughout Constellation.
Leonard was Constellation’s first Portfolio Manager but he already ran out of capacity around mid-2005 when the company was at $200mn revenue and ran ~30 BUs. So when he realized he couldn’t keep up tabs at the BU level andno longer could be the primary driver of M&A, Leonard started delegating responsibility to the Operating Group Managers for monitoring and coaching their BUs and deploying the majority of the FCF. You’d think the dilution of capital allocation responsibility would also dilute returns, but since then, ROIC has only expanded. Today, each of the Operating Groups is equivalent to how Constellation looked a decade ago and each Operating Group Manager is continuously delegating their monitoring, coaching, and acquisition activities down to their Portfolio Managers, expanding the organization and resuming the cycle.
Many would agree that this system wouldn’t work at Berkshire, so why has it worked at Constellation? It boils down to Constellation’s well-designed incentive system that works in a niche system but finds trouble if it expands out from the niche.
First, the incentive system pervades from the top instantiating a mentality of ownership down the organization. Bonuses depend on two variables: 1) ROIC and 2) net revenue growth, striking a good balance of how responsibly the Managers deploy capital. If Managers hold excess cash, they’re incentivized to send cash to HQ to lower the denominator in the ROIC equation. If they recklessly deploy FCF into bad Initiatives or acquisitions in an attempt to boost revenue growth, ROIC will suffer, and so will their bonuses. Because ROIC is so central to compensation, Constellation has struggled with making their Managers keep their profits to keep ROIC from running too wild. Second, instilling a long-term ownership mentality throughout the organization, Operating Group managers are required to earmark 3/4 of their tax-effected bonuses to purchase Constellation stock in the market, while all employees above a certain level of compensation must also purchase a certain amount of stock and it for >4 years.
That Constellation can trust so many in allocating its capital is a feat. It’s possible because every VMS acquisition is alike and every deal is pursued based on accumulated data from Constellation’s historical deals that allow capital allocation to remain rational, not emotional. Constellation has distilled its acquisition strategy down to systematic processes that can be learned by others. Almost no Manager within Constellation had done an acquisition before joining the ranks.
The incentive system also creates organizational career opportunities. The journey from Craftsman to Compounder is, of course, financially rewarding, but it’s also a realistic feat for any employee at the low ranks in the meritocratic system. The system works because employees are willing to stick around due to a high degree of trust that their performance and KPIs will be noticed and awarded accordingly.
(In Leonard’s HPC study, he found that only one other HPC has followed a strategy of buying 100s of small businesses and managing them autonomously. They eventually caved into increased centralization. Mark suspected the missing piece was a seamless web of trust in the system. ”If trust falters the BUs can be choked by bureaucracy.”)
Constellation looks for two types of businesses. Its favorite is when it buys directly from the founder who’s spent the better part of a lifetime building the business, instilling a long-term orientation in the enterprise. Constellation appeals to such founders because it offers something other potential buyers, such as VCs, don’t: a long-term safe haven that cherishes its autonomy. Constellation also dabbles in distressed assets like, for instance, a failed synergy experiment from a large corporation that has scooped up the VMS to integrate it only to divest it again. Such deals have been the most lucrative for Constellation during recessions.
For any large acquisition that the company pursues, the diligence, structuring, negotiating, and integration are led by a single Portfolio Manager, routinely a “Compounder”, who shoulders the responsibility for the process and the outcome. This Portfolio Manager and their team follow the hurdle rates set by HQ. This hurdle is what matters since Constellation is agnostic about how ROI is generated. It buys shrinking businesses if it meets the hurdle (even as mixing growing and contracting businesses in one conglomerate creates interesting cultural challenges). More often than not, non- or slow-growing businesses are the only kinds the Portfolio Managers can find that meet the hurdle. Even in an environment where throw themselves over all things tech, those businesses are deemed boring. Constellation will sometimes be the only buyer in its market corner, and promising a safe haven puts it in an attractive position to bid for lower prices. Akin to Berkshire, Constellation avoids bidding wars, offering one price to be accepted or not.
The key variable for Constellation’s growth story is the price paid for acquisitions. Taking each year’s revenue growth from acquisitions and the deal spend to assess the average implied P/S multiple paid in the given year, you can back into a historical <1 EV/S multiple paid.
But as mentioned at the beginning, Constellation has likely entered the third phase of declining returns as an HPC. For just the past four quarters, the company has generated almost $1bn in FCF, up almost 2x in 2 years. To sustain its growth by acquisition to the same degree, in the same type of businesses it historically has acquired, and at the same multiples, Constellation would need to swallow north of 140 companies next year and/or pay up higher multiples for bigger whales. Yes, for the next maybe 2-3 years, it’s going to be hard, but doable (in the 2021 letter, Leonard assured it will continue to invest most of its FCF in small and mid-sized VMS acquisitions at the traditional hurdle rate), but just a few years out, it’s going to become a challenge to find such an amount of good deals at <1x sales in VMS let alone to train or find a large number of Portfolio Managers as competent as the current Operating Group Managers (Topicus.com, a large acquisition, was bought at 1.8x sales). The question is not if but how slowly the return will revert.
For years, Constellation tracked large VMS deal prospects as a separate segment of the market. Historically, only 10%/FCF has gone to this segment with three large deals undertaken throughout its history (out of between 40 and 70 large VMS businesses that are sold each year). As Constellation becomes forced to compete in this market and move up the acquisition size ladder for VMSs, the peers become private equity funds more than anything. Competing with large PEs surely leads to paying up for quality and getting into the occasional bidding war. Keeping the traditional hurdle rate is going to dilute the prospects to potentially nothing, so it has to come down.
On the other hand, Leonard may not care whether Constellation invests in VMS, any other type of software, or software at all. Constellation just happened to stumble upon a decades-long value opportunity in the moaty VMS business and put all resources and efforts into becoming the best in that corner. The flip side is that it spent all its time crafting a narrow circle of competence and it’s not going to be easy to step outside that circle. You’ll hear more about this in the coming years’ reports and that HQ intends to do something about it on an experimental level. Constellation might pick a single industry with similar characteristics to VMS and spend a few years dabbling in that, seeing which investment processes already learned fit the given industry. For investors, that’ll require patience.
Meanwhile, investors will likely see returns of capital sooner rather than later. In his 2017 letter, Leonard wrote that he’d advise investors to look at the per share growth in FCF available to shareholders (FCFA2S) rather than the combined ratio of ROIC+OGr going forward as a proxy for growth in intrinsic value. Note that a measure like ROIC+OGr is only valuable as long as the business remains capital-light, the economics of the acquisitions are similar to those in the past, and the company can continue to reinvest all its FCF. So in changing it for growth in FCFA2S, he seemed to suggest either or all of the following: 1) that Constellation will not be able to reinvest all FCF generated going forward, 2) that returns will slow down so that past returns on capital mislead the expected future returns, 3) that HQ will see buybacks as a best-case alternative to acquisitions when the hurdle is lowered and if the right prices present themselves, and/or 4) that dividends become a regular thing due to the aforementioned points.
In 2013, Constellation paid a bunch of money to a group of consultants to get some estimates on the value of Constellation. At the time, Mark found the stock fairly priced, but the most noticeable conclusion about the sensitivity was this:
Varying the organic growth assumption has a tremendous impact on the intrinsic value of a CSI share. Add in another 2.5% organic growth to the baseline assumption and you get more than double the intrinsic value. Subtract 2.5% from the baseline organic growth assumption and you lose almost half the intrinsic value of the stock. You can see why so many software company CEO’s are growth junkies.
By the time of the report, the stock was priced at 20x FCFA2S. It was also a time when organic growth was higher and you had little to worry about in terms of acquisition runway. Today, at a CAD48bn market cap, where organic growth is stagnating and M&A opportunities likely look to contract, the price is at almost 40x TTM FCFA2S. I don’t know whether that’s cheap or expensive. The conclusion I can make is that there’s a capital allocation premium baked into the price—a premium which may or may not be justified depending on whether you believe either that organic growth will pick up to a significant degree (unlikely), that the company can continue vacuuming up many hundreds of small VMSs at <1x sales over the next decade (also unlikely), and/or that Constellation can expand its M&A machine outside of VMS to a significant degree (perhaps likely). My conservative nature keeps me from making such assumptions.
But I may turn out wrong. Mark has always been modest in his statements regarding the future of Constellation which is a testament to the fact that even one of the best capital allocators in the world is subject to a probability distributional future, not sure things. But Constellation has repeatedly surprised about its ability to scale a strategy of VMS rollup, managing them autonomously through a trusting culture. The most impressive feat, and the most overlooked part of the story, is how HQ has succeeded in distilling not only operational processes but also investment processes down to systematic steps to be used by the Operating Groups and every one of its Portfolio Managers. Perhaps this time isn’t different and Constellation has something else up the sleeve. At least, if you’re an investor, that’s what you pay for at the current price.