Reading Mark Leonard's Shareholder Letters: Part 2 (2013–2021)
Part 1 covered what Leonard looks for in a business. This part is about how he runs one, drawn from the letters between 2013 and 2021, which became less frequent and more concise over that period.
Three things run through them. He measures the company with a number he devised, watches it decline exactly as he predicted, and then tells shareholders the number has stopped being useful. He refuses to centralise a company of 200-odd businesses, while admitting his own data does not prove he is right. And he spends several pages on what makes a good business unit manager and a good director, which reads oddly in a shareholder letter until you see that those people are what holds the structure together.
1. What the numbers taught me
Return on invested capital is the profit a business earns divided by the capital tied up in it. Leonard’s version has two parts.
The profit figure is Adjusted Net Income. It is reported profit with non-cash items added back, mainly the amortisation of intangibles. When Constellation buys a software company, accounting rules require it to write down the value of the acquired customers and technology over a few years, even though no money leaves the business. Adding it back assumes those assets are not losing value.
A lot rests on that assumption, and he says so. He warns that acquisitive companies can grow revenue and adjusted earnings while impairing the assets they bought, so that what looks like a return on capital is really a return of capital. This is why he publishes the maintenance revenue table every year. Between 2012 and 2016, price increases added 5 to 7 percentage points of growth each year, while customers lost took away 4 to 5. Customers stay and accept paying more, so the base is growing on its own.
The capital figure is Average Invested Capital, and it is not taken from the balance sheet. It began as an estimate of what shareholders had put into the company, and has been kept as a running tally since: add each year’s adjusted profit, subtract dividends, add share issuances. It measures what shareholders have left in the business.
He then adds organic growth to the return and calls the sum the Combined Ratio. I could not follow this at first, because profit already includes growth from every source. What I was missing is that ROIC is a level rather than a rate of compounding. It tells you what the capital earns this year, not that the same capital will earn more next year without anyone adding to it.
Take ₹100 of capital earning ₹30, with all profit reinvested at the same return. If the existing business stays flat, the next year brings ₹30 from the old capital and ₹9 from the ₹30 just deployed, so profit grows 30%, which is the same as the ROIC. If the existing business instead grows 5% without needing new capital, it contributes ₹31.5 rather than ₹30, and profit grows 35%. Acquisitions do not need a separate term because they bring their own capital with them.
Leonard sets two conditions for the sum to mean anything. The company has to keep investing nearly all its cash at similar economics, and the organic growth has to genuinely cost nothing. In a capital-hungry business, growth needs investment first, so adding the two counts the same rupee twice. He describes the Combined Ratio as a reasonable but somewhat overstated proxy for growth in intrinsic value, and says that higher acquisition multiples or significant debt would make it misleading.
ROIC + Organic Net Revenue Growth, continued from Part 1
| Year | Adjusted Net Income (US$M) | Average Invested Capital (US$M) | ROIC | Organic Net Revenue Growth | ROIC + Organic Growth |
|---|---|---|---|---|---|
| 2012 | 172 | 491 | 35% | 2% | 37% |
| 2013 | 207 | 585 | 35% | 4% | 39% |
| 2014 | 274 | 739 | 37% | 3% | 40% |
| 2015 | 371 | 965 | 38% | -3% | 35% |
| 2016 | 395 | 1,261 | 31% | 1% | 32% |
| 2017 | 463 | 1,622 | 29% | 4% | 33% |
The 2015 organic figure was pulled down by currency movements, which cost 6 percentage points of maintenance growth that year.
Maintenance Revenue, continued from Part 1
| Maintenance Revenue movement | 2012 | 2013 | 2014 | 2015 | 2016 |
|---|---|---|---|---|---|
| Maintenance Revenue (US$M) | 510 | 725 | 1,015 | 1,170 | 1,400 |
| Growth from | |||||
| ↳ Acquisitions | 15% | 34% | 32% | 15% | 16% |
| Organic sources | |||||
| ↳ New maintenance | 8% | 10% | 10% | 8% | 9% |
| ↳ Price increases | 5% | 6% | 7% | 5% | 5% |
| Attrition | |||||
| ↳ Lost modules | -2% | -2% | -4% | -2% | -4% |
| ↳ Lost customers | -4% | -5% | -5% | -5% | -5% |
| Total Organic Growth | 8% | 8% | 8% | 7% | 5% |
| Estimated effect of FX | -1% | -1% | -1% | -6% | -2% |
| Total Maintenance Growth | 22% | 42% | 40% | 15% | 20% |
The later letters separate currency effects into their own row, which Part 1’s table did not carry, and restate some earlier years. Certain totals do not reconcile exactly due to rounding, as Leonard notes each year.
Between 2012 and 2017, profit grew about 22% a year while invested capital grew about 27%. When capital compounds faster than profit, the return falls, and that is what the last three years of the first table show.
He had already told shareholders it would happen. In the 2015 letter he published a study of other high-performing conglomerates, including Danaher, Dover, Roper and Jack Henry. Their returns on total capital began around 21%, improved to a peak of 46% after about six years, and fell to 18% by 2015 as their acquisitions became larger and more expensive. He wrote that Constellation’s own 37% return that year was probably close to its peak, and that acquisitions would push it lower. In the same letter he said bigger acquisitions and lower hurdle rates on large deals would likely bring ROIC down as well.
Both came true. In 2016 he explained why: cash he was unable to deploy sat in the denominator earning nothing, and several larger acquisitions had been bought at prices that returned less. By 2017 he wrote that the Combined Ratio was no longer the best single measure of the company, and that he was moving towards growth in free cash flow per share instead.
Two cautions if you plan to use these figures. They are the company’s own definitions rather than audited measures, so they are not comparable with another company’s reported ROIC. And earlier years were restated when the definitions changed, which is why some of these numbers differ from the tables in Part 1.
2. On decentralisation
Mark often fielded questions regarding centralization of the organization for efficiency. While he understood the merits of it, the culture of trust and ownership he had built from the business unit level upwards would not be a good fit with a top-down approach of running the company. He preferred the speed with which lean teams could experiment and bring in new initiatives, and an open meritocratic culture.
He drew upon his experiences of operating in venture capital and also the fact that most vertical markets have competitors that punch above their weight. Best business practices also flow organically in an organization that rewards curiosity at the ground level.
He treats this as a preference rather than a proven doctrine, and says some of his own group managers were less convinced. He also notes that only one other conglomerate has tried buying hundreds of small businesses and running them autonomously, and it eventually gave in to centralisation. His view is that the model needs an unusually trusting culture and a long horizon. If trust weakens, bureaucracy takes over, and if short-term results matter most, the promise of synergies becomes hard to resist.
I found his approach interesting in terms of its applicability to the problem of centralization vs decentralization in other sectors and business models as well. For example, in a previous article I wrote about the clinic model for different fields of medicine. There is a need for a certain level of centralization of processes in terms of SOPs, data management etc., but at the end of the day each clinic is competing with other organized and unorganized set-ups in that particular locality, and should have the autonomy to operate and adapt according to the local environment.
I think this also extends to devising a structure of ownership, trust in the organization and opportunity for growth that promotes this ownership mindset from the ground up, which should be ideated on further by founders.
3. People and small teams
Mark found it easier to foster the aforementioned culture in a team structure that is less bureaucratic, aided by small teams that trusted each other and the vision of the organization.
His picture of the ideal business unit is specific. Five or six functional managers, each leading a working group of five to nine people. At that size everyone knows everyone, rules are few, and the focus stays on growing the business rather than dividing it.
The strain usually begins at 30 to 40 people. Teams outgrow their ideal size and get split, a layer of middle managers appears, and attention shifts from customers towards internal coordination. Past 100 employees, he worried that even an excellent manager who knows the vertical well would struggle to grow faster than the industry. At the end of 2016, 158 of Constellation’s 193 business units had fewer than 100 employees.
The data does not settle the question, and he says so. Small units in his portfolio were no more likely than large ones to grow well or be profitable. Units under 100 employees looked slightly better, but excluding a couple of outliers reversed the finding. He said he was going on conviction until better evidence arrived. Within the company the idea is not universally accepted either. Some groups split large units into smaller ones, while another buys several businesses in the same industry and runs them separately.
What he looks for in people follows from this. He wants employees who work hard, treat others well, keep learning and share what works, and he names who should not last: sycophants, spin-doctors and mercenaries. Harder to remove, he says, are managers who run their businesses on habit and folklore rather than asking questions and testing things.
For business unit managers he looks for several years of experience in the vertical. When Constellation buys a company, the manager almost always comes from that company, because the knowledge sits there. In 2016 it made 40 acquisitions, of which 30 came from owner-managers.
Keeping these people gets as much attention as finding them. Compensation is tied to long-term outcomes. By the end of 2015 there were over 100 employee shareholder millionaires, and he hoped for five times that number within a decade. New business unit managers get an additional bonus programme, simply to keep them around long enough for the compounding to become visible to them.
A Craftsman runs one business unit well. A Player/Coach buys a second business while still running the first. A Portfolio Manager gives up daily management and advises several. A Compounder keeps finding acquisitions and deploys the cash his portfolio produces. Asked what he would tell his own 35 or 40-year-old self, he said stay a Craftsman: work closely with the best customers in your vertical, make small tuck-in acquisitions, and grow revenue per employee. He calls it the most satisfying job in the company.
4. On boards
Leonard separates two roles a director can play. Governing means monitoring, and he treats it as table stakes. Coaching means knowing the industry and the company well enough to actually help.
He works through what coaching would take. An outside director spends about 30 hours a year in board meetings, perhaps double that with preparation, and up to 200 hours if deeply engaged. At 200 hours a year, 10,000 hours of deliberate practice takes 50 years. So a director who does not come from the industry has little hope of coaching unless they join in their 40s or 50s and stay for decades.
This is why he dislikes term limits. Where governance is already good, he compares limiting a director’s tenure to firing a high performer on their tenth anniversary.
He also sets out why good candidates say no. Only two reasons seem to him compelling enough to justify hundreds of hours a year: having a meaningful part of your wealth in the company and wanting to watch it closely, or wanting to learn something you can use elsewhere. He then points out that he cannot forecast intrinsic value growth above 12% a year, which is not enough to attract capable capital allocators, and that very few people want to study a decentralised model this closely.
5. Closing thoughts
Whether a similar company could be built in India is not something these letters can answer, but they do suggest what would have to be true.
Constellation’s buyers are mostly small organisations, and a large share of them are governments or government-related bodies. When the company last disclosed the split, in 2019, close to half its revenue came from such customers. These buyers hold their software for years and pay for it annually, with the price rising a little each time. The annual increase is a meaningful part of the growth in its maintenance base, and without it the businesses would not be worth the prices Constellation pays.
In India, the question is whether that kind of buyer exists in enough numbers, and whether software is bought in a way that allows a vendor to hold on to that pricing power over decades. How government software gets procured here, and how that differs from North America, is a subject in itself and deserves a separate piece rather than a paragraph.
Reading these letters has given me a clearer view of three things. The value of measuring what actually matters, and of being intellectually honest about it for the sake of everyone involved. The value of trust, and how it compounds when a company’s structure is designed carefully to promote what it says it values. And a better sense of the questions I should be asking when I look at a business.
Source: Constellation Software shareholder letters and annual reports
https://www.csisoftware.com/investor-relations/