Constellation Software: capital allocation is the product.

The thesis

Most companies are built to run an operation. Constellation is built to allocate capital. It buys small vertical-market software businesses (unglamorous, mission-critical tools embedded in narrow industries), takes the cash they reliably produce, and redeploys it into more of the same, only ever at a high required rate of return. The operating businesses are good. The machine that decides where their cash goes is the actual asset, and it is what compounds.

What they buy, and why it works

The targets are deliberately boring: software that a marina, a bus company, or a municipal department runs its whole operation on. Each one is too small for a large acquirer to bother with, deeply embedded in its customer's workflow, and almost never churned. Individually they are modest. Bought at the right price and held forever, a portfolio of hundreds of them throws off a large, durable, growing stream of cash, exactly the raw material an allocation machine needs.

Crucially, Constellation rarely sells. Selling forces you to find a new home for the proceeds and pay tax on the way; holding lets the cash keep compounding inside the structure. The model is closer to a permanent-capital holding company than to private equity.

The revenue engine

Two engines run at once. The operating engine is the maintenance and subscription revenue of hundreds of sticky software businesses: high-retention, low-drama, and only modestly growing on its own. The capital engine is acquisitions: that operating cash, plus modest leverage, redeployed into new businesses at a disciplined return. The second engine is the one that drives the result. Organic growth is steady; acquired growth, compounded over decades at a high return, is what produced the track record.

The discipline that makes it work

Founder Mark Leonard wrote a shareholder letter most years from the mid-2000s until 2018. They are among the clearest writing on capital allocation in business, and the recurring theme is a single rule: deploy capital only above a defined hurdle rate, regardless of the pressure to do otherwise.

A hurdle rate is the minimum return a deal must clear to receive the cash. Set it high and hold it, and two things follow. You walk away from a lot of tempting, mediocre deals, and the deals you do make compound hard, because you only ever bought at a strong return. It sounds simple. Holding it is the entire difficulty: there is always cash burning a hole, always a "strategic" deal, always a quarter where the bar feels inconvenient. Disciplined allocators become average ones not through one disaster but through a hundred small relaxations.

The moat and enterprise-value drivers

• Decentralised operations, centralised allocation. Each business runs itself; the capital decision is made centrally and held to the standard. This lets the company scale to hundreds of units without the centre becoming a bottleneck.
• A reputation as a permanent home. Founders who want their software business cared for, not stripped, sell to Constellation specifically because it doesn't flip. That reputation is a sourcing advantage no new entrant can buy quickly.
• Compounding itself. Decades of disciplined redeployment is a moat of arithmetic: hard to replicate because it requires both the rule and the patience to hold it through cycles.

What would break the model

• Scale vs. the bar. The bigger the cash pile, the harder it is to find enough deals that clear a high hurdle. The central tension of the model is deploying ever-larger sums without lowering the standard.
• Competition for small VMS. More acquirers chasing the same businesses bids up prices and compresses returns, directly attacking the hurdle.
• Discipline drift. The whole edge is the refusal. Any erosion of that discipline, even gradually, quietly turns a great allocator into an ordinary one.

What I'd copy, what I'd avoid

Copy

• Treat "where does the next dollar go?" as the main job. Past a certain size, allocation matters more than operations. A merely-good operator who allocates brilliantly beats a brilliant operator who reinvests carelessly.
• Write the bar down and defend it. A return threshold you'll abandon for a tempting deal is a wish, not a threshold. Commit it to paper; give it to someone whose job is to challenge you when you want to relax it.
• Centralise capital, decentralise operating. Free your operators; keep a tight, consistent hand on where the money compounds. Most companies do the reverse.

Avoid

• Copying it onto a business that can't run itself. The model only works because the units mostly self-operate. If yours needs constant attention from the centre, you can't allocate your way out of an operating problem.
• Confusing activity with discipline. Doing deals is easy. Doing only the deals that clear the bar, and sitting on cash otherwise, is the hard, valuable part.

On the figures: the ~35% annualised total shareholder return since the 2006 IPO, the founding date, and the IPO year come from public market data and company filings, treated here as directional rather than to-the-decimal. The characterisation of the hurdle-rate discipline draws on Mark Leonard's published shareholder letters; the exact threshold is not something I'm quoting a precise figure for. Verified: the 1995 founding by Mark Leonard, the 2006 TSX listing, the vertical-market-software acquisition model, and the buy-and-hold (rarely-sell) approach.