How to Analyze a Software Company

What makes a software company different

A software company differs from most businesses in one decisive way: the cost of making a second copy of its product is close to zero. Building the software is expensive, but selling another subscription costs almost nothing, so once the product exists, additional revenue drops through at very high margins. That single fact shapes every number worth reading on a software income statement, and it stands in sharp contrast to the physical businesses covered elsewhere in this series, such as the factories and inventory in how to analyze a semiconductor company, where every extra unit carries a real manufacturing cost.

It also changes what you are buying. A well-run software business sells a subscription that renews, not a product that has to be re-sold from scratch each year. The revenue arrives whether or not the sales team wins a new logo, which makes the future far easier to forecast than for a company that starts every year at zero. This is the recurring-revenue model examined in recurring revenue business models, and it is the reason software commands the valuations it does.

Microsoft is a useful anchor because it shows the model at full scale. In fiscal 2025 it reported revenue of about $281.7 billion and net income near $101.8 billion, a net margin of roughly 36 percent, according to its 10-K. Its largest reporting line, server products and cloud services, brought in about $98.4 billion, and its Microsoft 365 commercial cloud added another $87.8 billion. Most of that is subscription or consumption revenue that shows up again next year. The job of analysis is to judge how durable that repeating stream is, and what it truly costs to run.

The metrics you need to analyze a software company

Four measures tell you most of what you need to know about a software business, and together they are how you analyze a software company past the marketing: the share of revenue that recurs, net revenue retention, gross margin, and free cash flow counted honestly. Growth matters too, but only once you know the revenue is high quality. The reason these four carry so much weight is that they describe the durability of the revenue, not just its size, and durability is what a subscription model is supposed to deliver.

Recurring revenue is the portion that arrives on a subscription or contract rather than as a one-time sale. The higher the share, the more predictable the business. Companies report this in different ways, so look for annual recurring revenue (ARR) or remaining performance obligations, the contracted revenue not yet recognized. Microsoft's commercial remaining performance obligations and its large deferred revenue balance, about $64.6 billion of current deferred revenue at the end of fiscal 2025, are a measure of sales already booked and waiting to be earned.

Net revenue retention (NRR) is the single most revealing software metric. It asks how much a cohort of existing customers spends this year compared with last, after upgrades, downgrades, and churn.

Net revenue retention = revenue this year from last year's customers / revenue last year from those same customers

An NRR above 100 percent means the company grows even if it never signs a new customer, because the ones it has keep spending more. Best-in-class enterprise software often reports 110 to 130 percent. An NRR below 100 percent means the base is leaking, and new sales are being used to fill a bucket with a hole in it.

Gross margin shows the structural economics. Software gross margins usually sit between 70 and 85 percent, and the mechanics behind that figure are covered in gross margin. A margin well below that range suggests heavy cloud-hosting costs or a services-heavy mix that scales less gracefully than pure software.

The Rule of 40, read skeptically

The Rule of 40 is a popular shortcut that says a healthy software company's revenue growth rate plus its profit margin should add up to at least 40 percent. A company growing 30 percent with a 15 percent margin scores 45 and passes; one growing 15 percent with a zero margin scores 15 and fails. It is a quick way to sort a long list, and that is all it is.

Treat it skeptically for three reasons. First, it treats a point of growth and a point of margin as equal, when durable profit is usually worth far more than fast growth that may not last. Second, the margin it uses is often an adjusted figure that adds back stock-based compensation, which flatters the score. Third, a business can clear 40 for a year or two on the way to running out of customers, so a single reading tells you little about durability.

Consider a hypothetical example. A company growing revenue 50 percent a year with a negative 15 percent free-cash-flow margin scores 35 and fails the rule, yet may be a superb business investing hard into a huge opportunity. Another growing 10 percent with a 32 percent margin scores 42 and passes, and may be a mature, cash-rich compounder. The rule ranks them almost equally, which is exactly why it should inform your thinking rather than settle it. Use it to find candidates, then judge each one on the quality and durability of its revenue.

Stock-based compensation: the modern trap

The biggest trap in software analysis today is stock-based compensation, the practice of paying employees in shares rather than cash. It is a real cost, because every share issued shrinks the slice owned by existing shareholders, yet it never leaves the bank account, so it quietly inflates the cash-flow figures that investors rely on.

Here is why it matters. Reported free cash flow adds stock-based compensation back, treating it as a non-cash expense. That is technically correct on a cash basis, but misleading on an ownership basis, because the company has handed away part of itself to pay its staff. A business can look strongly free-cash-flow positive while its share count creeps up two or three percent a year, steadily diluting the owners it is meant to serve.

Microsoft shows the honest version of this. In fiscal 2025 it recorded about $12.0 billion of stock-based compensation, equal to roughly 4 percent of revenue and about 12 percent of net income, per its 10-K. Crucially, it spent about $18.4 billion buying back stock in the same year, more than enough to offset the shares it issued, so its diluted share count stayed roughly flat. That is the discipline to look for. A company issuing heavy stock-based compensation without buying back enough to offset it is transferring value from owners to employees, however healthy its cash flow appears. The mechanics of that offset are covered in share buybacks explained.

The defense is simple. Read free cash flow, then check the diluted share count over five years, since that is where dilution actually shows up. If shares outstanding are rising, subtract the dilution from the return you expect, and treat any margin that conveniently excludes stock-based compensation with suspicion.

What good looks like

A high-quality software business shows a recognizable pattern across several years, not one flattering quarter. Revenue is mostly recurring and growing at a healthy clip. Net revenue retention sits comfortably above 100 percent, so the installed base expands on its own. Gross margin holds in the 70s or higher. And free cash flow is real, with a diluted share count that is flat or falling rather than drifting up.

The table below sketches the difference between a durable software business and a fragile one. The figures are illustrative round numbers, not any single company.

Two habits keep the analysis honest. Compare a software company only with other software companies, because the margins and multiples look nothing like an industrial or a retailer, and a number that looks alarming beside a bank may be perfectly normal for a subscription business. And weigh the durability of the revenue against the price, since the whole appeal of the model, predictable repeating income, is often already reflected in a high multiple. The stickiness that protects that revenue, high switching costs and deep integration into a customer's workflow, is a moat much like the brand loyalty examined in how to analyze a consumer brand, and the way predictability supports a valuation is the theme of pricing power explained.

Where to go from here

Analyzing a software company comes down to judging how much of its revenue truly repeats, whether its customers grow more valuable over time, and what profit survives after the shares it pays out. Start with the mechanics of the model in recurring revenue business models, then read gross margin to see why software economics look the way they do. When you are ready to test a name, open the Microsoft financials on Tenet and trace the recurring revenue, margins, and share count for yourself.