Selling software once and renting it continuously are not two versions of the same business. The switch between them reshaped the industry more thoroughly than any change in the software itself.
The most consequential shift in software over recent decades was commercial rather than technical. Companies stopped selling a product outright, as a licence, and began supplying it continuously by subscription. The two models are barely comparable in their economics, and the difference accounts for a great deal of how software companies now behave.
Under a licence, the sale happened once. Money came in, revenue was recognised, and growth required either new customers or persuading existing ones to upgrade. Revenue in any given period depended overwhelmingly on the deals closed in that period, which made it lumpy and hard to forecast. Past success did very little to support present results.
Subscription inverts that. A customer pays for continuing access, so a customer won this year still produces revenue next year and the year after with no further selling. Revenue accumulates rather than resets. Each period opens with whatever the existing base contributes, and new sales build on top of that foundation instead of replacing what came before.
The model also distorts early financial results in ways that are easy to misread. Winning a customer costs money immediately, in sales effort and marketing spend, while the revenue from that customer arrives gradually over the following years. A company growing fast is therefore paying now for income that lands later, and its reported profit can look poor or negative while the underlying economics are perfectly sound. Run that logic backwards and it becomes uncomfortable. Such a company could improve its reported profit substantially by simply stopping customer acquisition, which would be a bad decision wearing the costume of an improvement.
What matters, then, are questions about the customer relationship rather than about any single period's profit. What does acquiring a customer cost? How long does that customer stay? How much do they generate across that span? Does their spending grow as they use more? The answers determine whether the acquisition spending is an investment buying a durable stream or merely a cost buying very little.
Retention carries particular weight, because the entire model rests on it. If customers leave at any meaningful rate, the accumulated base leaks, and new sales have to refill the hole before they produce any growth. A business adding customers impressively while losing them steadily may be sprinting in order to stand still. The departure rate, and whether the survivors spend more over time, reveal considerably more about the business than the headline growth figure does.
Underneath all of this sits a structural feature of software worth naming directly, which is that serving one additional customer costs remarkably little once the product exists. As the business grows, a larger share of each additional dollar of revenue can reach profit. That is the source of the profitability these companies eventually reach, and it is also why spending heavily on growth beforehand can be entirely rational.
The structure is worth setting out because it is so often misread, particularly when a growing software business gets judged against conventional measures of current profit that its model is not built to produce at that stage. Understanding the economics is where any further question has to begin. It is not a statement about what any particular company is worth.