The Real Product Is Not the Software. It Is the Cage.

By Ken Ho

There is a particular kind of trap that works not by preventing entry, but by preventing exit. The Victorians had a name for it: the roach motel. “Roaches check in, but they don’t check out.” The technology industry has refined the concept into something rather more sophisticated — and rather more profitable.

Consider the modern software-as-a-service subscription. It is presented as a convenience: a tool, available on demand, for a manageable monthly fee. No large upfront investment. No servers to maintain. The pitch is seductively simple, and it is, for the most part, genuine — at first.

But over time, something shifts. The tool becomes a system. The system becomes a workflow. The workflow becomes the way an entire organisation thinks about its operations. And then, quietly, the cost of leaving exceeds the cost of staying. Not because the product has improved, but because the accumulated weight of data, integrations, and habit has made departure unthinkable.

This is not an accident. It is, for many SaaS companies, the entire point.

The economics of captivity.

In 1999, two economists at Berkeley published what amounted to a blueprint. Carl Shapiro and Hal Varian’s Information Rules set out a framework for understanding how value is captured in digital markets. Their central insight was striking: in the information economy, profits come not from the quality of the product but from the cost of leaving it. “It’s the lock-in that generates the profits,” they wrote, “not the product itself.”

This was not a warning. Shapiro and Varian were strategy advisors, explaining how to construct defensible businesses. The venture capital community even has a metric for the result: Net Revenue Retention. An NRR above 100 per cent means existing customers spend more each year than those who leave — a figure only possible when switching costs make departure irrational. Top-quartile SaaS companies achieve NRR above 120 per cent. The market rewards captivity.

A brief history of the cage.

The strategy did not originate with SaaS. IBM built its dominance in the 1960s by bundling hardware, software, and services into a stack that made migration functionally impossible. Microsoft refined the approach: Windows and Office together created perhaps the most powerful lock-in computing has ever seen. When open standards threatened that dominance, Microsoft’s internal response — catalogued in leaked documents known as the Halloween Memos — was to adopt those standards, add proprietary extensions, and steer customers back into the walled garden. The internal name was blunt: “embrace, extend, and extinguish.”

Apple then perfected the model for the consumer era. Its tightly integrated hardware, software, and services generate roughly $85 billion a year in services revenue alone. The green bubble that appears when an iPhone user texts an Android device is not a technical limitation. It is a retention mechanism. The lock-in is not merely technical. It is social.

The SaaS playbook.

The modern SaaS company deploys a portfolio of mechanisms, each reinforcing the others. Data gravity — a term coined by the technologist Dave McCrory around 2010 — describes how accumulated data attracts applications, analytics, and further data to the same platform, making extraction exponentially harder over time. Cloud providers monetise this directly: data transfer in is free; data moved out is taxed at five to nine cents per gigabyte.

Ecosystem bundling compounds the effect. Microsoft 365 packages email, documents, storage, video conferencing, and messaging into a single subscription. Leaving any one component disrupts the whole. When Microsoft added Teams at no additional cost, Slack — then the dominant standalone messaging platform — filed an antitrust complaint with the European Commission. The complaint was withdrawn after Salesforce acquired Slack, which rather proved the point: the only viable response to platform envelopment was to be absorbed by a larger platform.

Then there is “land and expand.” A product is offered free to individual teams. Adoption spreads. The vendor is patient. Gradually the tool is embedded in daily operations. Only then does monetisation begin in earnest — and by that point, the cost of ripping it out far exceeds the cost of the licence.

The enshittification cycle.

The writer Cory Doctorow has given the lifecycle a memorable name: enshittification. Platforms are first generous to users — subsidised by investor capital, offering excellent service at low or zero cost. Once users are locked in, the experience degrades. Quality declines. Prices rise. And yet users stay, because leaving has become too difficult.

“Here is how platforms die,” Doctorow writes: “first, they are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the remaining value for themselves. Then, they die.”

Not every platform follows this trajectory to its conclusion. But the mechanism — that lock-in enables degradation — is widely observable. Amazon Prime’s insertion of advertising into a service sold as ad-free. Twitter’s restriction of API access, destroying an ecosystem of third-party tools. The steady creep of enterprise SaaS pricing. These are the predictable consequence of a business model in which the product is not the tool, but the trap.

The next frontier.

The latest iteration may be the most consequential. Microsoft Copilot, Google Gemini, Salesforce Einstein — AI tools embedded in existing ecosystems, trained on proprietary data, and available only through a single vendor’s subscription. The intelligence these systems accumulate cannot be exported. Data gravity, already powerful, is amplified by machine learning.

As Tim Bray, a former Amazon vice-president, observed upon resigning in 2020: “The trick is to make it easy to get data in, and hard to get data out. That’s not a bug; it’s the business model.”

There are countervailing forces. The EU’s Digital Markets Act requires gatekeeper platforms to offer data portability and interoperability. The cloud-native engineering movement promotes portability as a first principle. Open-source alternatives offer exit routes. Yet these face a fundamental asymmetry. Lock-in is profitable. Portability is not.

The cage, after all, is most effective when its occupants do not realise they are inside one.