Convenience is usually a real capability. A single login replaces repeated registrations, automatic synchronization reduces duplicate entry, and a familiar interface lets a person leave attention for the task itself. To deny the value of convenience merely because a service creates dependence would be to demand that everyone continually pay meaningless friction for the sake of independence. The problem appears when the accounting for convenience stops at the entrance. Joining takes only a few minutes, while leaving requires moving files, re-proving eligibility, learning another set of procedures, recovering contacts, and absorbing the errors of the transition period. The price has not risen, yet a person's future options may still shrink because of migration costs.
We continue the fictional learning program. Zhou Yao uses the platform provided by her course institution to store her work, receive tutoring, and contact classmates, and she also uses the platform account to certify her progress. Debt pressure makes her wish to move to a less expensive community course. The new platform allows same-day registration, and the old platform also offers "one-click export". Only after downloading the files does she discover that the work descriptions, revision histories, tutor comments, peer relationships, and credential status each remain in different places.
This thought experiment corresponds to no real product and offers no method for circumventing platform restrictions. It only breaks migration down into tasks: what can be copied, what must be re-explained, who is responsible for correcting errors, and who maintains the life and relationships during the transition.
The Full Switching Costs Beyond the Entrance
A registration page measures the steps needed to join; it cannot represent the exit after a year of use. The service completed the setup for Zhou Yao early on, and as works, reputation, and relationships accumulate, the objects that leaving must handle grow ever more numerous. This need not come from deliberate design. Any long-term collaboration produces shared history, and hospitals, schools, communities, and friends cannot be replaced the way a blank account can. Switching costs are not in themselves evidence of impropriety. The old platform placed courses, files, and communication behind a single entrance, so Zhou Yao did not have to remember multiple paths. After migrating, she must check device compatibility, file versions, deadlines, and contacts, and also explain to her family why her study arrangements have changed again. The interface displays "export complete"; that only means the system task has ended. If the exported material cannot be recognized by the new environment, the work of understanding and repair is transferred to the individual. Technical success and completion in life must be checked separately. The platform may have fully discharged its contractual obligations while Zhou Yao still bears real costs. A judgment of responsibility cannot ask only whether a breach occurred; it must also ask whether the costs were knowable in advance, whether they arose from one party's choices, who obtains the benefits, and whether the person has an alternative she can actually bear. The full price of convenience therefore includes money, time, learning, relationships, errors, and opportunities. Not every item can be converted precisely into an amount, but omitting them miswrites migration failure as a user's lack of perseverance.
On the old platform, Zhou Yao submitted work by simply choosing a course and uploading a file. After transferring out, she must first identify which files are the final versions, then convert proprietary annotations into plain text, re-enter the background of each work, contact her tutor to confirm whether the comments can be cited, and finally explain to the new course why her progress is inconsistent with her certificate status. Listing the tasks is not meant to exaggerate the difficulty. It restores "just switch platforms" from a single choice to a set of actions, and it makes visible the steps that could be simplified. That a file opens does not mean its meaning is intact. Fonts, timelines, links, and version relations within a work may change; plain text preserves the sentences yet loses who changed what, when, and why. Details irrelevant to ordinary reading may be crucial to a credential review. Open formats can reduce dependence, but they also have limits. A standard cannot anticipate every function, and pursuing complete compatibility would constrain innovation; if sensitive records were copied without limit, leakage would expand. Portability requires selecting objects according to the task, rather than demanding that all data flow freely forever. Devices are also part of the relationship. New software may be free yet demand higher performance, so a surface price of zero hands the hardware cost to the user. Conversely, a new system may lower device requirements, and technical progress does genuinely reduce total cost. Judgment must compare real tasks and cannot be settled in advance by "new" or "open".
Zhou Yao is familiar with the shortcuts of the old platform; the new environment can accomplish the same tasks, but the motor memory must be formed again. Course deadlines did not pause for the migration, and learning the new tools competes with completing the work for the same stretch of time. That the individual bears part of the learning cost can be reasonable. Any new capability requires practice, and demanding that the service provider eliminate all unfamiliarity would freeze change. The question is whether the platform repeatedly forces migration, whether it provides a stable transition, and whether its advertising deletes learning time from the cost. When a user stays in place because time is short, one cannot immediately infer that she prefers the original platform. Staying may express satisfaction, or it may express an inability to pay for the switch at the moment. When observing choices, affordability must be included in the explanation.
Zhou Yao exported her work files and message records. The new platform receives the files, yet it does not know the task structure of the old course, and it cannot convert "seventy percent complete" into its own credits. Data are one record of experience; continuing eligibility depends on rules of interpretation. Two institutions, even when seeing the same facts, may reach different judgments about quality, timeliness, and responsibility. The old platform cannot unilaterally guarantee that the new platform will recognize the credential. The new platform likewise has the right to maintain its own course standards and cannot automatically award outcomes merely because a user arrives with records. The two sides can provide mapping notes, samples, and manual review, so that Zhou Yao knows which content is recognized and which must be redone. Full equivalence is not a necessary condition; foreseeable difference is itself part of the capacity to migrate. If a platform advertises that "all assets belong to the user" while allowing only the download of data that cannot be interpreted, the language of ownership conceals the conditions of use. Possessing a copy is not the same as being able to continue acting elsewhere. Critics, for their part, should not write off every non-convertibility as lock-in. Professional training, cultural context, and shared practice may involve genuine differences. Evidence is needed to distinguish what is technically portable but artificially blocked from what inherently requires fresh judgment.
Zhou Yao's tutor's evaluation is not an isolated piece of data. It embodies the tutor's reputation, the course standards, and long observation, and a new institution need not treat a single "performed well" as a universal qualification. Yet if the old platform refuses to provide the basis on which the evaluation was formed and displays only an internal grade, Zhou Yao cannot prove the work she has already done. The accumulation belongs to a shared relationship, yet the exit is controlled by one side. Reputation comes from others' judgments of one's actions and cannot be copied by the person herself the way a file can. Allowing Zhou Yao to carry all her high scores without any disputes would injure the receiving side's judgment; resetting all reputation to zero would let the platform appropriate long labor. The more reasonable migration carries verifiable facts, works, and the sources of evaluations, while permitting the receiving side to reinterpret them. Disputed statuses should also be preserved, but they must not travel forever under a vague risk label. If credentials are tied to debt, employment, or housing, migration costs are amplified further. Leaving the course may cost Zhou Yao her repayment discount or an entry point to a job, so what she faces is not a single software choice. The future commitments discussed in Chapter 8 here become exit conditions. A more dominatory arrangement would convert the certification that ought to belong to learning outcomes into a favor that cannot be taken away. The more successful the person, the more assets remain inside the system, and the more leaving resembles abandoning the past. Pointing out this inference serves to examine the right to proof and the consequences of exit; it is not a lesson in how to design lock-in.
Zhou Yao knows classmates, tutors, and potential collaborators. The platform can export contact names, but it cannot guarantee that the other parties are willing to move to a new channel, nor can it copy the trust formed through shared discussion. Relationships have multiple parties, and therefore cannot be declared portable by one person. Protecting others' privacy and choices makes relationship migration inherently incomplete. If all communication took place only inside the platform, Zhou Yao can hardly notify anyone after she exits. The platform has not deleted the relationships, but it controls the paths of access. Providing limited notification and a choice of contact channels lets the parties to each relationship decide for themselves whether to continue. The platform must also prevent harassment and bulk abuse and cannot open all lists. The right to migrate is not a right to obtain other people's information. Design should let each participant authorize their own contacts, rather than treating the relationship network as an asset package of the user. Offline communities also have switching costs. Leaving an organization means losing shared places, rituals, and acquaintances, and this is no reason to find the organization malicious. The difference with digital platforms is that interfaces and permissions can close many relationships at once within a short time. Maintaining multiple contact channels can increase resilience, but it also increases the costs of attention and security. An individual should not be required to run a permanent backup network for every service; providers and public infrastructure can jointly bear basic continuity.
Zhou Yao discovers that the timestamp of one work was recorded incorrectly. The old platform says the account has entered the closure process; the new platform says the error comes from the old record; neither has the authority to modify the other's material. The interface state of "migration complete" does not mean responsibility is complete. Errors move along with the data, and the capacity to correct them should also have a path. The old platform is responsible for correcting the facts it generated and notifying known recipients; the new platform is responsible for updating its local decisions. If only the source data is repaired, the old copies may still continue to produce consequences. The person should be able to see the scope of significant propagation, though not necessarily to inspect all internal material. Correction notices, version marks, and independent review can reconcile responsibility with privacy. Keeping a limited correction window after account closure does not mean the platform bears the service indefinitely. Deadlines can be set according to consequences: high-impact credentials need longer protection than interface preferences. Zhou Yao, for her part, must preserve reasonable material and raise specific errors in time. An institution cannot attribute every loss to the individual, nor promise to rebuild the entire history without evidence.
Investment and Loyalty Must Not Be Confused
Zhou Yao has already spent months learning the old procedures and building a portfolio. Continued use avoids renewed investment; transferring out may be cheaper in the long run. Past costs should not govern the future within an abstract decision, yet they genuinely affect the time and resources available today. "Do not mind the sunk costs" works as a reminder not to keep pouring resources into a bad project for the sake of unrecoverable spending, but it cannot make the labor of migration vanish into thin air. Part of the skills already formed remains usable; another part is effective only in the old environment. Proficiency lets Zhou Yao work faster, and this is a capability the platform created or supported. If every capability had to be universal, specialized tools could never develop. Specificity in itself is not harvesting. When the provider keeps changing the compatibility conditions so that prior investment can be retained only through continued purchase, specialized capability may be converted into bargaining power. Establishing such intent requires evidence from versions, contracts, and conduct, and cannot be inferred from a single update. Users also overestimate switching losses, ignoring the gains of a new environment out of familiarity. Trials, parallel periods, and small-scale migrations can produce comparative material. They have costs, yet they are more testable than demanding one-time faith in some future. Public discussion should avoid calling those who stay lazy and those who leave traitors. Both choices may be reasonable responses to different circumstances of time, risk, and relationship.
The platform sees a high renewal rate and concludes that users endorse the service. Renewal is indeed one piece of supporting evidence, especially when alternatives are reachable, prices are transparent, and data are easy to migrate. If exiting would interrupt credentials and relationships, however, renewal mixes preference with constraint. Interpreting the mixed result entirely as devotion lets the system use the costs it created itself to prove its own value. Zhou Yao keeps logging in perhaps to finish the last course, to export material, or to contact classmates. The behavior shows she still needs the entrance; it does not show that she accepts all the rules. Likewise, a low switching rate does not automatically prove lock-in. The alternative products may be worse, and stable relationships also have value. An audit needs to ask about the tasks of exiting, the reasons for failure, the time involved, and the consequences, not merely compare headcounts. One extreme inference is that an organization can keep exit nominally free while tying past achievements, present relationships, and basic credentials to the entrance, layer upon layer. No one is forbidden to leave, yet the one who leaves must become, again, a person without history. Criticizing such a structure requires making the switching tasks public, making key achievements provable, and limiting the collateral consequences in unrelated domains. Judgment of "free choice" therefore cannot stop at whether the button exists; it must also ask whether a person can use it without surrendering irreplaceable conditions of life.
Unified standards can let files, credentials, and identities move across systems, reducing repeated proof. They may also lower the entry costs for small platforms, giving users real alternatives. Yet excessive concentration of standards can also spread one classification into every environment. An error in one place travels more easily, local differences are forced flat, and the shared interface becomes a new single point of dependence. One can share work files without sharing a complete behavioral profile, and carry verified credentials without carrying irrelevant social scores. The more specific the objects of interoperability, the more easily its expansion can be reviewed. Different domains do share facts, but uses should be re-authorized. That technology can read something does not mean institutions should invoke it. The cross-domain evaluation discussed in the previous volume can, amid switching convenience, likewise form a total credential. Security restrictions sometimes need to block immediate export, for example to protect others' information or to investigate anomalies. Exceptions should state their object, time limit, and review; "security" must not become a blank justification for permanently closing the exit. Maintaining standards also requires resources. If the costs fall entirely on small institutions and individuals, the demand for openness may instead concentrate the market. Public support, layered obligations, and shared tools can distribute the costs.
Who Bears the Costs of Migrating and of Staying
If Zhou Yao actively pursues a new course that suits her better, she has reasons to bear part of the learning and arrangement costs. The benefits of the choice accrue mainly to her, and the institution has no obligation to pay for every preference. If the old platform unilaterally discontinues service, raises a key price, or changes the use of outcomes, handing the entire migration to the user is harder to justify. The party that controls the change has the stronger reason to provide a transition. If the institution earns stable revenue from lock-in, it cannot treat exit merely as private consumption; since the user benefits from customized service, she likewise cannot demand that all specialized investment be absorbed by others. The terms of the relationship need to be stated at entry. People with low income or heavy caregiving burdens face the same migration tasks, and their opportunity losses may be greater. Providing assistance is not an assertion that they lack ability; it is a recognition of differences in time and resources. Subsidies may also induce unnecessary switching, or let public funds absorb the costs of a commercial strategy. Support can target key materials, basic credentials, and continuity of service, rather than guarantee that every choice is cost-free. Genuine fairness is not an equal split of every hour. Who creates the change, who controls the interface, who draws the benefit, and who can absorb failure together determine responsibility.
Exit is not the only issue. After a platform update, Zhou Yao may be willing to stay on the old version yet be required to upgrade equipment and procedures within a short time. She has not changed providers, and a transition still occurs. Maintaining an old system carries security, compatibility, and staffing costs, and indefinite support cannot be demanded. A provider may retire a version, but high-impact changes should come with notice, data export, and a reasonable transition. Old features may impede improvement, and long-term parallel operation also creates more failures. What continuity requires is that important tasks have a migration path, not that every interface remain unchanged forever. Platforms on which public services or critical credentials depend bear higher obligations, because the consequences of a failed upgrade exceed consumer inconvenience. Pure entertainment preferences can tolerate faster change. Obligation follows consequences, not technical names. Users, too, can push risk onto others by delaying. Clear deadlines, phased reminders, and supported migration can make responsibility visible. After a deadline, if someone still refuses a necessary security update, limiting some functions may be reasonable. If a forced upgrade simultaneously expands data use or cancels exit, it cannot be explained by maintenance alone. The objects of change should be stated separately, so that necessary technical migration does not smuggle in unrelated power.
When choosing a platform, Zhou Yao should be able to see how key outcomes are exported, who recognizes the credentials, how errors are corrected after closure, and which relationships cannot be carried automatically. The information need not predict every future; it should at least cover high-consequence dependencies. During the service period, the platform maintains readable formats, version notes, and limited parallel access. Relationship contacts are authorized by each party, and sensitive material is layered by use. One must not count only successful downloads. One should check whether the new environment can open the key files, whether credential disputes have review, whether propagated errors are corrected, and whether the transition interrupted basic opportunities. Organizations can periodically simulate a small number of voluntary migrations to discover shared dependencies. Testing does not require publishing security details, nor can it use the failures borne by participants as proof that the system is reliable. Public or industry infrastructure can preserve a minimum of common formats and credential descriptions, so that every exit does not start from zero. The centralized facility itself still needs versioning, oversight, and alternatives. Zhou Yao may in the end choose to stay. Staying after a comparison with a feasible migration says more about the value of the service; even if she still stays because of relationship costs, the institution should at least not package that constraint as total endorsement.
Zhou Yao also uses plugins to store material, pays installments, and lets a recruiting agency read her course work. The old platform provides the interface, but the consequences are distributed among multiple parties. The platform cannot guarantee for all third parties, and the third parties cannot, because they depend on the interface, be exempt from their own explanations. A person needs to know which key relationships will change when the account closes. Payment obligations may continue, links to public works may break, and copies of credentials already issued may still be in use. Exiting the main platform completes only one layer. The service can provide a list of connections and their statuses without making every decision on Zhou Yao's behalf. High-consequence third parties need their own correction and termination entrances, and cannot send her permanently back to a closed account. The richer the ecosystem, the greater the convenience, and the harder it is for a single contract to describe the entire future. Layered disclosure, shared standards, and clear responsibilities are more usable than one blanket consent. Third-party innovation can break the original platform's lock-in, and it can also establish new dependence. The value of an open interface must be judged together with its permissions, continuity, and replaceability.
Turning Exit Promises into a Verifiable Budget
At entry, Zhou Yao can reserve habits of organizing her material; the platform can provide stable export; public courses can recognize part of her existing achievements. The capacity to switch is formed jointly by many parties. The budget is not only money; it also includes transition time, assistance, parallel services, and a margin for errors. A person without margin may see a better path and still be unable to move. Preserving every old version and maintaining every interface would consume resources and drag down current quality. What to keep should be decided by the importance of the outcomes, their consequences, and their alternatives, rather than treating "possibly useful someday" as an unlimited obligation. Nor must an individual run all alternatives simultaneously. High-consequence dependencies take priority for readable copies and independent contacts; low-consequence preferences can accept being set up again. When an organization reports retention rates, it can also sample and record the tasks required to exit, the reasons for incompletion, and the collateral losses. Measurement will not automatically yield a verdict on value, but it can stop convenience from being narrated only by the speed of joining. The purpose of a switching budget is not to encourage frequent departure, but to make staying and leaving both decisions whose consequences can be observed. A recoverable exit in fact makes long-term cooperation more credible.
The platform states "free to leave at any time". The sentence expresses an attitude, but it says nothing about how data, credentials, payments, and relationships will end. Only when Zhou Yao actually begins the migration does she discover what the promise refers to. The service can make the promise checkable through documentation, sample exports, and limited trial migrations. The samples must approach real complexity and cannot merely display a blank file. One successful export cannot represent every account, and one failure does not prove that the platform as a whole is lock-in. What should be recorded are the data types, versions, receiving environments, repair times, and the tasks that still require human hands. Third parties can help with verification, but the boundaries of their tools and criteria must also be public. If a certification mark checks only the download button, it leaves semantics, credentials, and error correction outside. User feedback can reveal the costs in life, but it cannot replace technical and contractual verification. Someone who finds migration easy may have simple tasks; someone who finds it extremely difficult may have unusual relationships. Layered material is more explanatory than average satisfaction. Providing tests does not mean exit is without risk. The platform should state truthfully which relationships and functions cannot be carried, and the receiving side should state which credentials it will not recognize. Making losses explicit lets the person make arrangements; it is not an automatic exemption for the losses. Only when the promise survives a real migration does convenience include recoverability. A "freedom" that cannot be verified may still be genuine, but it is not enough to support high-consequence dependence.
Halfway through her migration, Zhou Yao finds that a key evaluation cannot be recognized; if the old account has already been destroyed immediately, she can only bear the loss between two incomplete environments. A limited fallback period gives her the chance to verify results and revise decisions. Fallback increases synchronization conflicts and security costs, and cannot be extended indefinitely. The platform can freeze new activity, keep the material read-only, and state a final closure date. Prices, courses, and technology will change, and a person cannot demand, from one failure, the restoration of all historical conditions. Fallback protects only the key outcomes and error corrections within the transition; it does not cancel the legitimate reasons for ending a service. High-consequence data must remain provable after closure, while low-consequence preferences can be set up again. Layering keeps recoverability from becoming a demand to freeze innovation.
The next chapter discusses why attention cannot be priced by duration alone. Switching costs occupy not only hours; they also interrupt continuous tasks, increase vigilance, and force a person to rebuild context again and again. The same three hours may have different effects on future capability. Convenience can also raise the price of exit: a system can, by providing real convenience, let outcomes, credentials, and relationships gradually become continuable only within its own entrance, and then interpret the person's not leaving as free loyalty. This process can be formed jointly by commercial incentives, administrative convenience, and technical path, without presupposing any secret intent. Criticism must also preserve its counterexamples. Long-term relationships necessarily leave history that cannot be copied, professional tools require specialized capability, security and privacy will restrict portability, and a person choosing a new service should also bear proportionate costs. The keys to judgment are whether conditions were knowable at entry, who controls the changes, whether key outcomes can be proven, whether errors can be repaired along the chain of propagation, and whether leaving requires the person to abandon parts of life unrelated to the original task. Convenience can reduce the friction of action; it should not, through hidden switching tasks, take possession in advance of a person's future choices.