Choice Is a Product Feature
Once a payer can use one app to reach someone on another, an app has to earn a user’s choice rather than inherit it from a closed network. This chapter examines that change through UPI’s user experience: why people may want different apps for the same payment, how they learn about alternatives, and what happens when reliability becomes observable. The practical test is whether interoperability gives users a usable choice and gives providers a reason to keep improving.
Preference is not the same as popularity
A payment app is not valuable for only one reason. Two people may want to transfer money to each other but still prefer different services. One may trust a bank with which they already have a financial relationship. Another may need a particular language, accessibility feature, or way of organizing payments. A merchant may value a different set of tools from a household. These are not irrational inconsistencies. They are differences in what people want from the interface through which they handle money.
In a closed-loop system, those preferences are constrained by the need to coordinate. If one person chooses App 1 and the other chooses App 2, the payment may fail simply because the systems do not connect. One person must give up their preferred service, both must maintain multiple apps, or the parties must use cash. The app with the largest installed base can therefore win decisions before a user compares its quality. Popularity is doing the work that compatibility should have done.
Interoperability breaks that link. When the payer’s and payee’s apps can connect through a common payment system, each person can choose an interface that fits their own needs. The relevant counterparty is still reachable, but the counterparty’s app no longer dictates the choice. This is a direct user benefit: a system with several differentiated apps can deliver the full value of that variety instead of forcing everyone into the same one.
The distinction is important for interpreting UPI’s early growth. The IMF’s evidence shows that, during the period when UPI was unfamiliar, more than half of users accessed it through an app produced by their own bank. A familiar bank was a credible entry point for a new financial service. Over time, the share using their own bank’s app fell, while users increasingly moved toward a small number of fintech apps that offered more features. The shift is consistent with people learning about alternatives and choosing services that better matched their preferences or perceived quality.
It is also consistent with a strong affiliation effect. Across the full sample, transaction values for a given app-bank pair were ten times higher when the app and bank came from the same entity. That result does not prove that every user was motivated by trust, and it is not a standalone causal estimate of trust. It does show that the relationship between an app and an existing financial institution can materially shape early usage. Familiarity can help an ecosystem get started, but it need not determine which app users prefer after they gain experience.
The mechanism can be summarized without treating preference as a matter of abstract consumer taste:
The benefit is not that every user will switch. A user may be satisfied with their current app or may value a familiar bank relationship. The benefit is that the choice can reflect those reasons rather than being forced by the inability to reach a counterparty. This makes the presence of differentiated apps economically meaningful. Without interoperability, variety exists on paper but may be unavailable at the moment a payment must be made.
That distinction gives “choice” two separate meanings. The first is freedom to choose an app that fits a personal preference. The second is freedom to choose a better service when quality differences become apparent. A system can support the first without guaranteeing the second: users may have several interfaces available, but remain with the familiar one if changing apps would break their payment relationships. Interoperability removes that structural penalty. It does not tell a user which service is best; it lets the user make that judgment without having to solve a separate coordination problem.
The value of that freedom is partly revealed through the sequence of UPI adoption. Users initially relied heavily on their own banks’ apps, when familiarity and an existing financial relationship reduced the uncertainty of trying digital payments. As the platform matured, usage shifted toward fintech apps with more features. The evidence does not observe each user explaining that decision, so it cannot prove that every move was caused by a specific feature or by reliability. It does show why a market that permits different interfaces can reveal preferences over time. Early trust and later performance need not point to the same provider.
That freedom also changes what providers are competing to supply. In a closed network, a provider can retain users because leaving means losing access to people and merchants who remain inside. Under UPI, an app can lose a user to a rival while the user’s payment relationships survive. The provider must therefore compete more directly on the service that the user can observe: ease of use, useful features, trust, and successful completion of transactions. The market signal is not simply that another logo exists. It is that the user can act on a difference between the services.
Reliability makes switching consequential
Quality is harder to see than an app icon, but payment failure makes it concrete. The IMF uses an app’s average transaction-decline rate as a proxy for quality. A decline is an attempted transaction that does not complete, for example because of a technical problem. Users may not know the exact source of a failure, but repeated failed payments make an app less useful. A provider that offers a more reliable experience can attract usage if users are free to move without losing their counterparties.
The UPI evidence is consistent with that pattern. A one-percentage-point lower decline rate in a given quarter is associated with approximately four percent higher total transaction value in the following quarter. The estimate accounts for app-specific differences in average usage and for national time trends through app and time controls. Its direction is intuitive: fewer failed attempts make an app more valuable, and users can reward that value when cross-app access removes the need to follow the incumbent network.
Transaction value is a useful outcome because it records whether payments are actually being processed, rather than merely whether an app has been downloaded or listed as available. But it is not a complete measure of user welfare or a pure count of users. A larger value can reflect more transactions, different transaction sizes, or other differences in usage. That is why the comparison is most informative when read with the app and time controls and with the conceptual mechanism, not as a standalone scorecard for app quality.
The relationship also works through perception rather than perfect information. Users do not need to know an app’s measured decline rate to notice that payments fail less often or more often. Providers, in turn, face a practical retention problem: if a user can keep paying the same counterparties through a competing interface, repeated failure is less easily protected by the incumbent’s network. Improving reliability may require investment, but the expected benefit is no longer limited to keeping users who have nowhere else to go.
The wording of the result matters. It is an association from app-level panel data, not a fully isolated causal estimate of reliability’s effect on every user. A decline rate may capture problems elsewhere in the chain, not only the front-end application. Providers that already attract more capable users or process different types of payments may also differ in ways that the measure does not capture completely. The evidence supports quality-sensitive choice, but it does not justify the stronger claim that lowering the decline rate by one percentage point will mechanically raise every provider’s transaction value by four percent in every market.
Even with that qualification, the measure reveals a competitive channel that app counts miss. If one app fails less often, users have a reason to prefer it. If users can switch while keeping access to their existing contacts, the better service can gain usage without first persuading those contacts to migrate. The threat of that movement gives an incumbent a reason to improve. It cannot assume that users will remain solely because everyone they know is already reachable through the same closed network.
The UPI pattern fits this interpretation in two ways. First, users moved away from their own banks’ apps toward major fintech apps that offered more features. The source mentions tools such as automatic bill payments, rewards, credit, investment, and insurance products, as well as innovations designed to reduce payment friction. These features are not automatically evidence of better welfare; they are examples of how providers can compete for the user-facing relationship once the basic connection is shared. Second, the largest apps showed especially large improvements in decline rates even as the number of transactions they handled grew rapidly. Providers were not merely collecting users; they were operating under pressure to handle a larger flow successfully.
This is a dynamic view of competition. Entry can make the incumbent’s position less secure, but the prospect of entry and switching matters even before a rival becomes the largest provider. A user who can try a better interface without abandoning a counterparty gives the rival a path to grow. The incumbent then has to decide whether to improve, add useful features, or rely on other ways of retaining the user. Interoperability changes the payoff from those decisions because poor quality is less safely protected by network effects.
The same mechanism explains why the value of “choice” cannot be measured only by the number of available providers. Imagine two markets with ten apps. In the first, each app is a silo. Users choose the app their counterparties already use, and a technically superior entrant cannot build a user base without first recreating a whole network. In the second, the apps share a rail. Users can try a service because it is more reliable or better suited to them, and the provider can win a subgroup without controlling every transaction around that subgroup. The app count is identical; the competitive meaning is not.
There is a further reason to keep preference and quality distinct. A user may choose a bank app because of trust even if another app is more reliable. Another user may prioritize fewer failures over a pre-existing bank relationship. Interoperability does not decide which value should dominate. It permits different users to make different decisions and lets the market reveal which combinations of trust, features, and performance attract sustained use. That is a more demanding form of competition than simply making every app identical.
It is also a bounded form of pressure. The IMF reports that the largest three apps eventually facilitated more than 95 percent of payer-side transactions. Quality improvements can therefore coexist with concentration. The evidence does not say that users’ ability to switch permanently prevents large providers from emerging; it says that interoperability gives users a route to alternatives and gives providers an incentive to compete on quality. Other forces may later narrow that route.
For now, the direct lesson is clear. Interoperability turns app variety into a potential user benefit by separating the choice of interface from the identity of the counterparty. It turns reliability into a potential competitive variable by allowing users to reward better performance without abandoning their payment relationships. But a better choice for an individual is not yet a solution to the market’s two-sided entry problem. The next question is how a new provider can reach enough users to become viable when merchants and consumers still depend on one another.