The study
We examined the mobile applications of ten scheduled commercial banks:
| State Bank of India Punjab National Bank Canara Bank |
Union Bank of India Bank of India Bank of Baroda |
Axis Bank ICICI Bank HDFC Bank IDFC First Bank |
The set was chosen to balance legacy public sector institutions against newer private sector ones. Journeys covered account statements, tax deduction certificates, interest certificates, nominee-related (henceforth ‘nomination’), KYC status, branch location, and, where the participant held the relevant product, Public Provident Fund (PPF) account opening and loan-related documentation.
Journeys were completed by student users at Krea University in task-based walkthroughs, with each step captured, in early 2026. See more on the methodology and evidence in the companion note.
It is worth being explicit about what this design can and cannot support. With a single participant per bank, there is no basis for claiming how long a task takes, how frequently users abandon it, or how one bank compares with another on effort. Those are properties of people, and establishing them requires a sample, a timing protocol and repeat trials.
However, what such a corpus can support are claims about the interface itself: what a screen is called, what state it reports, whether it contradicts itself, and where a journey terminates. These are properties of the application and do not vary with who is holding the phone. We therefore treat this as an interface audit rather than a usability study, and organise the findings under three questions:
| A | Where the journey ends |
| B | What the service is called |
| C | What the interface decides on the customer’s behalf |
These are discussed below.
Outcome: where the journey actually ends
Feature availability across the sample was high. Nearly every bank offered nearly every service examined in some form. Availability, however, is not the same as completion.
Journeys in this sample terminated in materially different places. Some delivered the requested document inside the application. Others delivered it only to a registered email address, so the artifact does not arrive in the customer’s hands within the session. Two banks served what is nominally an in-app branch locator as an embedded rendering of the bank’s public website, complete, in both cases, with the website’s login button visible, to a customer who was already authenticated via the app.
A further category of journey did not conclude digitally at all. Across the sample, several journeys ended in an instruction to visit a branch. This occurred for KYC completion at three banks in the set. It occurred for nomination at two, one of which caps mobile nomination requests at four per account per calendar year and states that changes beyond that require a branch visit with charges as applicable. A third bank offered no facility to add or amend a nominee at all; the only action available on the relevant screen was deletion.
Reducing branch dependence is a substantial part of the rationale for investments by banks in digital channels.
Where an essential journey terminates at a counter, the digital delivery is partial, and it may be worth measuring that partiality directly rather than inferring service quality from adoption figures.
Language: what the service is called
The most consistent finding in the sample was inconsistency of vocabulary. The same service carries a different name at nearly every bank examined.
An account statement appears variously as a statement, a request for a statement, an emailed statement, and, at three banks, under a digital passbook metaphor inherited from branch-era artifacts (possibly used due to familiarity with the word ‘passbook’ for its customer base). At one bank it is filed within a payments and transfers section; at another it sits within a section of trending features, adjacent to travel bookings.
Tax certificates show the same divergence with higher stakes. Some banks use the plain-language description, some the statutory form number, and two present both the form number and the plain-language name as separate destinations on a single screen, with nothing to indicate whether they lead to the same document. At one bank, the label used denotes a different statutory instrument from the one the customer is likely seeking. Nomination appears as maintenance, as registration, and as addition or modification. A search for KYC most commonly surfaces the term ‘re-KYC’.
Internal register (syntax used within the bank’s core-banking systems) also leaks into customer-facing screens. The sample contained a compliance status expressed as a single letter, a status described as complied, a date field rendering an error string, and help text listing internal product scheme codes.
Because participants frequently reached services through in-app search rather than menu navigation, search behaves as the de facto information architecture, and its quality varies considerably across the apps studied.
Comparison across providers and the articulation of a grievance both depend on a shared vocabulary, and the evidence here suggests that shared vocabulary does not presently exist for these services.
Trust and consent: what the interface decides for you
A third set of observations concerns whether the interface deals straightforwardly with the customer at points of decision.
The first question is whether the customer is told the truth about their own account. In one instance, a bank displayed an affirmative KYC confirmation alongside, on the same screen, a message stating that the central KYC record was not updated and that a branch visit was required. In another, there were two routes to the same service, one through the menu and one through search, and these returned different answers, one functional and one indicating the feature was forthcoming. Elsewhere, a stated product benefit appeared directly above a constraint contradicting it. At two banks, the investment journey collected personal information or required a statutory declaration before displaying the applicable interest rate or minimum contribution.
The second question is whether the customer is permitted to decide. The sample contained a statutory declaration to be agreed by customer, agreeing to Terms & Conditions displayed, before the interest rate, minimum contribution and other product features were shown for an investment product. And, as noted, another contained a nomination screen on which the only available action was deletion of existing records.
The bar already exists inside the sample
It would misrepresent the evidence to present these as sector-wide deficiencies. For nearly every weakness observed, a counterexample appears at another bank within the same ten.
- ✓One bank consolidates every certificate type into a single screen, crossed with account type, and offers an optional digitally signed statement, which is the form institutions typically require for verification.
- ✓One displays the financial year and the corresponding assessment year together, removing a common source of error in tax documentation.
- ✓One resolves every relative date preset into absolute dates before the customer commits.
- ✓One states a specific KYC validity date rather than a status flag. One warns the customer before handing off to a browser.
- ✓One offers documents bilingually.
Notably, these examples are distributed across both public and private sector institutions.
The variation in this sample appears institutional rather than sectoral, which suggests these are design and prioritisation choices rather than constraints of capability or cost.
Conclusion
This was an exploratory audit of ten applications over a single period, and it should be read as such. Applications change, and at least two features in this sample changed state during the observation window.
What it does suggest is that the documentary layer of retail banking has been substantially digitised, while the compliance-facing layer, the services that exist because a regulator requires them, has been digitised less completely and less consistently. That is close to the inverse of what the emphasis on payments would lead one to expect.
A structured programme could establish this properly, and two measures seem worth constructing.
First, is the proportion of essential non-transaction journeys a customer can complete entirely within the application, without recourse to a branch, a call, a browser or an email, and
Second, is a nomenclature index recording how many distinct names the market uses for a single service.
Both are countable, both are repeatable, and both would be more informative about the state of digital customer service than adoption volumes alone.
See more on the methodology and evidence in the companion note.
Dhruv Jesrani is a UX designer and an independent consultant working across brand and product, with a Masters in Communication Design from Parsons School of Design, New York. Data for the study was collected by Kunal Sen, a student of the MBA program at Krea University.