First-Time Banking Study

Problem — professional did not always feel understandable

A bank's public website asks people to compare products before they have learned its vocabulary. For a first-time visitor, choosing an account or estimating a mortgage is not a routine navigation task: the interface shapes whether the decision feels safe enough to continue.

This independent academic study evaluated BMO's public personal-banking website. It did not involve BMO as a client. Participants generally described the site as trustworthy and professional, yet the task evidence showed a different layer of the experience: people could reach pages and still remain unsure which option, tool, or result mattered.

Method — test complete decisions, not isolated screens

The detailed evaluation round included six participants aged 18–34. None used BMO as their primary bank; they brought expectations formed by TD, RBC, or CIBC. Each moderated session used think-aloud across three scenario-based tasks: compare chequing accounts and find opening requirements, choose a no-fee cashback card, and calculate a monthly mortgage payment before locating pre-approval information.

Scenario tasks fit the question because the friction often appeared between pages. A page-level heuristic review could identify inconsistent labels, but only an end-to-end attempt showed whether a participant chose the wrong calculator, missed a result, or abandoned a comparison feature.

Behaviour

Task success, step errors, path taken, and observed hesitation

Interpretation

Think-aloud comments and a semi-structured debrief

Benchmark

System Usability Scale after the session

Perception

Product reaction words to separate visual trust from task clarity

Triangulation

The method separated what people did, how usable it felt, and how the experience was described

01 · Observe

Behaviour in context

Moderation and observation captured hesitation, wrong turns, and workarounds that a post-task score could not explain.

02 · Interpret

Trust and friction can coexist

One completed form selected both “Professional” and “Cluttered,” a useful reminder that visual credibility does not guarantee decision clarity.

03 · Benchmark

An average can hide the spread

The 72.1 mean sat above the conventional 68 benchmark, while individual scores ranged from 62.5 to 82.5.

This was a five-person university team. We rotated moderator, observer, and note-taking roles across sessions and synthesized the shared evidence. I also translated the findings into interface concepts; those concepts are hypotheses, not evidence of an implemented outcome.

Evidence — trace each finding back to a task step

We marked every success criterion by participant and recorded the observed problem beside it. This kept memorable quotes from outweighing behavior. A finding became important when the same issue appeared across participants or across more than one product journey.

0102

1. 01 · Evidence pattern

   The same compare failure appears in both the chequing and credit-card tasks, which made it a system-level issue rather than a single-page defect.

2. 02 · Critical friction

   The mortgage task combines wrong-path navigation, missed results, and incomplete pre-approval discovery—the densest cluster of failures in the study.

The overall SUS score was 72.1, but the task evidence explains why that aggregate was not enough. Participants could rate the site as broadly usable while still relying on manual comparison or continuing to search after the answer was already on screen.

Findings — the site exposed products before decision logic

People compared from memory because comparison stayed peripheral

Four of six participants skipped the chequing-account comparison feature. They scrolled between cards and mentally retained fees, bonuses, and requirements. The same behavior appeared in the credit-card task, where only two participants used the comparison tool.

010203

1. 01 · Missed capability

   Four of six participants did not use the comparison feature; they scrolled between cards and compared details from memory.

2. 02 · Labels without a decision rule

   Names such as “Performance,” “Premium,” and “Plus” describe tiers but do not tell a first-time visitor which situation each account fits.

3. 03 · Requirements arrive late

   Three of six participants could not find opening requirements while they were still deciding between accounts.

Requirements were treated as application details, not selection criteria

Three of six participants could not locate the requirements for opening an account. The information appeared later in the application path, after people had already begun comparing options. For a first-time visitor, eligibility belongs beside fees and benefits because it determines whether an option is viable at all.

The mortgage journey created two different kinds of wrong turn

Two participants entered the affordability calculator before recognizing that the scenario required the payment calculator. Once inside the correct tool, all six participants missed the monthly payment displayed at the top and scrolled downward looking for it. Navigation vocabulary and result hierarchy compounded each other.

“Wait, that's the monthly payment? I scrolled right past it.”

0102

1. 01 · Choice before comprehension

   Two participants entered the affordability tool before realizing the task required the payment calculator.

2. 02 · Opportunity

   A short goal-based selector could ask what the visitor wants to learn, then route them to the appropriate calculator.

0102

1. 01 · Present but unseen

   All six participants missed the monthly payment at the top and continued scrolling for the answer.

2. 02 · Decision context

   Keeping the result beside the inputs would make the relationship between a changed assumption and the payment immediately visible.

Response — convert findings into testable hypotheses

The concept work explored two responses: a comparison view that explains why an option may fit a person's stated priorities, and a calculator that keeps the monthly payment prominent while assumptions change. Both reduce the need to remember values across pages.

The AI recommendation and “smart tip” treatments in the concepts go beyond what the evaluation proved. The study supports visible rationale and proactive guidance; it does not prove that AI is the preferred or most trustworthy delivery mechanism. A follow-up concept test should compare rule-based guidance, guided questions, and an AI-assisted explanation before selecting an implementation.

Learn — completed is not the same as confident

Binary task success concealed important hesitation. A participant could eventually reach the calculator and still select the wrong tool first; they could enter every value and still miss the number the tool existed to provide. Pairing step-level behavior with think-aloud data made those gaps visible.

The sample was small, younger, and primarily mobile-first, while sessions occurred in a controlled setting. The concepts were not tested and the study cannot support claims about conversion, application completion, or live business impact. The next round should include older participants and recent newcomers, test the concepts on mobile, and measure both first-path accuracy and whether people can explain why an option fits them.