A glass-style direction
A more contemporary visual treatment with layered cards and a different presentation of account information. We wanted to understand acceptance, interpretation and the learning involved.
A/B concept testing with 34 participants to choose a design-system direction—combining visual preference with observed payment and navigation behaviour.
Research Lead, jointly owning design direction with the Design Lead. I coordinated research, design and UX writing outcomes, reviewed work and connected findings with delivery decisions.
A utility app and website serving approximately seven million users in the project context needed a shared design direction. We had two alternatives: a contemporary glass-style interface, or an evolution closer to the existing visual language.
The starting point was not a product that users uniformly disliked. Participants said the existing experience supported their main needs: viewing and paying bills. The challenge was to understand whether a more substantial visual change would be accepted—and whether the new concepts still supported those familiar tasks.
A more contemporary visual treatment with layered cards and a different presentation of account information. We wanted to understand acceptance, interpretation and the learning involved.
A design closer to the existing experience. We used it as the alternative for evaluating visual preference and the clarity of key interactions.
How might we select a design-system foundation that users want to adopt, while making essential bill-payment and account-management tasks clear enough to complete?
Which overall look and feel would users prefer, and why?
Could they pay, change account selections and adjust an amount?
How did they interpret usage information and respond to light and dark modes?
Project evidence: Concept Study, pp. 6, 15–22. Concept descriptions and product scale also reflect my project account.
I led the research and jointly owned the design direction with the Design Lead. We combined contextual questions, moderated usability tasks and A/B concept evaluation so that a visual preference would not become our only decision criterion.
We had two concrete directions and needed users to experience them before choosing. The questionnaire asks participants to compare payment, usage displays and dark themes, then explain their overall preference.
This made the decision more specific than asking whether they liked a new design in isolation.
A preferred interface can still make a task harder. Observing actions helped us identify hesitation, interpretation errors and recovery, then decide which interactions needed to change within the selected direction.
The task block was planned for 45 minutes, allowing both structured activities and follow-up questions.
The project called this A/B testing: a moderated comparison of prototype concepts and preferences. It was not a live traffic experiment measuring conversion uplift. NN/g’s live A/B guidance describes a different statistical setup. Here, task percentages describe the tested sessions and help prioritise refinement.
NN/g reference [1]Project evidence: Concept Study, pp. 6, 8, 23–27.
We recruited 34 participants. The inclusion criteria focused on recent product use, adult account responsibilities and the ability to engage with the prototype in a supported language.
Existing app or portal users who had used the platform within the previous three months.
Aged 18 or above, with a smartphone or computer, and comfortable using digital services.
Able to read and understand Bahasa Melayu or English.
| Group | Planned coverage | Why include them? |
|---|---|---|
| Regular users | 34 overall | Understand navigation and discoverability in bill-viewing and payment journeys. |
| Older users | 8 aged 55+, using the app independently | Explore hierarchy and whether screens could be understood without help. |
| Younger users | 8 aged 18–25, using the app independently | Explore expectations of modernity and functionality. |
| Low-vision users | 5; colour-vision differences desirable | Investigate font sizing, contrast and zoom-related needs. |
These are overlapping recruitment targets, not four additive groups. The report states that older-user recruitment was met and younger recruitment was nearly met; it does not provide a verified completed low-vision count. Concept Study, pp. 7, 10.
Reported counts total 34. Concept Study, p. 9.
The published age chart totals 31. The grey remainder represents the difference from n=34, not an inferred age group. Concept Study, p. 10.
IT was the largest occupational group (13), followed by telecommunications (4) and architecture (3). This gave us useful feedback from digitally comfortable users, but it also limits how confidently we can generalise to people with low digital confidence. Location was similarly concentrated in the central region.
Our initial expectation was that older users might prefer an incremental visual change. My recollection of the sessions was more nuanced: older participants also responded positively to the contemporary direction. The report does not publish an age-by-preference cross-tab, so I present that as a qualitative learning rather than a quantified age effect.
Project evidence: Concept Study, pp. 9–11; age-related interpretation from my project account.
The discussion guide connected existing habits with what participants saw and did in the prototypes. Questions about multiple homes were particularly useful: the person monitoring a bill was not always the person paying it.
| Stage | Key questions / prompts | What we wanted to learn |
|---|---|---|
| Before the tasks | How many homes do you pay electricity bills for? Where do you view and pay? Which payment methods do you use? | Account responsibilities, existing habits and channel choices. |
| First impressions | What do you see here? What do you think this icon does? What would happen if you tapped it? | Understanding before exploration or explanation. |
| Account interpretation | Whose usage is shown? What does a favourite bill mean? How many bills would this payment action include? | Whether account scope and payment scope were clear. |
| A/B comparison | After using both designs, which payment experience do you prefer, and why? Which usage display is easier to understand? | Experience-based preference, with reasons. |
| After the tasks | What do you think of colour, text size and spacing? What would you change? Which design would you choose to use? | Overall acceptance and priority improvements. |
Show how you would pay. Explain what you understand when bills appear in the payment list.
Select the parents’ and rental-home accounts, then change the parents’ payment amount to RM100.
Switch from light to dark mode, return to the main screen and describe the experience.
We also explored login icons, registration and recovery entry points, account usage and chart comprehension. The guide introduced prototype limitations and included follow-up probes to understand participants’ reasoning.
Asking what a control means before explaining it helps reveal interpretation problems. A specific account-and-amount scenario tests more than the easiest payment path. Asking “why?” after preference makes the findings actionable: it helps distinguish visual appeal from information clarity. For a repeat study, I would also document and counterbalance concept exposure order; the report does not establish that procedure.
Project evidence: Concept Study, p. 8 and Appendix A, pp. 23–27. Rationale is my interpretation of the study design.
The reach makes design decisions consequential.
A focused sample for concept direction and usability learning.
The study combined qualitative observations with descriptive task metrics. Its sampling plan referenced NN/g, recruited 34 participants and allowed for further recruitment if patterns had not formed. The report states that findings continued to repeat and that the team reached thematic saturation.
We needed to understand acceptance, confusion and recurring behaviours. NN/g recommends analysing interviews as research progresses and extending recruitment while important new themes are still emerging. Scope and participant diversity affect how much research is needed.
NN/g reference [3]Sample planning depends on the metric, acceptable uncertainty and decision risk. The report cited a 28-participant benchmark; the assumptions behind that option are explained below. Task-level sample sizes still need to be considered separately.
NN/g reference [2]In industry-led research, a 90% confidence level can be acceptable when the consequences of uncertainty are manageable. NN/g describes this as an option for practical UX research, with 28 participants needed for a binary metric at a ±15 percentage-point margin of error under its stated assumptions.
Our 34-participant study informed an iterative concept decision, supported by observed behaviour and repeating themes. This provides a rationale for the scale of the research; it does not establish that every reported task result achieves that confidence or precision.
Read NN/g’s sample-size assumptions [2]For a large population, sample requirements are driven primarily by the metric, precision and risk assumptions—not a fixed percentage of all users. Industry research can accept a different precision–risk trade-off when the decision warrants it; it still needs those assumptions to be explicit. The task subsets in this report cannot inherit the confidence or precision of the full sample.
The study supports a formative direction decision. Recruitment was concentrated geographically and occupationally, and some task denominators are not reported. A live A/B experiment to establish conversion uplift would require a separate sample calculation and experiment design. The report’s saturation statement concerns repeated themes, not statistical precision.
Project evidence: Concept Study, p. 6. NN/g links below explain the conditional sample-size guidance.
Percentages reproduced as reported, including source rounding. Overall study n=34; the report does not provide raw preference counts. Concept Study, p. 22.
The overall preference supported Concept A as the visual foundation. But task performance showed where its interactions needed refinement. That distinction was the most valuable outcome of combining the two methods.
Reported percentages, not population estimates. Per-concept task denominators are not explicitly stated, so they are not assumed to be 34. Concept Study, p. 15.
| Outcome | Concept A | Concept B |
|---|---|---|
| Completed first attempt | 46.67% | 76.47% |
| Completed after multiple attempts | 46.67% | 23.53% |
| Could not complete | 6.67% | No failures reported |
Users struggled to work out how to change the amount and account selection from the payment review screen. Concept A was more appealing overall, but its lower first-attempt completion made payment refinement a priority. This was not a reason to discard the whole concept; it was a reason to separate the visual foundation from the interaction details.

In Concept A, participants noticed and understood the payment action but hesitated before selecting it. The report describes a clearer relationship between the amount and adjacent payment action in Concept B.
Use Concept B’s payment placement to inform the selected direction. This directly addresses the observed hesitation, without claiming eye-tracking measurements or proven conversion uplift.
Principle tags explain the design rationale retrospectively; they are not additional study results or a separately scored heuristic audit.
Account selection & amount editing · Concept Study, pp. 15, 26
First-attempt completion was 46.67% for A and 76.47% for B. Some participants recovered after several attempts; A also recorded failures.
The scenario deliberately combined two homes with a custom amount. It exposed friction that the successful full-payment task did not reveal.
Keep the selected account and amount visible, provide recognisable edit controls and let users review the changes before committing.
Account usage discovery · Concept Study, pp. 20–21
Some stopped looking for the chart. After an explanation, they understood that the page resembled home and that the chart required scrolling.
Visual consistency is useful, but making an overview and a detail page too similar can obscure their different purposes.
Use a clear account heading, a visible route back and a recognisable usage section. Retain the bar-chart format participants found easier to understand.
Reported account distribution, Concept Study, p. 12.
Some participants monitored accounts for relatives or business outlets while someone else made payment. Others paid individually because they needed a separate receipt for each transaction.
Favourite accounts could help distinguish bills they usually paid from accounts they only monitored.
Rounded percentages from Concept Study, p. 13; response base is not specified. Methods are generalised to remove provider names.
Participants described established habits around saved banking payments, and some felt safer using their banking app. Better button visibility therefore addressed only part of the problem. Payment design also had to respect account ownership, receipt needs and channel trust.
| Screen / interaction | What we learnt | Design implication |
|---|---|---|
| Onboarding | Some users skipped when allowed; others read to discover what had changed. | Keep a clear skip route and make the value of the update easy to scan. |
| Login | The facial-login icon was sometimes mistaken for a camera or scan action; an identity-service logo was unfamiliar. | Pair unfamiliar authentication symbols with clear labels. |
| Account usage | In Concept A, a subpage looked too similar to the home screen. Users repeated the path or missed the chart below. | Differentiate account detail from the home overview and make context visible. |
| Usage chart | Participants preferred a bar chart for understanding usage; familiarity with the existing display helped. | Choose chart form for comprehension, not visual novelty. |
| Dark mode | Both concepts recorded 100% first-attempt switching in the reported task results. | Carry theme support into component and token planning. |
People could welcome a contemporary visual direction while still needing familiar, explicit task cues. The strongest solution combined the preferred look and feel with clearer payment placement, account context and information hierarchy.
Project evidence: Concept Study, pp. 12–22. Design implications are synthesised from the findings; they are not all verified release outcomes.
Preferred visual foundation
Clarity from Concept B and observed tasks
A coherent basis for continued refinement
I translated the study into a direction the team could work with: carry forward Concept A’s visual language while resolving the specific usability issues before treating the library as a stable foundation.
Align the team on the contemporary direction and define how it behaves across light and dark themes.
Improve amount editing, account selection, payment prominence and the distinction between overview and detail.
Develop component guidance, semantic token naming, responsive behaviour and complete interaction states with the Design Lead and developers.
The result was a direction decision and a refinement agenda, not a claim that every screen in the winning concept should ship unchanged. Subsequent discovery, SME reviews and usability validation extended the work into registration, authentication, account settings and professional-user journeys.
Direction evidence: Concept Study, p. 22. Subsequent process: MVP Research Report, pp. 6–8, 29–42. Design System responsibilities reflect my project role.
Different study stages. They are not pooled into one sample or assumed to be unique participants.
The later draft report documents 15 participants and says findings were repeating. It includes different task subsets, so the overall study size is not the denominator for every task.
Each square represents one successful task attempt in the reported six-person subset. These counts describe prototype sessions, not live customer conversion.
Task location, instructions and recovery needed further attention.
The draft identifies this as a major discoverability problem. The design implication was to review labels and the location of account information.
Successful payments did not remove the need to improve supporting journeys.
The report’s recruitment targets total 16, while its completed sample is 15. Some demographic totals and task summaries also differ. I use the explicitly documented six-person payment subset and the internally consistent 2/2/2 password breakdown, rather than derive a combined success rate from incompatible totals. Detailed findings were still marked as under review.
Jointly set the Design System direction with the Design Lead; organise component libraries and semantic naming practices.
Review default, focus, error, loading and empty states, responsive behaviour, long content and accessibility considerations.
Document component usage and do/don’t examples, share a writing playbook, and guide designers and developers.
I reviewed semantic naming best practices with developers and worked with architects on feasibility. My coding background helped me translate interaction intent into implementation discussions.
I coordinated two researchers and shifted designers between prototype preparation, usability support, rapid fixes and existing deliverables. An additional stakeholder exhibition required presentation materials, speakers and a live feedback setup.
I guided eight designers through reviews and feedback in a shared reporting arrangement. Project management maintained formal administration and final sign-off; I contributed requirements, sequencing and review decisions.
My role was to help the team make a decision with evidence, make its limitations clear, and turn that decision into standards people could implement. The work demonstrates research-led delivery; the supplied evidence does not establish a measured production adoption uplift.
Project sources are cited by neutral labels and page number throughout. Source reports remain off the public website to avoid exposing client identities and participant information.
September 2025 · 28 pages. Primary evidence for the 34-participant A/B concept comparison, recruitment, demographics, questionnaire, task results and preference findings.
October 2025 · Later discovery and validation. Read alongside the later validation DOCX; its 15-participant study is kept separate from the original concept comparison.
Explains live experiment design; helps distinguish this prototype comparison from live conversion testing.
Explains why sample recommendations depend on confidence and precision, including the conditional 28-participant option.
Explains saturation, recruitment relevance and analysing as research progresses.
Jakob Nielsen’s principles inform the tagged recommendations for control, prevention, recognition, status and consistency.
Supports grouping the payment amount and its related action.
A lens for interpreting visual preference alongside task performance.