What a review actually looks like.

The order total never appears before payment

Delivery fee, service fee and the slot surcharge are all applied on step 3. A shopper who budgeted RM60 first sees RM73.40 with their card already out.

FixShow a live running total in the sticky bar from step 1, itemised on tap. Nothing else on this list moves the number as much.

Guest checkout is hidden under “More options”

Account creation is presented as the only path. The guest route exists but sits behind a tertiary link that reads as a settings menu.

FixLead with guest checkout and offer the account at the confirmation screen, once the purchase is already done.

Primary button fails contrast on the sticky bar

The mint fill against white text measures 1.9:1. In daylight on a phone it reads as a disabled control, which is precisely the moment it needs to read as the way forward.

FixDarken the fill until it clears 4.5:1, or invert to dark-on-mint. This is a token change, not a redesign.

The promo field outranks the thing you came to do

A full-width, focused promo input sits directly above the primary action. Shoppers without a code read it as a step they are failing; shoppers with one leave to go find it.

FixCollapse it to a single line of text under the total, expanding on tap.

The default screen is not the screen anyone works in

Dispatchers spend the day resolving failed deliveries. That queue is three clicks deep; the landing screen is a KPI dashboard whose numbers are read weekly, by someone else.

FixLand on the exception queue. Move the KPI dashboard to its own route for the people who actually use it.

Fourteen columns, none sortable, no sticky header

At the second screenful the column headings are gone, so dispatchers count cells across to work out which field they are reading.

FixSticky header, sort on the four fields that carry decisions, and let the rest live behind a column picker.

Delivery status is carried by colour alone

Amber and red dots are the only difference between “late” and “failed”. Those two states have completely different responses, and roughly one dispatcher in twelve cannot reliably tell them apart.

FixAdd a short text state beside the dot. Colour becomes reinforcement rather than the whole message.

Filters reset on every navigation

Opening a job and coming back drops the filter set. Dispatchers rebuild the same four filters dozens of times a shift.

FixPut filter state in the URL. It also makes shift handover a pasted link instead of a spoken list.

No case study says what happened after launch

All three end on a final-design shot. A hiring designer reads that as work that was never shipped, or shipped and never measured — and cannot tell which.

FixOne honest closing line per study. “Shipped in March; support tickets about step 2 dropped by roughly half” beats a hero shot. If it never shipped, say so and say why.

The first screenful is a mood board

Colour swatches and type specimens occupy the opening viewport. The problem statement is below the fold in every study.

FixOpen with the problem, who had it, and your role, in three lines. Move the visual exploration to the middle, where it supports rather than delays.

The process diagram is the same in all three studies

Discover → Define → Ideate → Design → Test, unchanged, with no evidence any of it happened. It reads as coursework rather than as an account of the project.

FixDelete the diagram. Replace it with the one decision you got wrong first and what changed your mind — that is what interviewers actually probe.

Eleven screens shown, two explained

The screen grid is doing the work of a portfolio, but nine of the eleven carry no caption, so the reader cannot tell which decisions were yours.

FixCut to four screens. Caption each with the decision it represents.