Self-initiated · Concept redesign · Nairobi
Reimagining how Kenyan households order cooking gas
EasyGas is an LPG delivery app used by Kenyan households. This is an unsolicited redesign — no client, no brief, no access to their data. I rebuilt the three journeys that carry the whole product: ordering, confirming where you are, and knowing the gas is actually coming.
- self-initiated, no client brief
- Concept
- core journeys rebuilt end to end
- 3
- friction points removed from ordering
- 5
- designed for how households actually order
- Mobile-first
01 · Context
An everyday errand, running on an awkward app.
Cooking gas is not a considered purchase. A cylinder runs out mid-meal and someone needs a replacement within the hour. EasyGas already had the hard part solved — real delivery, real coverage, real payment. What it did not have was an interface that matched the urgency of the task it served.
02 · The challenge
The product worked. The path through it did not.
- 01
The app asked users to confirm their location repeatedly, on a task they had done before
- 02
Ordering gas — the reason the app exists — was buried behind menus
- 03
Visual hierarchy gave equal weight to primary and secondary actions
- 04
No dark mode, on a product often opened in a dim kitchen at night
- 05
No shared component system, so every new screen restarted the design
03 · Approach
Fix the journey first, then build the system
I worked from the live app rather than from a brief. I walked the three journeys a household actually uses, marked every place the product asked for something it already knew, and rebuilt from there. The design system came last, once the flows had settled.
Analyse
Walk the live product
Ordering, re-ordering and tracking on the shipped app, noting every repeated question and every tap that did not move the task forward.
Simplify
Delete the repeated question
Persistent location with GPS auto-detect and saved addresses, so the app asks once and remembers after.
Design
Lead with the primary action
One unmistakable Order Gas action on the home screen, with everything secondary demoted rather than deleted.
Systemise
Build it to survive growth
A component library, type scale and colour system carrying both light and dark mode, so new features inherit the decisions instead of re-litigating them.
The fastest improvement was not a new feature. It was deleting a question the app kept asking.
The redesign
Four moves, journey by journey
01 — Home screen
Making the primary action unmissable
The original home screen gave equal visual weight to promotions, categories and the order action. The redesign demotes everything that is not ordering gas, and gives the primary action the largest, highest-contrast target on the screen.
02 — Location
Ask once, remember afterwards
Location was requested again at almost every stage of checkout. The redesign combines GPS auto-detect with saved addresses in a single persistent step, so a returning household confirms delivery without re-entering anything.
03 — Order tracking
Proof the gas is on its way
Once an order is placed, the only thing a household wants is reassurance. The tracking screen leads with delivery progress, estimated arrival and status, rather than burying them under order metadata.
04 — Design system
Components, a type scale, and a dark mode
A structured palette, reusable components and a typographic hierarchy — carrying both light and dark mode, because this is a product opened in a dim kitchen as often as in daylight.
What it resolves
What the redesign resolves
This is concept work, so there are no adoption or conversion numbers behind it and I am not going to model any. What it does resolve is structural: the ordering journey loses five repeated interactions, the primary action becomes the most prominent element on the home screen, and the product gains a component system that lets it grow without redesigning itself each time.
What I learned
Most usability problems in everyday products are questions the software already knows the answer to.
EasyGas was not badly built. It was built feature by feature, and the friction accumulated in the joins — a location prompt here, a menu level there, none of them wrong on their own. That is the failure mode of products that ship faster than they consolidate, which is most products in a growing market.
It changed how I open an audit. Before I look at a screen I map what the product already knows about the person using it, then find every place it asks anyway. That list is almost always shorter and more valuable than a list of things to redesign.
