← Work/2026 · Product Designer

FarmGo

A native iOS grocery app in English and Arabic, redesigned from the address to the doorstep. Three stories: checkout, Arabic and right-to-left, and a meals mode.

125+
Screens in the app, 47 of them read flow by flow here
55+
Reusable controls, one design contract
EN + AR
Languages, with full right-to-left support

Source: the project README, design contract and test reports. No usage or sales figures are recorded in the project files, so none are claimed.

In 60 seconds

Role
Product Designer
Timeline
2026
Team
Not recorded in the project files

The challenge

A grocery order is a string of small promises: where it goes, what is in stock, what delivery costs, when it arrives. The old app broke them. Its store reviews describe a location wall, sign-in that blocks shopping and a lost cart.

The insight

Certainty before decoration: address, stock, quantity, free delivery, fee and order status belong in the main flow, not behind menus. And a shopper outside the delivery area should still be able to see the store.

What I did

Redesigned the native iOS app in English and Arabic, and designed Kitchen, a meals mode, on top of it. Three stories follow: checkout, Arabic and right-to-left, and Kitchen. The rest is folded away at the end.

125+

Screens in the app, 47 of them read one by one for this page

Design judgement · Qualitative

1,128

Unit tests passing at the last recorded full run, 24 Feb 2026, with 80.05% coverage

Code review · Quantitative

7 of 7

Tasks met a pass bar written before anyone was asked. Participant details and timings are not published

Usability test · Mixed

None

Orders, retention and ratings are not recorded in the project files, so none is claimed

Not yet tested · None

The problemWho it is for

A grocery order is a string of small promises, and each one has to be kept.

Someone buying groceries on a phone wants to know where the order is going, whether the thing is in stock, what delivery costs and when it will arrive. FarmGo sells in riyals, delivers in about 30 minutes or at a chosen time, and has to work for people who read right to left. The screens are dense because the shopper is deciding quickly; the design problem is keeping every number in them true to every other.

Certainty before decoration: where it goes, what is available, what to do next.

The design contract in the project says every screen should communicate those three things quickly. Three stories show it at work: how checkout removes doubt, how language and place shaped the product, and how a new kind of shopping fits inside it.

Story 01Checkout

Say what it costs, and what happens next, before asking for payment.

The cart is where the promotion, the free-delivery threshold and the delivery fee all have to agree. Checkout then asks in a fixed order: where, how fast, how to pay, which points or wallet to use, and what it costs.

Cart · guest
Cart · empty
Cart · promo, free gift, free-delivery bar
Checkout · payment and wallet
Checkout · pay with points
Checkout · order items
Leaving checkout
Order confirmed

One sticky bar says the total

Cart and checkout end in a bar with the total, the delivery fee or the amount saved (“₼19.95 off applied”), and a single green action. Place order and Checkout never move.

Checkout in a fixed order

Map and address, Express (30 mins) or Schedule, Pay using (cash, cards, Mada on delivery), then Wallet and Points as toggles with the amount each is worth.

A last offer before leaving

Backing out of checkout shows “20% off · Save ₼19.95. Give up the offer?” with Quit and Checkout, so abandoning is a choice the shopper sees the cost of.

01

Say what delivery costs before the basket is full

What we saw
On a listing, the shopper is deciding what to add, and the delivery terms decide whether adding is worth it.
Evidence
Design judgement· QualitativeThe design contract's first principle is certainty before decoration: address, stock, quantity, free delivery, fee and order status are shown in the main flow, not behind menus.
Prototype· QualitativeListing header reads Minimum order and Delivery fee, and a progress bar underneath reads “Add 2.35 to enjoy free delivery”; the same bar rides above the tab bar on the category screen.
So we
Terms in the header, progress toward free delivery as a bar with a sentence, and the running total with the delivery fee beside the Cart pill.
Not
Revealing the fee and the minimum at checkout.
How we'd know
Does the bar change what people add? Compare average basket for sessions that saw the bar near the threshold with those that did not. No such comparison is recorded.

02

Let people outside the delivery area look around, with ordering off

What we saw
Pinning a location outside the service region used to end at an alert with two ways out: Close, or Search another location. There was no way to see the store.
Evidence
Code review· QuantitativeThe catalogue is filtered by a backend address id sent as a request header. Coordinates alone are never sent, so saving only a map point would drop the user into an empty app.
Design judgement· QualitativeProduct-owner approved spec (7 June 2026) for a third, green button on the alert: Continue to the default location.
So we
The green button runs the real address flow with a default Al-Khobar location, so Home and every listing load. A single flag then disables add-to-cart everywhere: the + on cards, Add to Cart, the long-press menu and Add Request. Only that button sets it; any address added or selected by hand clears it, and logout resets it.
Not
A hard stop at the alert, or saving a bare coordinate with no address behind it.
How we'd know
How many out-of-region users continue rather than leave is not recorded. I would count continue-versus-close on that alert, and how often a disabled + is tapped.

Tested

You pinned a place outside the delivery area. “Keep looking around the store, then tell me why you cannot add anything.”

Finds the green button, then states the reason without prompting

5 of 5 state the reason

Passed

At checkout: “Pay part of this with your points and the rest in cash. What will you pay?”

Amount stated, and what they call the points

5 of 5 correct and consistent

Passed

Each task met a pass bar written before anyone was asked. Participant details and timings are not published on this page.

Story 02Arabic and right-to-left

A grocery app for Saudi shoppers is two scripts and one set of local rules, not a translation.

What the project files can show is how language and place shaped the product: what the research kept in both languages, which local rules set the screens, and where the words have to agree.

The old FarmGo search with tap heatmaps, Arabic after a search on the left and English by default on the right
The old search with tap heatmaps, in Arabic and English. From the FarmGo research file.
Research observation· Qualitative

Both scripts were in the research from the start

The old search was heat-mapped twice in the research file: Arabic after a search, and English by default. The Arabic results lay out right to left, and the English default's Most Searched list holds Arabic words. My reading: search has to cope with both scripts on one screen. The totals behind the heatmaps are not on the board, so no figures are claimed.

Interviews· Qualitative

Saudi rules decided what stays on a screen

The client's note on the old Settings page reads: keep VAT and CR here, because those are the rules of Saudi. The redesign carries the same local context: Mada on delivery beside cash and cards at checkout, prices in the riyal symbol, a Ramadan Picks rail on Search, and a default location in Al-Khobar. The notes are stakeholder requests, not shopper research.

Design judgement· Qualitative

One type system for two scripts

The design contract sets type from one font manager, DM Sans for Latin and IBM Plex Sans Arabic, and takes colour, spacing and radius from tokens. A screen in either language is drawn from the same parts.

Design judgement· Qualitative

The words have to match across languages

Reading the screens, the loyalty programme has two names (Imtiaz and Mokafaa), the order stepper has two word lists, and the currency has three spellings. In two languages that doubles. The change I would make: one word list per concept, localised from a single source.

What this page does not show

Right-to-left support is stated in the project files, but the redesigned screens on this page are the English app, and I have no Arabic capture of them to show. The Arabic evidence is the research above. I would add right-to-left captures of Home, Cart and Checkout next to their English twins.

Story 03Kitchen mode

A new kind of shopping, inside the same app.

FarmGo sells groceries. Kitchen is the next step I designed for it: a second mode in the same app for ready meals, ordered for delivery or pickup, with weekly plans. It is in this case study because it tests the redesign. If the grocery screens are a real system, a new kind of shopping should fit inside them with its own colour and its own cart, without a second app. These screens are design work, drawn from the app's own components. I have no record of Kitchen shipping.

Mode gate
FarmGo home · mode menu
Branch picker
Kitchen home
Kitchen search
Search · no meals found
Item · weight and nutrition
Item · add-ons and sauces
Cart · configured meals
Checkout · delivery or pickup
Orders · delivery and pickup
Tracking · on the way

Branch before browse

Delivery or Pickup, branch, ETA and an Unavailable state with a stated reason come before the first Kitchen screen, and one Continue button confirms. Prices stay out of it.

A meal with rules

Item detail shows weight, nutrition for the selected weight, add-ons with limits (“Optional · up to 3”), a required sauce group and a running total in the Add to Cart bar.

01

Kitchen is a mode of the first tab, not a sixth thing to find

What we saw
Fresh meals needed their own identity without breaking the shared cart, orders, loyalty and checkout.
Evidence
Benchmark· Qualitative31 Careem iOS captures and 6 reference captures from other food apps read for how food apps order address, ETA, categories and customisation rules.
Design judgement· QualitativeThe first tab reads Home in FarmGo mode and Kitchen in Kitchen mode. A long press opens a native menu; a title chevron and a VoiceOver action give the same switch without a gesture.
So we
A one-time popover that teaches the long press, a visible chevron as the fallback, a branch and Delivery or Pickup choice before first browse, and one shared cart.
Not
A new tab, or a Kitchen shortcut on Home that leaves the tab bar unchanged but hides the mode.
What happened
Checked: whether someone can switch to Kitchen with no hint. Result: it passed.

Tested

Kitchen: “Switch to the meals side of the app.” No hint given.

Discovers long press or the title chevron, and time

4 of 5 within 20 seconds

Passed

Each task met a pass bar written before anyone was asked. Participant details and timings are not published on this page.

The trade-off

I chose to let people outside the delivery area browse a real store with ordering switched off, over a hard stop. The cost is a state where the plus is visibly disabled, and the screen then owes the shopper a reason.

What it did

What the files can show is engineering health: 1,128 unit tests passing and 80.05% coverage in February 2026, and a performance audit with a fix plan in June. Their outcomes for shoppers, such as orders, retention or ratings, are not recorded, so I do not state them.

MoreMore product decisions

The rest of the project, if you want it.

Everything the three stories left out: the other flows, smaller decisions, the design system, AI in the product, the research trail and the things I would change. Each part is closed until you open it.

Seven more flows, screen by screenHome and booths, category and listing, search, loyalty, orders and tracking, address and account, and edge states.
Smaller decisionsA quantity control that is a plus, a stepper and a badge; clearing the last user's data on logout; haptics as a vocabulary.
The design system underneathOne contract, tokens and 55 controls.
AI in the productFour features that are live in the app, and the three rules every one of them is held to.
The rest of the usability testThe other four tasks: the quantity control, the loyalty copy, order tracking and recovering offline.
How we knew: the evidence and the research trailWhat is recorded and what is not, then the old app, the client's notes, a competitor teardown and the store reviews.
What I would change nowTen things I see on the screens or in their specs, each with what it does today and what I would do about it.
© 2026 Alhasan HosniHome