Consumer UXMobileGoFoodOrder Pooling

Gojek · GoFood · June–October 2023

Redesigning the order tracking experience to reduce customer anxiety

In 2023, Gojek needed to improve profitability by reducing rewards and promotion costs while keeping food delivery affordable for Indonesian customers. To support this, order pooling was introduced but faced low customer adoption. As Design Lead for the consumer booking experience, I built and led a team to redesign the order tracking experience, reducing customer anxiety and enabling the scalability of this new logistics model.

My Role

Product Design Lead

Team

GoFood · cross-functional

Duration

5 months (June–Oct 2023)

Platform

Mobile (Android + iOS)

01
The Situation

Business shipped a new delivery model without touching the UX.

In 2023, Gojek was under pressure to become profitable. For GoFood, the lever was clear: lower the delivery fee. Order pooling — “Mode Hemat” — was introduced as the solution. Customers accept a slightly higher ETA; eligible orders get assigned to a single driver; logistics savings get passed on as subsidies.

The business ran an experiment. Orders started pooling. And the tracking screen that millions of users stared at while waiting for their food stayed exactly as it was — a map-first UI designed for regular, single-order deliveries.

It didn't go well.

A good logistics model had a UX problem. And that UX problem had a measurable, growing cost.

What broke

Order delay & driver complaint tickets

For pooled orders vs. regular orders on identical routes

~60%

ETA compliance for second deliveries

Down from baseline — eroding the one promise users were relying on

$177k

Potential revenue loss per month

As pooling was scaled back 40% to contain the complaint surge

How Mode Hemat worked
08:08▲◆▲
Ayam Geprek, Kemang
🛒
DeliveryChange
RegularRp11,000
30 min
Mode HematFree
46 min· Stays fresh
Delivery toRumah
Order with Mode Hemat

✓ You'll save Rp20,000 on this order

Checkout

Customer accepts higher ETA

Order eligible for pooling

08:08▲◆▲
On time · Delivery in 15 mins

Good food is coming

Driver is on the way to you

Galih Pambudi

B 1060 JEK

First delivery

Pooled order

08:08▲◆▲
On time · Delivery in 15 mins
First delivery×

Driver is delivering to another location first. Don't worry, it's not far from you. They ordered from the same resto as you.

Good food is coming

Driver is on the way to you

Galih Pambudi

B 1060 JEK

Second delivery

Pooled order

02
The Diagnosis

Three structural problems in the existing tracking screen.

The tracking screen wasn't designed for pooling — but it had deeper structural issues that would have broken any high-ETA, multi-state delivery model. We audited the current experience and found three root causes.

01

Map = center of attention

The design forced users to focus on driver movements on the map — not order states, not ETA. If the driver paused, changed route, or GPS lagged, users read it as 'something went wrong' and raised a ticket. Pooling made this worse: when drivers were unassigned or detouring to a first delivery, anxiety spiked.

02

ETA = buried in hierarchy

ETA information sat at the lowest level of visual hierarchy on the screen. With historically poor ETA compliance (now made worse by pooling's longer ETAs), users had learned not to trust the number — and there was no stronger signal to anchor to.

03

Linear order lifecycle

The state model assumed a fixed sequence: restaurant confirms → driver assigned → food picked up → delivered. Pooling broke this. Driver assignment and food prep could happen out of order, but the UI had no way to represent it — leaving users with unexplained states that looked like failures.

Current experience audit

Annotated existing tracking screen

Three callouts on the live UI: map dominance, ETA visual hierarchy, and the linear state model — showing exactly where the experience broke under the pooling model

HMW

How might we balance system status transparency with customer anxiety?

HMW

How might we design to change the mental model of customers to adapt the new logistics model?

How we measured customer anxiety

Time spent

on the order tracking screen

Proxy for anxiety — longer dwell time means more checking, less confidence in the system.

👆

Screen open rate

of the order tracking screen

How often users actively re-opened the screen mid-delivery — a signal of uncertainty.

🎫

Tickets raised

for ongoing orders

Support contacts during the active delivery window — the clearest signal of UX failure.

03
The Experiments

We didn't start with the final design.

Two targeted experiments ran before the full redesign — each with a specific hypothesis, a measurable outcome, and a deliberate learning that shaped what came next.

June

Exp 01

Turn off pooling pins & remove pooling nudges

July

Exp 02

Remove map from order tracking for all delivery orders

October

Final Design

Improve order status & ETA prominence; Revamp info cards

Experiment 01

Are we overcommunicating pooling?

June · No stat-sig result
Hypothesis

Pooled orders were perceived as more likely to be delayed. Reducing the visual prominence of pooling information — the dual customer pin, the banner nudge — would reduce anxiety and bring focus back to the ETA promise.

Experiment
  • Remove 2nd customer pooling pin, show only 1 pin on the map

  • Remove the banner nudge that explained the pooling model

  • Make pooling look less visually distinct from a regular order

  • Surface pooling context as part of the order status copy instead

Impact

No statistically significant impact — positive or negative — on any anxiety metric. No changes to driver↔customer chat initiations.

Learning

Over-communicating about pooling doesn't add anxiety. The problem isn't the labels — it's the underlying UI. The map and the state model are the root issue, not the pooling messaging.

Experiment 02

Do we really need the map at all times?

July · All 3 variants stat-sig positive
Hypothesis

The map itself — not the pooling information — was generating anxiety by forcing users to focus on driver GPS position. Replacing it with state-specific illustrations would reduce anxious checking and tickets.

Experiment
  • Variant 1: Replace map with illustration by default; user can switch back to map

  • Variant 2: Remove map entirely with no option to switch back

  • Variant 3: Show illustration only; map revealed once driver picks up the order

Impact

All three variants stat-sig positive. Variant 3 performed best: avg screen time −4.68%, tickets raised −6.89%.

Learning

The map is only useful in the last mile — when the driver is actually heading to the customer. Before that, it adds anxiety without adding information. Variant 3 gave users the map when it became meaningful.

04
The Solution

ETA-first. Illustration-driven. State-rich.

Four design principles, derived from what the experiments proved. Shipped in October 2023.

01

ETA focused

The arrival countdown becomes the dominant UI element — a large, prominent circle with time and on-time status. The map now supports the ETA, not the other way around. Users are anchored to time, not location.

02

Scalable states

New order state model that decouples food preparation, driver assignment, and delivery. Non-sequential pooling logistics are now fully representable — each state has its own distinct visual treatment and copy.

03

Delightful journey

State-specific illustrations replace the map during food preparation and pickup phases. A chef cooking, a driver flying. Contextual, emotionally resonant, animated — signalling progress without GPS dependency.

04

Cleaner UI

The three most common support queries — 'When will I get a driver?', cancellation, order edits — are surfaced as quick actions directly on the tracking screen. Reducing contact rate and friction simultaneously.

Before → After

Before

Original tracking screen

Map-dominant, ETA buried at the top, binary order states, pooling banner visible, linear lifecycle

After

Redesigned tracking screen

ETA countdown leads, state-specific illustration replaces map, simplified driver card, quick-action support surfaced inline

Key states

Chef has started cooking

Illustration phase — map hidden, ETA leads (Arrival in 30 mins · On Time), contextual copy ('Mm— if only you could smell the aroma')

Food is coming your way

Driver en route — map now revealed, ETA still leads (Arrival in 8 mins · On Time), driver card with quick-reply pre-fills

Frequently raised issues panel

Self-serve support surfaced inline: 'When will I get a driver / I have to cancel / I want to edit my order'

Driver card redesign

Before → After: Driver info card

Old: photo, name, plate, health badge, tip CTA, delivery details — all stacked. New: compact card with rating, trips, pre-filled chat quick replies

05
Outcome

The numbers that followed.

−5.8%CCU late ticketsStat-sig reduction in support contacts during the active delivery window
−3.4%Screen time spentStat-sig — less anxious checking, more confident waiting
−2.5%Screen open rateStat-sig reduction in reactive mid-delivery screen opens
What this unblocked

The redesign directly unblocked the full rollout of Mode Hemat. Pooling had been scaled back 40% while the UX was broken — the new experience gave the business confidence to scale it.

[Add: whether pooling fully scaled back up, any broader adoption of the illustration model across other GoFood order types, or team/leadership recognition.]