Grab · Metrics · Mid · 45 to 60 minutes

Cut post allocation ride cancellations

Grab is a Southeast Asian app that connects passengers with drivers for ride hailing and other services. Answer this product management case and get it graded line by line against a weighted rubric.

Metric design
Experimentation
Product judgment
Marketplace thinking
Structured communication

The situation

You are a PM on Grab's rides team. The Head of Rides has asked you to tackle cancellations that happen after a driver has already been allocated to a passenger but before the trip starts. These cases are especially painful because both sides had high intent, time was already spent, and neither side gets value when the trip falls through. Service design research says some cancellations are unavoidable, but there may be avoidable ones tied to uncertainty during the wait before pickup. Several driver-partners have said that when passengers hear from them on the way, cancellations seem less likely, but this is only anecdotal so far. Your job is to decide how to approach the problem as a product and metrics question, not just ship a guess. The challenge is that Grab operates across different cities and rider behaviors, and any change must help passengers without creating extra burden for drivers or masking deeper supply issues. Leadership wants a clear experiment plan for the next planning cycle.

What to produce

Write a 1 to 2 page product experiment brief.

Your answer must

Define the problem precisely and show how you would isolate the post allocation cancellation stage from other cancellation types.
State your core hypothesis and describe 2 to 3 candidate interventions related to communication or the waiting experience after driver allocation.
Specify a metric framework with one primary success metric, at least 3 supporting metrics, and at least 2 guardrail metrics that cover both passengers and driver-partners.
Design an experiment plan that explains unit of analysis, control versus treatment, segmentation by city or ride context, and how you would separate timing effects from message content effects.
Explain how you would interpret possible outcomes, including what you would do if engagement rises but cancellations do not, or if passenger results improve while driver outcomes worsen.

Constraints

You can only change the in app experience after driver allocation and before pickup.
The first test must be lightweight and should not depend on new driver training, call center operations, or phone or SMS incentives.
The solution must be workable across multiple cities, with room for localization if the data supports it.
Do not assume every cancellation is bad, your plan should account for unavoidable cancellations and avoid pushing users into low quality trips.

Format

Use sections for problem definition, hypotheses, metrics, experiment design, risks, and decision rules.

How Grab actually played it

Grab looked at bookings that were cancelled after a driver was already allocated, a costly failure for riders and drivers. The team designed experiments around automated chat messages, testing wording, tone, emojis and the delay before the message is sent. A PM can learn how to turn a vague pain point into a testable set of variants.

Isolate the moment of failure (post allocation cancels) before designing a fix.
Test copy, timing and tone as separate variables instead of shipping one guess.
Cancellation cost hits both sides of a marketplace, so measure both.

Built from How Grab Experimented with Chat to Drive Down Booking Cancellations by Grab Engineering.