How Grab brings new products to life through experimentation

Experiments to improve user experience

One type of experiment Grab runs routinely is improvements to user interface and user experience on the app.

The GrabFood homepage revamp is an example of that. It was an experiment that saw quantifiable results for both users and merchants.

Looking at the conversion and bounce rates of users that visit the GrabFood homepage, the team made some hypotheses about the design of the homepage that could increase conversion. Conversion means users actually follow through with an order, instead of “bouncing”, or navigating away from the page.

The team ran experiments to test out how the position of tiles and carousels affects conversion. Then, it also tested the effectiveness of certain changes in the way merchant information is displayed.

From these experiments, the team learned a lot about how to improve the GrabFood homepage. One change that led to measurably improved conversation was placing the “Order Again” button high up, where it is seen by users without scrolling further down the page.

Not only did this make it more convenient for users to re-order a dish they already had before—it also led to an increase in purchases, which benefits merchants.

Experiments in the real word

Experiments at Grab don’t stop at optimising the design and user flow. Experiments can also dramatically change the user experience in the real world.

Take the creative feature Quiet Ride, which lets passengers discreetly indicate that they prefer driver-partners to not make conversation with them. This idea also began as an experiment in some countries before launching in more markets. Passengers who participated in the trial were presented with the option to ask for a conversation-free ride in the chat interface that opens on the app after a booking is confirmed. And it proved popular.

In Malaysia, for example, passengers who chose Quiet Ride were more likely to give 5-star ratings than those not on Quiet Rides. Such results encouraged the team to move the feature forward, and make it more user-friendly in its final form. This feature has now rolled out in five countries, and exists as a simple toggle button a passenger can select after booking a ride. 

Experiments can fail

Not all experiments lead to new features. They’re not meant to. One important function of experimentation is to quickly sieve out ideas that look good on the drawing board, but don’t work in the real world.

AI-generated food images started as the idea to use AI photos of dishes for merchant-partners who did not have resources to take professional food photos to illustrate their digital menus.

Grab ran a pilot with a small pool of merchant-partners, making sure to label photos clearly if they were generated by AI. This feature was not well-received, generating negative feedback by users online, complaining that the photos did not represent the actual product accurately. 

While the intention was to help merchants make their menus more enticing, the use of AI-generated images in this instance clearly didn’t resonate with users. Though this experiment failed, it validated the ideas behind experimentation with small groups of users. The negative feedback was important in helping Grab’s product teams understand what works and what doesn’t, helping them grow as they embark on their next product development journey.

With more experiments coming up, expect more out-of-the box features at Grab in the near future.

Innovation isn’t easy. New features can affect already existing ones in unexpected ways. Or a new product may turn out less useful than initially assumed. It’s an opportunity to learn and iterate quickly to deliver a better experience.