ZomatoDiscovery5 min read

Choosing where to eat, not what to order

A discovery surface holding reviews, photos and delivery in one place.

The surface: The discovery side of the app: search and cuisine collections, the restaurant page with its ratings and photographs, the review section, and the switch between dining out and delivery.
What the user wants: I do not know what I want. Show me something I will not regret.

Collections instead of a search box

The home surface leads with curated and algorithmic collections such as cuisines, occasions and trending places, rather than requiring a typed query.

Recognition over recall

A user with undefined intent cannot type a good query, because they do not yet know the answer. Offering browsable groupings lets them discover the shape of their own preference by reacting. The collection is doing the job a friend does when they list three options.

The dual mode

The app separates dining out from delivery, with different ranking, content and offers behind each.

One surface, two jobs

Choosing a restaurant to sit in and choosing food to arrive at home are different decisions with different criteria, and ambience matters for one and packaging for the other. Splitting the mode lets each be ranked honestly. Merging them would have produced a compromise that served neither.

The rating with its count

Restaurant ratings are shown alongside the number of ratings, and separate ratings exist for delivery and dining.

social proof: People look at what others did to decide what is correct here. with a confidence signal

A rating without a count is close to meaningless, since a perfect score from three people says little. Presenting the two together lets the user weight the signal themselves. Separate delivery and dining scores stop a great kitchen with poor packaging from being unreadable.

User photographs

Restaurant pages carry large volumes of user submitted photographs of dishes and interiors alongside official images.

Unmediated evidence

A styled menu photograph tells the user what the restaurant wants to be, and a user photograph tells them what arrives. The visual gap is the most efficient quality signal available in this category. It works because it is hard for the restaurant to control.

Reviews written for other diners

Reviews carry reviewer profiles with follower counts and review histories, which makes some reviewers into recognisable local voices.

Identity raises contribution quality

Anonymous reviews are cheap to write and cheap to fake, while a named reputation gives the reviewer something to lose. Building status into the reviewer profile buys better content without paying for it. The same status mechanism is what makes the reviews worth reading to strangers.

A rating everybody scores four on

A schematic distribution. When almost every restaurant lands in the same band, the number is still being computed, still being displayed, and no longer separating anything. A discovery product is judged on whether its signal has spread, not on how the grid looks.

What not to copy

Ratings on Indian delivery platforms cluster in a narrow high band, which compresses the signal to the point of uselessness. When almost every option shows a four, the number has stopped being a filter and become decoration.
Paid placement and promoted listings sit inside the same ranked lists as organic results, with disclosure that is easy to miss. A discovery surface that sells position is not a recommendation, and users read it as one.
Review authenticity is a persistent and unsolved problem across this category, with incentivised and coordinated reviews difficult for any user to detect. The platform's interface asserts more confidence in the score than the underlying data supports.
Restaurants have repeatedly disputed rating changes, commission rates and the visibility they get from the platform. A discovery product whose supply side distrusts the ranking has a structural problem that better UI cannot fix.

The takeaway

Discovery products are judged on the quality of the signal, not the beauty of the grid. A rating everyone scores four on is a decoration you are paying to compute.

Finished the teardown? Bank it and the day counts toward your run.

Where the principles come from

Influence: The Psychology of Persuasion, Robert B. Cialdini
Recognition Over Recall, Nielsen Norman Group
Predictably Irrational, Dan Ariely

Written from public behaviour of the product, not from inside it. Interfaces change often, so treat the flow described here as of the time of writing and check the live product before quoting it.