Improving Headout's search
An improvement to the existing search feature to increase it's adoption and efficacy
Timeline
1 month
incl. Design, BE and FE
Team
3
Product, design and dev
Role
Product designer
THE PROBLEM STATEMENT
Why is it required?
As the company's focus shifted towards increasing revenue from organic users (rather than paid campaigns), it became increasingly important for us to improve Headout's search functionality.
Obvious issues with current search feature
Buggy functionality
Inaccurate search results, leading to little or no conversion
Poor visibility of the search feature for users
Lack of clarity on the different types of search results available
From a tech perspective, the search feature's code had not been updated in two years
Some nomenclature before we move ahead:
Experience page: A single product page providing all necessary details for users to book an experience. (e.g., Guided tour of the Colosseum, Roman Forum, and Palatine Hill.
Collection page: A group of all experiences associated with a single point of interest, landmark, or location. (e.g., Colosseum)
Category page: Displays all experiences belonging to a specific category. (e.g., Guided tours, Airport transfers, Cruises)
City page: Displays all experiences available in a specific city, including entry points for collections, individual experiences, and categories. (e.g., Rome)
OBJECTIVE
How do we help users find what they are looking for with the least amount of hassle?
THE RESEARCH PHASE
How did we arrive at what we want to do?
What we learnt from other products?
Showing and differentiating top products in the list
Smart default search page
What we did data tell us?
Currently, roughly XX% of users coming to Headout use the search feature, and this number decreases significantly as users move between pages:
City page > Collection page (group of all experiences from one point of interest) > Category page > Experience page.Search adoption from the collection page is only around ~4.5% of all users searching.
More users click on the search result suggestion dropdown compared to the actual search results page.
60-70% of users click on one of the search results; this applies to both desktop and mobile users.
What are people clicking on (from most to least)? Experiences, then city, and lastly collections.
Search adoption from the collection page is only around ~4.5% of all users searching.
Search adoption from collection page is only around ~4.5% of all users searching.
FIRST STEPS
Our vision for search
Search should directive, than exploratory
Users are more likely to search for a city they are visiting or the place they want to visit.
How do we make it possible to show "most relevant search out for the user"?
User doesn't know what experiences are or collections are. Does categorisation even make sense?
Taking bigger bets
Considering the lower traffic to search feature, we want to take bigger bets.
(Few examples in the end of this case study)
SOLUTION
Search adoption
Differentiation
We tweaked the placement of the search bar on home screen to give focus on it.
On scroll, we show the search bar on the top so that user doesn't have to scroll back to the top to search anything.
Changed copy to show users a glimpse of what exists on Headout and what they can search for.
Focused search everywhere
We worked to make the search suggestion dropdown more focused and for the user to be less distracted from other elements on the page.
We added search bar to all the prominent page where there is a chance of user jumping between different experiences/collections.
Mobile & app changes
Having search in the bottom nav made it harder for the user to search anything once they are already browsing through the pages inside the product.
We decoupled it from bottom and made it a part of top navigation which enables us to show search entry point anywhere.
SOLUTION
Search efficacy
Ranking of the search results
We tweaked the ranking order of search results to be 2 experiences, 2 collections and 1 city.
When the city is a direct match or close match, the city result shows up on the top because highly likely that user was looking for a city.
Experiences are first results to show because it's the path of booking with least friction
Collections were also shown to show users the number of options and encourage exploration
Ranking of the search results
For experiences, we stuck with the usual, the city of the experience, image and in future, we plan to add scarcity based boosters (like bestseller and selling out fast)
For collections, we added no of experiences available in that collection and slight visually different treatment, in addition to the city and image.
For cities, we used an icon and country name. This combined with it's ranking when users are searching for city helped us give a better experience.
Design and code improvement
Made visual improvement to have a more focused search view
Clear empty state design
The developer refactored the whole code to improve the search's performance and improved the no of people able to see proper results by 33%.
LEARNINGS
How did the revamp go for us?
All of these combined improvements led to seeing us:
Search adoption
+X%
Double digit increase
Search efficacy
+X%
Double digit increase
Direct CVR increase of
+X%
The search adoption was measured by no of people clicking on search and actually start typing.
Whereas for efficacy, it was measured by no of people clicking on search suggestions list (people being able to find exactly what they are looking for).
These changes led to an X% increase in CVR across devices.
WAY AHEAD
Ideal solution
There are bunch of optimisation and experiments we wanted to run post the first set of experiment, but we were not able to go forward with them due to prioritisation changes in the new quarter.
These is the ideal version that I would have wanted us to experiment with.
We noticed a sizeable chunk of users searching for category and subcategory. Add those to the search results as well.
Run experiments on showing all collections or all experiences, to compare between least friction results to conversion vs more exploratory results.
Add boosters to results to check for it's efficiency.
One more experiment we could have run is date level search. Something in similar lines to Airbnb and other hotel bookings.