Unified Search and Discovery
In LBB we always strive to provide best of user experience to explore and shop various brands and retailers. It is a platform to connect local businesses with the people and this complete journey to explore and shop is done in many different ways.
With time as LBB has exponentially grown with active users, products, content and on-boarding local businesses, the platform becomes more complex. Thereby making search and discovery an integral functionality for the user as well as for the business.
This post is divided into three major parts- Product Structure, Challenges and Solution.
Product Structure
LBB has 2 primary sections that completes the platform. One section is Place(home page), where user can discover new places, cafes, local retailers, events or content from multiple categories. Other section is Shop(e-commerce), where user can buy products online. Each section has it’s separate list of categories and listing.
Challenges
The two sections mentioned above had their respective search bar, thereby creating a confusion in user to decide where to search. Adding to this there were multiple UX units and analytics to be handled and maintained on our end.
Hence it was important for the team to revamp the search funnel i.e. moment the user starts to type till the landing on relevant result. Thus the problem was to have an impeccable search that helps user to easily discover and reach the exact item they were looking for.
Search is a way the user talks to your platform.
Solution
Unified Search
There were some key user behaviour that indicated need for a unified search:
- Queries on search bar of Place section had many queries that were related to Shop. Eg: Users were typing ‘Casual Shoes’ in search bar of Place section. This was resulting in drop-offs.
- Open searches (search that is not triggered by using autocomplete suggestion) were very high as compared to the autocomplete searches, showing scope of improvement in the autocomplete suggestions.
Unified search also need a common autocomplete framework that not just handle results from both the sections but also help user easily distinguish between them.
Eg: Query like ‘Casual’ can have Place results (Casual Dining) and Shop results (Women’s Casual Shoes, Mens Casual Shoes, Women’s Casual Shirt etc.)
There were multiple prototypes that were tested within team and some of our active users. Each test aimed to make user search for a query and test how quickly and easily they can reach to the item searched.
After multiple iterations, calls and meeting up in cafes with end users our team had an autocomplete framework that works uniformly across platform assisting every query.
Now unified search was solving first part of the problem of triggering one common search across all sections, but there was a second part in the search funnel i.e. channelizing the search to the relevant landing page. This becomes important in case of open searches where we don’t know the intent of the user.
Eg: Open search like ‘Coffee’ can be triggered to see results either in Places(coffee shops, coffee makers etc.) or in Shop(Coffee printed shirts, coffee coloured dress etc.)
To handle this problem, we introduced a new page - Disambiguation Page.
Disambiguation Page
This page handled cases specifically for open searches. As user is able to search all the sections across platform, any open search takes the user to the disambiguation page.
The page helped user to channelize their intent of search to a particular section. What looks like a mere choice to select the section in app, triggers the type of search indexing(Place or Shop) in the backend.
Since search funnel is a type where users want to spend least time and land to the relevant result as soon as possible! Hence to make it further clear and fast, a glimpse of top 3 results with images are shown in the disambiguation page. Our users love images! They could access either directly the result or see the list of results related to the query search.
Better the results → Faster the user reaches the item searched → Better user experience.
Relevancy of Results
As mentioned above there are two different search indexing that we use, one for each section- Place and Shop. Indexing used depends upon the user’s type of selection in the disambiguation page.
Each of the indexes has a list of attributes and each attribute has a defined score depending on the priority. As the user searches a query, these attributes with their respective score provide a sum of score called as relevance score. We use this score to rank the results.
With all these iterations, testing and processes, unified search was introduced to refine the search funnel. As LBB further grows in conjunction with rapidly changing market, feature like search serves as an integral base to the whole ecosystem to evolve further.
Will talk more about the new and exciting features in the future posts.
ABC. Always be clappin’