Adding AI to Search
Designing with data and documenting design
Senior Product designer (Team lead)
Charter: Discovery - Search
with
Somanshu: Associate Product Designer
This started as an idea by the product managers for a company hackathon. Somanshu an associate in my team and I worked on this together. The goal was to bring Ai into search. The initial idea the PM wanted to peruse was to have a conversational tone to the search, we ideated on the same and most of the ideas landed on a chat experience, we were clear this wouldn’t work for our audience. as our research showed that people had a tough time spelling things right and preferred not to read. As the concept was already sold we reached a compromise to give users a tappable prompt which they could search with.
There were parts of the experience I was not okay with such as the banner to enter AI search. I wanted the user to launch directly into AI search and have an opt out option rather than needing them to opt in. Or in case a opt in was required at least lead with a toggle to switch on the AI mode. Both ideas were shot down in favour of the banner to save Development time during the hackathon.
Once the hackathon was over we conducted a user research using the version we had built to get feedback to improve the version further. we learn that
-
Users missed the entry point for the search ai search
-
On the AI search page people did not read the entire prompt but fixed on keywords in the prompt to guess what was being suggested
-
The categories of prompts were confused for genres of content but were appreciated and people wanted to see more of it.
We collated internal feedback to come up with the following changes that were expected
-
Combine the AI and general search into one experience
-
Extend search to other types of content
-
Create a new rail or content display format that can encourage users to watch content
-
Create Advertising and branding opportunities
-
Create a distinct visual experience for search
-
Reduce latency for results
As the first step we mapped out the flow we were aiming to build and looked at the adjacent data for the same. we knew 12% of the users used search and 75% of the users who searched ended up watching a piece of content through search. The ask was to improve the success rate of search. I voiced my views on the strategy, in my view if the search was performing successfully majority of the time we should consider improving top of the funnel to get more users to search. I voiced it to the head of design, but in vain as the success metric was locked at Search CTR it was decided we would focus on the bottom of the funnel.
The PM wanted us to deliver one screen at a time, they scoped the timeline limited to the internal feedback, to optimise for development bandwidth. I decided to work on the entire search flow instead as that allowed us to have a holistic view on the user experience. I worked on the structure of the search landing page and Somanshu took over the detailing of that page while I worked on the rest of the flow. This allowed us to meet the timelines the PMs and Dev wanted while not compromising on the user narative.
I collaborated with the data analyst in the team I found that they had maintained a list of user issues they were trying to get prioritised into the road map for a long time. I incorporated their suggestions into screens we built, as well as the user feedback our research team had found. I was able to showcase a solution to the team and relate it back to core user issues we already had found. The new experience was widely appreciated and the result screen from this presentation was incorporated into the road map.
Impact
-
Parts of the presentation even though didn’t make it to the road map immediately inspired a host of designs going forward such as the language rail, got translated into the new selection language rail, and later the same visuals were incorporated into the language packs going forward.
-
Enabled a long standing user insights to the roadmap
-
Launched a new Search experience
-
Our search was able to increase CTR from 79% to 81%