About this Case Study
This session will cover the development of an integrated search feature for a large-scale e-commerce platform.
Over the course of 8 months, our scrappy team took a broken search experience and created something that’s quick and relevant for the people using it. We will cover the highs and lows of tackling search with limited resources in an fast-growing organisation, as well as practical IA strategies for people working on different search projects.
This project required the collaboration and compromise of UX and engineering disciplines to land on a solution that worked best, while also establishing a stronger information architecture culture.
Here are some topics we’ll cover:
1. The Google effect: how other search technologies influence user behaviour, whether we like it or notPeople want Google, but our original search feature was AltaVista. The results were never what people expected. They were limited to certain topics and frequently came up with 'no results' results (an oxymoron?). So it had high drop-offs in use. Since people often expect search to suck when it isn’t Google, our users didn’t complain about it. But that didn’t stop us from advocating for a better search experience. Have you ever discovered that people were talking to your search like they were having a conversation with a person? We did. And this was another thing that influenced the restructuring of our search results.
2. The relationship between elasticsearch and keyword taggingWe had a complex problem to solve when it came to the technology. Some of our results in the product had to use elasticsearch, which makes search slower. All other results had to be hard-coded in a YAML file. These results surface quickly when a user looks for the information (yay!) but it’s very high effort to create (boo!), prone to human error, and for version 1, queries had to match exactly to our keywords. Definitely not like Google.
3. Designing the IA of search resultsIterate, iterate, iterate. Because our search results were a combination of elasticsearch and hard-coded, it took us a while to find the sweet spot of just enough information. We wanted our results to do a lot of heavy lifting: not only giving people exactly what they’re looking for, but also teaching them where to find it in the product and surfacing other related results. This wasn’t totally doable, and we found during the build stage that what works on paper doesn’t always translate to real life.
4. Finding a balance between marketing and UX, and fighting for better search experiencesThis is something we didn’t anticipate: a hard-coded search is a lot easier to manipulate. Marketing 'hackers' beware! Our stomachs dropped when the marketing team asked to surface 'discounts' in search. Our primary goal was to keep results relevant. We found a nice balance between surfacing marketing tools and resources, without turning search into an advertising platform. And surprisingly, that compromise helped build relationships across teams.
5. Harnessing the power of search data to reveal user intentionsYou might think that seeing a failed search result is a bad thing. But what people search for reveals a lot about their intentions. Tracking what information users are looking for in your product or feature can reveal what gaps you have - not only in your search results, but in your product or website overall. People using search have a goal in mind, and using search data can help identify where you’re not helping them meet their goals, and where there are opportunities to add new results. Search data also taught us that people use search to navigate, and if it works well, they’ll use it as a navigational tool (sometimes in favour of menus).
Overall, our goal with this session is to give tangible takeaways that anyone working in search can apply to their practice.
About the Speakers
Sarah Folkes is an information architect turned researcher, focusing on information systems.
Across her career she’s worked on IA, usability, taxonomy, and metadata design.
Selene Hinkley is a content strategist with a background in information architecture. She led the quick search project.