Have you tried to search on Meetup for an event to attend or a community to join? Historically, Google has done a better job at that than Meetup itself, but we have been working hard to change that. Roughly 10% of Meetup users are already seeing far better search results, and over the next few months we will be gradually rolling this out to everyone.
Search is a fascinating problem at Meetup. It has to cover interests, events, groups, time constraints and location constraints. At any given point in time, there are around 300,000 events around the world that someone could theoretically decide to attend, and over 300,000 groups to possibly join. However, because events happen in-person, there are interesting challenges around population and event density.
Our goal with search is simple. We want to help more members find events that match their interests, and thus help more organizers run thriving events and communities. Our early tests of our rebuilt search engine are showing very promising signs, with a 12% increase in event RSVPs and a 10% increase in people joining groups.
What Have We Done So Far?
First, we needed to invest in a new technical foundation for search. To enable quick and effective search improvements for our customers, we decided to integrate with ElasticSearch, a leading search and analytics engine.
After building the foundation, we focused our initial efforts on improving the relevance of event search results. It’s important to us that if our organizers are going through the effort of scheduling events, any interested members can easily find them when browsing on Meetup.
At a more nuanced level, we set a goal to be able to quickly and accurately parse and match customer search queries to the hundreds of thousands of events organizers have scheduled, ultimately returning the most relevant events for our customers.
To accomplish this, we leveraged ElasticSearch to:
- Break down customer search queries to understand the intent of the search keywords
- Analyze, standardize, and store the primary keywords found in event and group titles and descriptions
- Assign relevance weighting to the matches between search queries and event and group information
- Use location and time data to rank and filter events that make sense for the customer
- Display the most relevant events, with key event details, back to the customer searching
- Store and analyze search and performance data to measure the effectiveness of the search experience
To understand how this new ElasticSearch-based experience performed compared to our existing search experience, we launched an A/B test on our new logged-out homepage (another ongoing experiment). After four weeks, we saw clear signs that the new search model is meaningfully better at helping members discover relevant events.
In our work, we learned that the most commonly searched for keywords are topic based queries like “hiking,” “board games,” and “data science.” However, because Meetup is so diverse — a fact we’re extremely proud of — search has a very long-tail distribution of search queries, with the top 1,000 queries accounting for only 50% of the total.
With a better search engine, we’ve been getting even better signals of where the supply and demand sides of our marketplace are mismatched. We can use this information to better set member expectations and to help organizers spot opportunities for community growth.
What’s Next?
We’re excited to incorporate our initial learnings, in addition to ongoing feedback from our members and organizers, to deliver more search improvements in the coming months. In Q1, we will continue to improve the search relevance algorithm, add exposure to the new search experience on desktop and mobile web, and bring the improved experience to Android and iOS. In Q2, we will improve the group search experience, and add more personalization and filtering capabilities, and increase exposure of the new search experience.
Last modified on June 23, 2021
