In February, everyone’s fancy turns to matters of the heart. And for many soul mate-seekers, that means logging in to an online dating site like Zoosk. With 35 million visible members, Zoosk’s Behavioral Matchmaking™ engine matches people using the company’s own data and algorithms.
Zoosk also relies on New Relic APM to track its systems’ performance, during recent important holidays like Valentine’s Day as well as all year long. According to Brian Backhaus, Zoosk’s VP of Engineering Applications, “We track pretty much every possible metric as you go through the site. We use that to tailor the optimal experience and hopefully help you find someone that you care about.”
New Relic helps Zoosk make sure poor performance doesn’t degrade the user experience, keeping the MTTF (Mean Time to Flirt) as low as possible. “We’re a dating site,” explained Doug Wehmeier, Zoosk’s Vice President of Platform and Infrastructure. “We’re always looking for better ways to increase the speed with which we can match people with other members they’re compatible with.”
One aspect of performance Zoosk monitors particularly closely is how quickly the server responds to a request once a user loads a page and can send a message. “We built a dashboard with New Relic showing how long it took us from the time we got the request,” said Backhaus. “We inline all of our statics on our mobile website. We track how long it takes on the server to inline the JavaScript, the HTML, and the CSS—and we made some improvements based on that. It was just one part of the Mean Time to Flirt, but it’s an area where we used New Relic to drive our performance forward.”
Test and test again
New Relic also aligns with Zoosk’s extensive testing regimen. “At any given time,” Backhaus said, “we have between 20 and 50 different A/B tests going on across Android, iOS, Web products, anything like that. We got to the point where developers were spending so much time just flipping flags on the code that we needed something a little better. So we built a system we call Mission Control that lets the product managers really control what sets to distribute our users into.”
Zoosk uses New Relic to monitor whether testing a change in its search algorithm, for example, affects performance, which could influence the test results. If a test isn’t performing properly, “it could definitely be because we’ve increased the response time by 200 milliseconds,” noted Backhaus. That performance hit wouldn’t show up in the Mission Control interface right away, but the New Relic monitoring would reveal it.
It’s not always easy to draw a direct connection between performance optimization and increased user activity. “We can’t always say definitively ‘this performance increase caused this influx in flirts,’” said Wehmeier. At the same time, recalled Backhaus, “We’ve seen cases where we launched a feature based purely on performance optimizations and interactions jump up a bit. We definitely use New Relic to watch those types of things.”
On the other hand, poor system performance can definitely result in lower activity. “We have had instances where we’ve made some changes in our network routing layer that added a certain percentage of latency or overhead to the request,” said Wehmeier. “We could see the effects through our entire ecosystem, on the amount of messages, likes on our carousel experience, and other stuff like that. We’ve definitely seen engagement drop off from slowdowns, and increase when we’ve made things faster.”
All about engagement
For a dating site like Zoosk, engagement is the name of the game. New Relic helps Zoosk promote engagement among its users—and hopefully, keep producing engagements.
Beach image courtesy of Shutterstock.com. Zoosk profile image courtesy of Zoosk.