Row vs Column-Oriented Databases | SentinelOne

Scalying New Heights: Moving the Needle on Search at Glassdoor

Bhawna Singh is the CTO of Glassdoor and also one of the founding members of ENG, the peer network of VPEs, and CTOs from leading SaaS companies. Bhawna started with Glassdoor five years ago, and she was promoted to CTO in mid 2020. She started at Glassdoor with a dual mandate to rethink and rebuild the jobs search business, and also to open the new remote office in San Francisco with engineering and product presence. It was an opportunity to impact not just Glassdoor but all of its users and their careers, and there were a few moments of fear as well as elation. 

Glassdoor’s mission is to help people everywhere find a job and company they love, by powering workplace transparency to help people make informed decisions. They shine light in professional workplaces because bad things happen in the dark. When Bhawna joined, her first job was to set up a new office in San Francisco with engineering and product presence, to tap into a broader talent pool. The site would be a Center of Excellence, with its initial focus on talent to rebuild the search vertical for jobs business. “Our users were not engaging with our jobs vertical, and we knew that if we turned it around, we’d have another strong revenue stream for the company. I was a new leader who was going to move a few teams from our Mill Valley office to start the remote San Francisco office and attracting talent was the first course of action.”

The purpose of the new talent was to improve search, but the biggest question was where to start. 

We had to start measuring search.

“This was a big missing piece.  We knew our relevance was bad, but we didn’t have a measuring unit to say how bad. I believe if we can’t measure it, we can’t improve it, so before we could start working on improving the quality of search, we had to identify the measuring metric, the process of measuring it and building trust in that metric, because if everyone does not believe in the measurement metric, then we won’t be effective and we won’t rally behind moving the needle on it.”

They would need to figure out both offline and online tracking, all while working on the tactics of architecting the platform, the deployment infrastructure and monitoring setup. “I remember we faced a major performance challenge a few months into the effort. We got so heads down on improving the quality of the system that for a brief period, we took our eyes off system performance. So by tuning the system for quality, we made the system heavier and less performant. And by this time, we were receiving more traffic on the system.” 

As more and more users started to engage with the feature, the performance issues impeded scale. “Since we were still a very small team, we had to pull back from working on quality improvements to invest in performance. This is when we built load and performance testing features in our deployment pipeline to make sure we could compare performance of every release that went out.”

But as with most technology objectives, the hardest part wasn’t the technology. It was the people. “Moving people from a location they’ve worked at for many years, with all the amenities like free lunch, gym, and other perks, to another office with none of that, and following a new leader they didn’t know yet, can be hard. A few months after we moved to the new office, I realized that failure was not an option anymore. I was responsible for the people I had hired.”

I remember this heavy weight on my shoulders.

Bhawna immersed herself in the job and took guidance from the leaders of the company, and they were very helpful. And she knew they had made it when the team reached their relevance goal. “We celebrated big! We had reached the target relevance, putting us at parity with competitors in the market. And the users had started to engage at a healthy rate.” 

Now the team could turn their attention to revenue growth. “That was a transformational moment; the feeling that we have arrived. Our platform was modernized, we had good infrastructure around it, supporting the deployment and monitoring. We had also hired some great talent. Everything felt that it was in place.”

They also kicked off a migration of the whole web app technology to smaller applications, and a modern tech stack. “A big portion of our site grew into a monolith written in JavaScript and Java backend, and we planned to migrate the site into smaller known applications with frontend written in React and Graph QL to aggregate and load data. We identified a list of core services such as authentication, tracking, and few others. Our platform team built these core services following API first principles, which enabled the migration effort for application teams. There were multiple phases to this migration effort and at the completion of every phase, there was a significant improvement in our engineering efficiency. I would say we have come a long way, but engineering efficiency is an objective we continue to invest in, as we strive to build engineering excellence.”

On reflection, Bhawna shares a few lessons for what to do, and not do to. “Sealing the new location with few people from our existing space really worked for us. Having people who have the experience, reflect the company culture, and are driven by the mission are key to building out a new office location.”

Now, for things I would do differently….

“In an effort to improve the quality of the system, we took our eyes off the performance. It’s a hard balance to strike when you have a small team with a large charter and with lots riding on their success. I would push myself to strike a balance between infrastructure investment and innovation. In our case, we focused more on building new features to meet our quality goals, while the infrastructure improvements lagged until we were hit by performance issues. After we hit our relevancy and quality goals, we did strike a balance, and were able to invest more in infrastructure and system improvements. We also had a larger team by then.” 

“As we kicked off the migration effort of our web app layer to the new tech stack, we had the plan for early phases laid out which helped us track our progress. But the effort dragged after sometime due to lack of urgency without the end-to-end migration roadmap in place. Its key for large migration projects to have a high level roadmap in place along with communication cadence to keep respective teams up-to-date on the progress to ensure steady rallying around in getting it across the finish line.” 

“As part of migrating the web pages to the new tech stack, we sliced the migration effort into common sections of the page instead of migrating full web pages in one go. With this approach we were able to compare page performance with new and old technology faster and were able to adopt sections that tested well and narrow our focus on addressing the sections that failed.”  

That was a huge learning and something I would do differently.

About the Author

Christine Heckart is CEO of Scalyr, which provides the industry’s only Event Data Cloud for log management and event analytics to support incident management, alerting, or to power custom services and dashboards.  Available as an OEM option, Scalyr supports customized analytics and applications through APIs, and is more cost effective and performant than “free” open source. Scalyr also curates a peer-network of VPEs, CTOs, and top technical executives at leading SaaS companies called ENG (Engage, Network, Grow). To learn more about Scalyr or ENG, visit