Upcoming event October 14: Activate, the Search and AI Conference REGISTER NOW

Close Message

Customer Stories

3x
Increase in Documents Served
20M
Queries Per Day
P95
Under 400 Milliseconds
Woman on computer

The Challenge

An outdated Solr search infrastructure couldn’t scale to a 3x increase in job postings in Talroo’s inventory while reaching its goal to have a P95 of under 400 milliseconds. With a small team, Talroo needed a partner to help modernize its search platform and capabilities.

Our Plan

By automating indexing and maintenance jobs and making updates to indexing and query pipelines simple—without the need for code changes, full deployments, or developers from the infrastructure team—Talroo could meet SLAs with its partners, and exceed user expectations.

The Results

Partnering with Lucidworks, Talroo was able to deploy its new search experience seven months faster than DIY. During peak times when requests reach 4,200 per second, latency goals are met and regularly surpassed. Triple the documents added no additional latency to the system.

‹ ALL CUSTOMERS

Modernizing the Job Hunt with Talroo

Employers match with the best candidates thanks to an AI and data lead job search platform

Talroo is a data-driven talent attraction platform that uses AI to connect job seekers with employers. An answer to businesses’ needs to more efficiently recruit top talent, Talroo believes quality hires happen when a candidate’s qualifications and passion intersect with job requirements. By using data and AI, Talroo helps employers reach the right candidate, at the right place, at the right time online.

Behind the scenes, Talroo functions more as a big-data, adtech platform than a traditional employment board. “We have owned-and-operated properties like jobs2careers.com, and in addition, publishers rely on Talroo to power their job search plumbing just like merchants rely on Amazon,” explained Andrew Kidder, Principal Software Engineer at Talroo. “We have to have sub-second response times in order to provide ads to those sites so that they meet their SLAs. It’s crucial for us to respond in a timely fashion.”

Jobs need to be advertised until a designated KPI has been reached, and then a posting needs to be removed. This means indexing can’t happen once or twice a day, it has to happen continuously.

Loads Increase

When Talroo’s advertised jobs were set to triple, it knew its current Solr search architecture would struggle to scale to the increased loads. Talroo needed an updated infrastructure that could support peaks of 4,200 requests per second while providing a satisfactory response time. The goal was latency under 400 milliseconds for 95% of users. And of course, results had to be highly relevant.

All this had to be achieved while indexing roughly 146 additional, new advertised jobs per second and serving over 20 million queries a day. There’s no such thing as a quiet period for Talroo.

Partnership Enhances Strength and Experience

With a small internal team and an aggressive roadmap, Talroo knew there was an opportunity to run faster by partnering with a market leader. “My team has fingers in several different pies,” said Kidder, “and search is a major one, but it’s one of many.” Lucidworks’ history of Solr expertise and its AI powered search tool, Fusion, were a good match for Talroo’s business model.

“For me, it is important to have partners that see our success as their success. Throughout our relationship with Lucidworks, we have experienced their commitment to ensuring Talroo’s success in this very complex upgrade,” said Tony Mills, Talroo’s VP of Technology.

The benefits of the Talroo-Lucidworks partnership were quickly quantifiable. Mills estimated that their new search experience was in production seven months faster than if they had attempted it on their own. Since revenue is directly correlated with the number of job listings Talroo is able to handle, seven additional months of handling three times the jobs meant great gains for the company.

With Fusion in place, Talroo was able to provide quick, relevant search results that kept job-seekers on the site longer, clicking on more jobs, and applying, ultimately, to the right position.

Triple The Job Postings, No Added Latency

Benefits stretched beyond the quick implementation of Talroo’s upgraded search experience. Fusion’s microservices architecture makes the system flexible and easy to maintain. Changes to indexing and query pipelines are managed effortlessly, essentially by filling out forms that can immediately go into production. For Talroo, this meant no more struggling with code changes or full deployments, no disruption to the flow of the software, and no need to involve developers from the infrastructure team. Far less employee time is required for search maintenance.

“We’re spending less time working on the operations of search and more time working on the features that utilize that search to help us increase revenue,” said Kidder. “The way that Fusion is set up allows anybody from our business analysts to our developers to go in and look at things, make changes, facet on things, find more information to make good decisions, and to do better search analysis.”

Analytics tools built into Fusion provide insights that improve operational efficiency by making troubleshooting a cinch for Talroo. Dashboards allow for easy monitoring so issues are quick to surface and quick to resolve since problems can be approached knowing exactly where changes need to be made.

Indexing, which happens constantly, has now been automated. Maintenance and clean-up jobs are automated as well, using out-of-the-box features that required no additional coding on Talroo’s part. This keeps Talroo’s index devoid of extra junk it doesn’t need, and doesn’t require any extraneous processing time or infrastructure management.

“With the increased job count—hosting three times as many documents as we were previously—we’re seeing an increase in our request counts, somewhere in the neighborhood of 3000 requests per second on average,” shared Kidder. “For response times, our goal was to have a P95 of under 400 milliseconds, and we are routinely surpassing that by quite a bit. Those metrics have been the key indicators that the pipelines, gateways, and extra infrastructure didn’t inject any additional latency into our stream.”

A lot of the customizations Talroo had built for its old Solr search environment weren’t needed anymore, facilitating a cleaner system that was much easier to manage. But ML models Talroo had developed for relevancy and wanted to continue using could be plugged into Fusion with limited changes.

The Search Frontier

All of these advantages put Talroo on the bleeding edge of what’s possible in search. “As we grow, which we are doing consistently, and as we scale up within the marketplace, Lucidworks is going to be there to make sure that they are meeting the objectives we set forth early on with our relationship, which were the latency requirements, relevancy, speed, and uptime,” said Mills. “And that partnership means that I have a lot less worry about a year from now, and can focus our engineers on developing and enhancing our data-driven product suites. If we’ve doubled our throughputs again and if we’ve doubled the amount of documents that we’re holding, I am positive that the system, a Lucidworks system, will be able to handle that, and make us very, very powerful in the marketplace.”

“Before Lucidworks, we physically could not handle the number of documents that we’re handling now. And Lucidworks has allowed us to do that at the same levels as when we were serving a third of those documents. That is a pretty impressive indicator of ROI,” said Kidder.

Mills added, “Partnering with Lucidworks has been one of the most forward-thinking things we’ve done.”

“Partnering with Lucidworks is one of the most forward thinking things we’ve done.”

-Tony Mills, VP of Technology, Talroo

Play Video

Explore More Customer Stories