Things working in Glassdoor Email Operations taught me

Glassdoor is a website where current and former employees anonymously review companies. Glassdoor also allows users to anonymously submit and view salaries as well as search and apply for jobs on its platform. Glassdoor is known for its consistent email campaigns and robust email automations, which are created and implemented on Responsys.

Identifying the obstacle:

At the time, there were six core automations running on top of Glassdoor’s weekly newsletters. One automation campaign in particular, the Know Your Worth campaign, had been noticeably underperforming for a while. The campaign was not reaching its targeted audience due to the conditions for the campaign not being met as often as anticipated, resulting in poor engagement rates.

With low engagement and lack of real volume for the Know Your Worth campaign, there was certainly a gap to fill.

Hypothesis Strategy:

Although it may seem simple enough to identify a new automation campaign, there were some key questions we asked and though about.

• Who is doing Email marketing well?

• What can we learn from these successful campaigns?

• How would this new automation exist with the other campaigns?

After doing industry research and analyzing email inboxes, we realized how e-commerce companies adopted the abandoned cart emails, which are automated campaigns that trigger once a user adds an item to an e-commerce shopping cart but does not move forward with the purchase.

The Apply Now automation campaign would work similarly to abandoned cart emails in which the user would be sent a campaign with specific job descriptions they were interested in and clicked on.

Implementation of Strategy:

Step one: Data Feasibility 

Once we identified the need to fill the gap in our email operations, we moved forward working closely with our engineering team on feasibility, data integrations, and delivery expectations. In order for the automation to work correctly, a user’s job search activity needed to be captured on the Glassdoor website. For example, if a user clicked on a job description, and up to three different job descriptions, that activity would be stored and would later trigger the automation email to be sent.

Step two: To the drawing board

We worked with the email front-end engineer to go over the functionality and design the email. The goal was to not boil the ocean, but to make a simple and engaging email that would be clicked on often. In addition to including the job descriptions the user clicked on, the email would also include other job descriptions that were similar to the user’s job activity.

Step three: Building and testing the Email

Once the email was coded, we imported the code to Responsys. Over a course of a few weeks, we then ran multiple tests to small sample sizes, including A/B testing subject lines, job descriptions, and company names. We also tested the frequency of the automations. Sending too many emails would result in diminishing returns as users would eventually ignore the email.

Step four: Start of great email marketing

Once we were comfortable with the test results, we decided to push the Apply Now automation live. The automation would be triggered the same day after a user clicks on a job description. 

Results:

The original Know Your Worth campaign had a volume, Open and click rate of 32,000, 19% and 3% respectively.

The Apply Now campaign became a success and quickly became a core automation.

The automation sent about 300,000 emails a week and had open rates of 24% and click rates of 5%, which blew the KYW campaign out of the water. Over the course of the first quarter, we saw a 5% increase in job applications.

Concluding Findings:

We used a very methodical approach in coming up with the Apply Now automation campaign, along with observing industry standards as well as working closely and communicating with different stakeholders.

Take as much time needed to test everything and always be observing ways to improve slightly. It was important to not try to boil the ocean, meaning to not find the solution right away. Using multiple iterations was the way to go.