Jobfeed, Textkernel’s labor market intelligence tool, is renowned for its accurate and up-to-date market insights, which can help recruiters make critical talent decisions. It’s not all about optimizing the hiring process, however. Jobfeed data has also been leveraged in the datasets of several research and policymaker institutions in the Netherlands and across the EU.
At Textkernel, we pride ourselves on developing products that deliver and maintain exceptional accuracy levels. We recognize that Jobfeed data has a critical role to play in helping businesses, educational institutes, non-profits, and government organizations make important decisions.
That’s why we constantly monitor Jobfeed data. In recent months, we noticed a disparity in our reports in comparison to independent economic indicators. When we started to look into why the numbers were off, we discovered a few interesting job posting trends, which we’ll discuss below.
A steeper uptrend than expected
Over the past few years, Jobfeed has reported a significant increase in the total number of job openings. That’s not surprising, as labor shortages have made it difficult for recruiters to fill many positions.
However, the uptrend reported by Jobfeed seemed too steep when compared to independent economic indicators. While investigating this difference, we realized that the tight labor market had led to a few changes in job posting behavior that were impacting the Jobfeed data, including:
- Catching candidates’ eyes. Lately, recruiters have had to find creative ways to get their vacancies noticed by candidates. These strategies include making significant changes to job postings so they appear different, as well as posting multiple vacancies for the same job with different job titles and locations.
- New aggregators. Since the pandemic, we’ve also noticed a number of new online job boards appear, most of them aggregators. Aggregators crawl the web and collect job ads from career pages, often posting a shortened or altered (and sometimes almost unreadable) version of the job ad on their website. Those altered copies are sometimes reposted on other aggregators, resulting in a chain of low-quality copies of jobs.
Both of these developments have led to an increase in the number of job posts identified by Jobfeed. The rise of new aggregators in particular has contributed to this trend, since the many changes aggregators make to job posts when copying them can make them difficult to deduplicate.
Currently, the smart algorithm we use to deduplicate job posts is identifying on average 5 posts for each unique job in the Netherlands. After examining the new market conditions that arose after Covid-19, however, we realized that the algorithm was sometimes failing to take into account these new tactics.
Textkernel’s approach
In order for recruiters to implement the best hiring practices, and for research and policymaker institutions to accurately understand employers’ future demand for skills, it’s critical that Jobfeed data reflects the actual job market as closely as possible. That’s why we’re working to take these market changes into account and are in the process of normalizing past labor market data by removing these duplicate job postings.
Of course, this needs to be done carefully to avoid deduplicating too many posts. To make sure we achieve this, we’ll undertake the following actions:
- Remove all jobs from specific aggregators that don’t contribute to the completeness and quality of the data in Jobfeed.
- Identify and remove low-quality content from job boards and other websites, focusing on shortened and/or unreadable text.
- Improve handling of single jobs that are posted for multiple locations on job boards.
The first action will account for most of the corrections we plan to make. We’ll update all data from 2019 on, and in order to provide immediate improvements, we’ll also remove specific aggregators starting the week of August 15th. The estimated reduction in the number of unique jobs is between 8% in 2019 and 23% in 2022.
While we’re currently investigating the best approach for the remaining two actions, we expect their impact on the normalization of the data to be smaller, since aggregators appear to be causing the bulk of the issues.
Once the improvements are made, they’ll automatically become visible in the Jobfeed Portal and API, giving you immediate access to the most up to date data. If you have a data feed you’re using and would like to receive a refresh of the data, please contact your account manager for help.
Please feel free to reach out to marketing@textkernel.nl if you have any questions. Accuracy is the cornerstone of our service, and we remain committed to delivering the best solutions and improving our data every way we can.
