Applicant Tracking Systems: Why do we need them and why are they still bad?
The bane of both job seekers and recruiters, a good ATS nowadays is still a bad assistant to the people it’s built to serve.
What is an Applicant Tracking System?
ATS, or Automated Tracking Systems, are computer programs or systems that are used to automate and streamline recruitment processes and employee management. ATS typically use algorithms and rules to process and analyze data, and to make decisions or take actions based on the results of this analysis.
In recruitment, an ATS might be used to automatically sort and filter job applications, to identify the most qualified candidates for a given position, and to schedule interviews with these candidates. This can save HR departments a significant amount of time and effort that would otherwise be spent manually reviewing and evaluating job applications. You must have read about the average time recruiters look at a CV, typically 7 seconds. That 7 seconds are usually aided by the ATS to make it more efficient to comb out the most suitable candidates for the jobs.
Why are they so popular now?
There are several potential advantages to using an ATS in the recruitment process. Some of these advantages include:
- Improved efficiency: ATS can help streamline the recruitment process by automating many of the tasks that would otherwise be done manually, such as resume screening and ranking. This can save time and reduce the workload for HR staff and hiring managers.
- Greater accuracy: ATS can help ensure that resumes and job applications are evaluated consistently and accurately based on the criteria that the employer has set. This can help eliminate bias (but hold on to that thought for a moment) and ensure that the most qualified candidates are identified and considered for the job.
- Better organization: ATS can help keep all of the relevant information about job applicants in one central location, making it easier for employers to track and manage the recruitment process.
- Enhanced candidate experience: ATS can provide a more personalized and efficient experience for job applicants, which can help improve the employer’s reputation and make it more attractive to top candidates.
- Cost savings: By automating many of the tasks involved in recruitment, ATS can help reduce labor costs and improve the overall efficiency of the hiring process.
It is the age of data, recruitment and human resource are just part of the equation. Long gone are stacks of paper resumé and people walk in to ask for an office job. Websites like Indeed, Glassdoor, LinkedIn may make it so easy for job seekers to find and apply for jobs without having to go anywhere, they are also the channels through which digital copies of resumé are collected and fed into ATS. Which leads to our next point.
How ATS process applications?
The way ATS processes applications can vary depending on the specific system and how it has been configured by the employer. In general, however, ATS allows employers to input certain criteria, such as keywords and experience level, and then automatically screens resumes and job applications to identify and rank potential candidates who match those criteria. The employer can then review the ranked results and decide which candidates to move forward with in the hiring process.
Some well-known applicant tracking systems (ATS) include Taleo, Lever, Workday, and Greenhouse. These are specialized software programs that help companies organize and manage job applications and resumes. They often include features such as resume parsing, applicant tracking, and job posting capabilities.
If you think about how these softwares are effective, is that they are built to process data. All the work history, skills, education, certificates, etc., need to be extracted, or parsed from CV files, typically from .PDF files. Any ATS would at least have the capability to parse a CV file.
CV parsing, also known as resume parsing, is a technology that helps to extract information from a CV or resume and convert it into a structured format that can be easily stored and analyzed. This allows companies to quickly and efficiently process a large number of job applications and resumes, and to extract relevant information such as a candidate’s education, work experience, and skills.
Once you have the data broken down and categorised, then they can easily be searched or even ranked according to any criteria.
Then why are ATS bad?
Parsing a CV, or resume, can be difficult for a few reasons. One reason is that CVs can be structured in many different ways, depending on the individual’s work experience, education, and other factors. This makes it difficult for a computer program to accurately interpret and extract the relevant information from a CV.
Another reason is that CVs can contain a lot of unstructured data, such as free-form text descriptions of work experience and skills. This unstructured data can be difficult for a computer program to interpret and process accurately. In a perfect world, we would all use one CV format.
Additionally, CVs often contain a lot of personal information that is not relevant to the job being applied for. This can make it difficult for a computer program to accurately identify and extract the relevant information from a CV.
Can we solve that?
Yes, machine learning algorithms can be trained to parse CVs, or resumes. In order to do this, the algorithm must be trained on a large dataset of CVs that have been manually labeled with the relevant information, such as work experience, education, and skills. This allows the algorithm to learn the patterns and structures that are commonly found in CVs, and to accurately extract the relevant information.
Once the algorithm has been trained on this dataset, it can be applied to new, unseen CVs to automatically extract the relevant information. Of course, the accuracy of the algorithm will depend on the quality of the training dataset, as well as the complexity of the CVs being parsed.
I recently led a project to build a custom machine learning algorithms for my last company to parse and process CV more accurately. It takes thousands of beautifully labelled CV files and months of work from many people, to reach a satisfiable level of accuracy and efficiency to rival a human. But that’s a (success) story for another day.
Safe to say, not a lot of ATS can parse CV perfectly everytime. Efficiently, yes. But there are always CV that can throw them off. Sometimes, the PDF is not a text-based file. Instead, some people could design a whole CV in Photoshop, save it as PDF and send it off, hoping a real human will read it instead of a machine. Parsing in this case, will return an empty result, as there is no text to extract but may be some meta data that’s not worth looking at. In this case, OCR is needed but even then, an effective Natural Language Processing engine that can understand the texts in the context of a CV is still more important.
Why else is ATS bad?
ATS can be biased in a few different ways. For example, ATS use algorithms to screen resumes and match them to job requirements. These algorithms are trained on data, and if the data used to train the algorithm is not diverse or representative of the population, the algorithm may be biased. This means that it could be more likely to reject resumes from underrepresented groups, or to prioritize resumes from certain groups over others.
There is a concern that applicant tracking systems (ATS) may be biased against certain racial or ethnic groups. This means that it could be more likely to reject resumes from underrepresented groups, or to prioritize resumes from certain groups over others.
However, it’s important to note that the extent to which ATS is biased is not well understood. There have been some studies that suggest that ATS can be biased, but the results have been mixed and more research is needed. Ultimately, it’s up to each individual business to carefully consider the potential for bias in their ATS and take steps to prevent it.
Another way that ATS can be biased is if the keywords used to match resumes to job requirements are not chosen carefully. For example, if a job posting includes a lot of masculine-coded language (such as “dominate” or “assertive”), the ATS may be more likely to match male applicants to the job, even if a female applicant is equally qualified.
And how can we avoid bias in ATS?
There are several steps that businesses can take to prevent bias in their applicant tracking systems (ATS). These include:
- Use diverse training data. It’s important to use a diverse and representative dataset to train the algorithm. This means labelling data with even more diverse criteria.
- Carefully select keywords. ATS often use keyword matching to screen resumes and match them to job requirements. It’s important to choose keywords carefully to avoid bias. For example, avoid using gendered language (such as “assertive” or “dominate”) that could bias the system against certain genders. Understanding contexts is still a developing aspect of ATS.
- Regularly review and test the system. Even with diverse training data and carefully chosen keywords, there is still a risk of bias in ATS. To prevent this, it’s important to regularly review and test the system to ensure that it is not biased. This can include conducting blind resume reviews and comparing the results to the results produced by the ATS.
The tools we create are merely tools, not some miracles that can do the job for us. In the case of recruitment, the human aspect is quintessential to the success of hiring, not how smart the machine learning algorithm is nor how fancy the ATS is built.
I am Henry, a Product Manager with a passion for recruitment and machine learning. I’ve been in Product development for 6 years, and job board development for 3. Recently, I joined LexisNexis UK to work on machine learning products, more specifically in NLP space.