Talent acquisition technology has mostly failed to help answer two simple, but vitally important questions:
- The Recruiter’s Question – of this group of applicants, which are the best ones to invite for an interview?
- The Candidate’s Question – of all the jobs currently available, which ones am I most qualified for?
The Recruiter’s Question
Picking the best applicants to invite to an interview is not an easy task. Doing it effectively requires the recruiter to have subject matter expertise for each position they hope to fill and clairvoyance to know what each individual hiring manager is looking for. Recruiters are often expected to look through dozens, maybe hundreds of resumes in a short period of time to pick out the best ones. In practice this amounts to recruiters performing a quick (few seconds) scan of each resume for “skills”, i.e. keywords, that match those on a job description. It’s no surprise that 77% of hiring managers say that recruiters’ candidate screening is inadequate.
This process of matching via keywords – e.g. a job description that says “must be proficient in Excel” matches a resume that includes the word “Excel” – does not lead to very good predictions. It wastes an enormous amount of recruiter and hiring manager time interacting with presumed matches (via phone screening or interviewing), only to discover e.g. the candidate’s experience with Excel does not match what the job calls for. In some cases this discovery occurs after the candidate is hired, which can be very costly. Such false positives happen repeatedly, resulting in hiring managers becoming increasingly disappointed with the candidates sent to them, and recruiters becoming increasingly frustrated that hiring managers reject so many of their candidates. This dysfunctional recruiter/hiring manager relationship is common to many organizations.
Exacerbating the problem is all the time recruiters and hiring managers spend looking at stacks of resumes of unqualified candidates. All the time spent focusing on unqualified candidates means less time spent on qualified candidates, which often results in the most qualified candidates being missed and failure to respond to applicants in a timely manner (if at all). False positives and false negatives both contribute to higher attrition rates across the hiring organization.
The Candidate’s Question
Jobs boards like LinkedIn and Indeed offer alerts and other matching services to assist job seekers in finding out about jobs that they would like and are qualified for. Nearly all professionals have tried these at one point or another and realized how terrible they are. The common frustration is that candidates are sent too many jobs that aren’t even close to what they are looking for and their time is wasted going through them to hopefully find one that is truly a good match. Although the jobs were ostensibly selected based on the candidate’s unique skill sets and experience (as advertised by the candidate), most are left wondering how most of these jobs were selected.
The answer is often by matching keywords on our resumes to keywords in the job description. The problem is, as it is with the recruiter’s question, that high-level summaries such as resumes and job descriptions don’t contain nearly enough information to make good predictions about who is qualified for a job let alone who the hiring manager is going to like.
The Common Root Cause
Screening candidates by trying to match resumes to job descriptions is the root cause of both of these failings. Resumes and job descriptions are, at best, high-level summaries geared toward advertising, containing a heavy bias toward false positives. They cannot provide nearly enough detailed information to make a good prediction about whether a given candidate would be qualified for a job.
Surface-level information such as “proficiency in Excel” or “willing to travel” may indicate a match, but once you go below the surface by asking each candidate specific questions about the job – e.g. if they are “comfortable creating pivot tables, using macros, and writing VBA code in Excel” and if they “can travel up to 60% of the month” – it becomes easy to see exactly where an applicant is not qualified for the job.
Figure 1 Matching with surface-level data results in too many false positives
Figure 1 illustrates a typical false positive arising from keyword matching and not looking below the surface, where all the disqualifying information is found. The false positive will result in the recruiter’s and hiring manager’s time being wasted on further interaction with this candidate, instead of actually qualified candidates.
False negatives also occur from looking only at surface-level detail, and these result in great candidates being missed because they didn’t include the right keyword or otherwise get past the hasty resume scan.
AI Doesn’t Fix This
There is no shortage of technology that attempts to improve this matching process with highly-structured skills taxonomies, advanced search tools, and AI. These technologies might save recruiters some time by speeding up the resume matching process, and might make better predictions than a human taking a few seconds to scan a resume. However, their predictive capability is severely limited by using only surface-level detail contained in resumes and job descriptions. Machine learning algorithms’ predictions are only as good as the data they use (to see for yourself, just ask ChatGPT who won the 2022 Oscars). AI algorithms trained to find resumes that resemble those of “good” employees learn to perpetuate and amplify the systemic bias that exists in the organization (prompting legislation at federal, state, and local levels) and their predictions will still result in a large number of false positives and false negatives because they don’t address the root cause: resumes.
The only way to save time and achieve better outcomes is to go below the surface and make predictions based on the critical details and context of the candidate’s qualifications and job requirements.
The Solution: Actual Intelligence
Better predictions lead to better outcomes, and the best way to improve predictions is with more data; not social media profiles or personality tests, but the critical details and context that pertain to the job. How do you get that data? You simply ask for it on the application.
The key to screening efficiently and effectively is asking highly-specific questions that have multiple choice answers that enable automation and create an even playing field for all applicants. When you ask all applicants the same set of questions pertaining to the qualifications for the job, you find out exactly what you need to know about every applicant instead of just what they want you to know.
Because all of the questions have multiple choice answers, applicants can more quickly go through the application (especially on a phone) while learning what the job really entails, and their answers are more honest and precise than what’s written on their resume.
When each answer has a point assignment you can automatically quantify exactly how well each applicant matches what the hiring manager is looking for and objectively stack rank all applicants. This eliminates the tedious work of looking at resumes and conducting screening calls with unqualified applicants.
When you track hiring manager sentiments and correlate them to selected answer options, you can use those correlations to better predict which applicants the hiring manager will like the most based solely on their answers.
You can also source qualified candidates with a much greater precision than boolean searches for keywords and you can make significantly better predictions about which jobs to recommend to applicants.
SmartRank is technology that is enabling all of the above benefits and more for its customers.
SmartRank eliminates resumes entirely from the screening process, which addresses the root cause of poor predictions: using surface-level advertisements to match applicants to hiring manager preferences. Instead SmartRank uses actual intelligence to deliver unprecedented match prediction capabilities that deliver more qualified candidates to hiring managers without placing unrealistic expectations on recruiters to act as clairvoyant subject matter experts. This addresses the two root causes of the frustration cycle that leads to recruiter/hiring manager dysfunction.
SmartRank uses machine learning where it makes sense and can be effective. For example, we use large language models such as ChatGPT to help generate highly-specific, role-based screening questions. This significantly reduces the investment and increases the return for hiring managers. The predictive quality of the questions gets better over time with reinforcement from actual intelligence gathered by capturing and associating answers with hiring manager sentiments and other outcomes.
Replacing resume-based screening with actual intelligence removes the bias associated with resume data and places every applicant on an even playing field. It enables massive automation that reduces the amount of time spent on unqualified candidates, enabling recruiters and hiring managers to focus on the most qualified ones. When empowered with SmartRank organizations spend much less time and money on the hiring process so they can hire the best candidates faster.
Focus on the best and automate the rest with SmartRank. Get Started.