Reference checks change a hiring decision roughly 1% of the time. That's not an assessment. That's a ceremony, and an expensive one. At an average cost per hire of $4,700 (2025 SHRM data, most recent available), recruiters are spending 45 to 90 minutes per candidate coordinating and conducting calls that nearly always confirm what they already decided.
This guide is published by Mokka, an AI candidate screening platform. We include ourselves alongside competitors and aim to be accurate about both our strengths and limitations. We're an early-stage company with limited ATS integrations during our pilot phase, and our seat-based pricing can get expensive for large teams.
The ritual persists because stopping feels more dangerous than continuing. Legal teams want a paper trail. Hiring managers want someone to blame if things go sideways. Recruiters want to check the box and move on. The result is a process that functions as institutional comfort food: familiar, low in nutritional value, and almost impossible to give up. Meanwhile, candidates coach their references on what to say, 25% of references don't show up for scheduled calls, and 73% of hiring managers have already made their decision before reading the notes.
The question isn't whether reference checks work. The data answered that. The question is why we keep performing them, and what we should do with that time and money instead.
The Anthropology of the Reference Check Ritual
Every culture has rituals that outlive their original purpose. Anthropologists call this cultural persistence, the tendency for practices to continue long after the conditions that made them useful have disappeared. The reference check is a textbook case.
References made sense in an era of small networks and stable employment. When someone worked at the same company for fifteen years and their manager knew them intimately, a reference call could surface genuine signal. The principal (the hiring company) needed information from someone who had observed the agent (the candidate) over time. Information asymmetry was real, and references reduced it.
That world is gone. The median tenure for U.S. workers hovers around four years. Managers change roles frequently. Candidates curate their reference list the way a museum curates an exhibit: carefully selected pieces that tell a coherent story. According to 2025 Harvard Business Review analysis, 89% of candidates provide references they are confident will speak positively, effectively pre-screening out any negative signals before the process begins.
The ritual persists because rituals serve social functions beyond their practical ones. Reference checks signal that an organization is "thorough." They distribute responsibility for a hire across multiple people, diluting accountability. They provide a procedural anchor, something everyone agrees to do, even if no one can articulate what it achieves.
Dr. John Sullivan put it bluntly in 2025: "Reference checks are the most expensive useless process in recruiting. They're theater, and we know the outcome before we make the call."
The Economics: Sunk Costs and Confirmation Bias
From an economic perspective, the reference check is a sunk cost fallacy in motion. Companies have built entire workflows around references. Applicant tracking systems have reference-check modules. Compliance policies mandate them. Recruiters have been trained to conduct them. The infrastructure exists, and no one wants to be the person who suggests dismantling it.
Dr. Charles Handler, an industrial-organizational psychologist, described the dynamic clearly in 2025: "The sunk cost is real. Companies have invested heavily in reference infrastructure, and no one wants to be the one who says this adds no value."
The numbers paint a stark picture. According to Schmidt and Hunter's meta-analysis (cited in a 2025 ATD research review), reference checks have a predictive validity of just 0.13 for job performance, among the lowest of all hiring assessments. For context, structured interviews hover around 0.51, work sample tests around 0.54. References barely outperform random chance.
Yet 82% of employers still conduct them, per a 2025 SHRM survey, even though 96% of references provide only positive feedback. The system is designed to produce a single output ("looks good, proceed") and it delivers that output with remarkable consistency.
Stacey Berk, Managing Consultant at Expand Consulting Group, named the mechanism: "Most reference checks are confirmation bias exercises. Recruiters have already decided to hire and are just checking a box."
The cost isn't just recruiter time. Reference checks add 3 to 5 days to time-to-fill. In a competitive market, that delay has real consequences. Top candidates get snapped up. Hiring managers grow frustrated. Recruiter capacity gets consumed by coordination logistics rather than evaluation.
Why Legal Protection Is a Mirage
The most common defense of reference checks is liability reduction. The logic goes: if we hire someone who causes harm, and we didn't check references, we're vulnerable to a negligent hiring claim.
The data tells a different story. 67% of talent leaders say they conduct reference checks primarily for "legal protection" rather than candidate evaluation, according to the 2025 Greenhouse Hiring Trends Report. But reference checks rarely surface the kind of information that would protect against negligent hiring claims. Former employers, fearful of defamation lawsuits, typically confirm only dates of employment and job titles. The references candidates hand-pick are even less likely to reveal red flags.
A September 2025 class-action lawsuit in Texas highlighted the gap. A company was sued for negligent hiring after a reference check failed to uncover prior misconduct. The reference call had been conducted. The box had been checked. The liability protection had not materialized.
California's AB 1842, passed in January 2026, attempts to address the fear side of the equation by limiting employer liability for reference checks conducted in good faith. The legislation aims to encourage more honest reference conversations. It's a thoughtful intervention, but it underscores the core problem: the current system produces such unreliable information that legislatures feel compelled to intervene.
The legal argument for reference checks is a signaling mechanism, not a genuine risk management strategy. Companies check references to demonstrate that they checked references. The act itself is the product.
What Actually Predicts Performance
If reference checks have a predictive validity of 0.13, what works better? The research is clear, and it's not close.
Structured interviews, work sample tests, cognitive ability assessments, and job trials all significantly outperform references. Even unstructured interviews, which have their own well-documented problems, perform better than reference checks at predicting job performance.
The emerging alternative gaining traction is backchannel references, reaching out to people who worked with the candidate but weren't hand-picked by them. Jennifer Dulski, CEO of Rising Team, noted in 2025: "The best signal comes from backchannel references, people who worked with the candidate but weren't designated by them."
This approach reduces the information asymmetry problem. A candidate's curated reference list is a marketing document. A former colleague you find through your own network is more likely to provide unfiltered signal.
LinkedIn's November 2025 launch of "Reference Insights," an AI-powered feature that aggregates informal endorsements from colleagues, represents an attempt to systematize this. The tool pulls signals from people who worked with the candidate but weren't designated as references, creating a more organic and less controllable feedback loop.
AI-based assessments are also outperforming traditional references by significant margins. Research published by Pymetrics (now Harver) in June 2025 found that AI-based soft skills assessments had 4x the predictive validity of reference checks for 18-month retention. Tools like Mokka take this further, using an AI Evaluation Agent that screens resumes and conducts AI pre-interviews, generating evidence beyond what any curated reference list can provide.
The Sunk Cost Trap: Why Smart Teams Stay Stuck
Economists distinguish between sunk costs (money and time already spent) and marginal costs (the cost of the next action). Rational decision-makers ignore sunk costs and evaluate only marginal costs going forward. Humans don't work that way.
The reference check is a sunk cost trap because companies have invested so heavily in the infrastructure, the policies, and the cultural expectation that eliminating it feels like admitting those investments were wasted. So the cycle continues.
Laszlo Bock, former Google SVP of People Operations, described the phenomenon with unusual candor in a 2025 interview: "We found references had zero correlation with performance. We kept doing them anyway because it felt wrong to stop."
That sentence should be printed and taped to every TA leader's monitor. Google, a company famous for its data-driven hiring practices, couldn't bring itself to abandon a process it knew didn't work. If Google couldn't do it, imagine the challenge for a mid-size company with a risk-averse legal department.
The trap is reinforced by diffusion of responsibility. No single person owns the reference check process. Recruiters conduct the calls because hiring managers expect them. Hiring managers expect them because HR requires them. HR requires them because legal advises them. Legal advises them because "that's what companies do." The circular logic is the point. It protects everyone from being the one who recommended eliminating the practice.
What to Do Instead: A Framework
Breaking free from the reference check ritual requires replacing it with something, not just removing it. Here's a framework that actually predicts performance.
Replace Curated References with Backchannel Outreach
Instead of calling the three names the candidate gives you, use your network to find people who worked with them at previous companies. Ask specific, behavior-based questions: "What was it like to work with them on a tight deadline?" "How did they handle disagreement on the team?" These conversations produce signal because the candidate couldn't prep for them.
Greenhouse's August 2025 "Reference Check Reimagined" framework recommends exactly this approach, pairing backchannel outreach with structured digital references that ask specific performance questions rather than the open-ended "tell me about so-and-so" prompt that produces generic praise.
Use Structured Pre-Interviews
If you want to know how someone performs, watch them perform. Structured pre-interviews, whether conducted by humans or AI, create a standardized evaluation layer that references can't match. Companies using structured reference interviews report 23% higher new hire retention, according to 2025 LinkedIn Talent Solutions data, but only 18% actually use a structured approach.
The gap between the 23% benefit and the 18% adoption rate tells you everything about the gap between what works and what's comfortable.
Shift Liability Strategy from Process to Documentation
If the legal concern is negligent hiring, the solution is a documented, structured evaluation process that shows the company took reasonable steps to assess the candidate. Structured interviews, skills assessments, and background checks all provide stronger legal protection than reference calls that yield "they're great, hire them."
Reclaim the Time
The average recruiter spends 45 to 90 minutes per candidate on reference coordination and calls. For a recruiter filling 30 roles per quarter, that's 22 to 45 hours, roughly one full week of work, spent on a process that changes decisions 1% of the time. Redirect that time toward sourcing, candidate experience, or structured evaluation, and the ROI improvement is immediate.
The Decision Matrix
Here's a mental model for Monday morning. For every step in your hiring process, ask two questions:
- What would I lose if I removed this step entirely? Be honest. If the answer is "nothing meaningful" or "just the comfort of having done it," the step is a ritual.
- What could I do with the time and money I'd save? If the answer is something that actually predicts performance (structured interviews, skills assessments, backchannel conversations), the tradeoff is clear.
Reference checks fail this test decisively. They consume time, add days to time-to-fill, produce almost no decision-useful information, and persist primarily because stopping feels riskier than continuing. That's the definition of a sunk cost trap.
The hiring process is a chain, and every link has a cost. The reference check is a link made of tissue paper: expensive to maintain, weak under pressure, and kept in the chain only because removing it would require admitting it was never load-bearing.
The most effective TA teams in 2026 are replacing ritual with evidence. They're using structured assessments, backchannel signal, and AI-powered evaluation to make decisions based on what candidates can actually do rather than what their friends say about them. The reference check is being replaced, one thoughtful team at a time.