A customer walks into a store and starts browsing. The employee who helps them doesn’t perform “rapport building,” pause, then switch to “needs discovery,” pause. They have a conversation, in real time. The employee reads the moment and adjusts as they go. Most pre-hire assessments don’t work the way that conversations happen. They test language in one exercise, behavioural in another, and cognitive in a third, then hand recruiters a stack of separate scores and ask them to figure out what it all means.
Recruiters don’t have the bandwidth for that anymore. Applicant-to-recruiter ratios keep climbing while completion rates keep falling, and stacking multiple single-skill assessments onto an already-long funnel makes both problems worse. The State of Frontline Hiring Report published by ICIMS late last year revealed that 60% of candidates drop out due to friction in the hiring process. That means six of every ten qualified candidates never finish what they were asked to do.
Trimming the assessment won’t necessarily fix the problem. A peer-reviewed article from the Journal of Applied Psychology found that cutting an assessment’s length by 15 minutes has only raised completion rates by 1 to 2 percent. The bigger drag on completion, and on data quality, is asking candidates to prove five different things in five different formats when the job actually asks for all five at once.
Research Reveals Stacking Reduces Reliability
HiringBranch’s research team, led by Chief Research Officer Assaf Bar Moshe, put skills scoring to the test on one of the highest-volume roles in the frontline economy: customer service. Four expert human evaluators scored nearly 500 candidate conversations across four core skills: acknowledgement, positive language, empathy, and active listening.
The individual skill scores were then compared against human ratings. Scored one at a time, no single skill got recruiters very far.
Even averaging four separately-scored skills only modestly improved accuracy over any one skill alone. It took an assessment built to read all four skills together, inside the same conversation, to substantially close the gap with expert human judgement. These findings don’t mean that individual skill scores are worthless. It’s that no single dimension, or even a weighted stack of dimensions, captures the judgement a trained evaluator forms by watching how a candidate actually manages a conversation from start to finish.
Explainability Is the Other Half of the Story
Combining skills into one score only works for recruiters if they can still explain the result. That’s where generic “LLM-as-a-judge” scoring runs into trouble because it can only say a response was strong or weak, and it never tells the recruiter which job-relevant behaviours drove that decision.
Regulators are no longer willing to take hiring outcomes on faith. New York City’s Local Law 144 already requires independent bias audits and public disclosure for automated employment decision tools, and the EU AI Act’s high-risk obligations for employment systems are now in effect. That means vendor conformity statements are becoming a standard procurement requirement even for employers who don’t hire in the EU.
A combined score built on a documented taxonomy, one that maps candidate behavior to specific, observable actions within a conversation flow, gives recruiters a defensible answer to “why did this candidate get this score.”
The Job Was Never One Skill at a Time
None of this is really about assessment methodology for its own sake. It’s about matching the way candidates are measured to the way the job actually gets done. A retail employee reads a browsing customer, builds rapport, uncovers a need, and handles an objection in one continuous exchange. A healthcare worker absorbs a complaint, asks the right follow-up, and delivers a hard message in the same breath. Scoring those moments one skill at a time was always going to miss what made the interaction work, or fail. This new research from HiringBranch suggests recruiters don’t have to keep making that trade-off between a realistic measure of performance and a fast, defensible hiring process. They can have both, in a single conversation.
This research was shared by one of our network partners, HiringBranch, a company that identifies top talent based on demonstrated skills for customer-facing roles
