Week 24 Day 6: Reference Checks That Actually Work

Standard reference checks are theater. The candidate provides references who will say positive things. The reference says positive things. The hiring manager checks the box. Nobody learns anything. The reference check that actually works asks the reference questions that the candidate could not have anticipated -- questions about specific behaviors, not general impressions.

Here is the reference check protocol I use. First rule: always call the reference. Email reference checks produce sanitized, carefully worded responses. Phone calls produce spontaneous reactions -- the hesitation before an answer, the qualifier that follows praise, the topic the reference avoids. These signals are invisible in email and invaluable on the phone. Second rule: ask behavioral questions, not assessment questions. Instead of 'How would you rate this person's teamwork?' ask 'Can you tell me about a time this person had a conflict with a teammate? How did they handle it?' The reference was not prepared for this question. Their answer comes from genuine memory, not from the script the candidate provided. Third rule: listen for what is not said. If you ask 'Would you hire this person again?' and the reference says 'She is very talented' -- they did not answer the question. That non-answer is data. If you ask about collaboration and the reference describes individual contributions, that is data. People do not omit positive information by accident. Fourth rule: ask the diagnostic question: 'If I were going to manage this person, what is the one thing I should know that would help me get the best out of them?' This question gives the reference permission to share a concern disguised as coaching advice. The answer often reveals the exact dynamic that will emerge in the first 90 days. Fifth rule: call at least one reference the candidate did not provide. A former colleague you identify through LinkedIn or mutual connections will give you a more candid picture than the hand-picked references.

The reference check protocol is informed by research on the validity and limitations of reference checking in personnel selection. Taylor, Pajo, Cheung, and Stringfield (2004) in their meta-analysis found that reference checks have a validity coefficient of only 0.13 for predicting job performance when conducted in the traditional format (assessment questions to candidate-selected references). However, structured reference checks using behavioral questions have a validity of 0.29 -- comparable to the validity of personality tests and significantly higher than the return on time investment for most other selection methods. The phone-versus-email distinction is supported by research on 'communication richness theory' (Daft and Lengel, 1986), which demonstrates that richer communication channels (phone, video) convey more information through paralinguistic cues -- tone, pace, hesitation, emphasis -- that are absent in lean channels (email, text). Research by DePaulo and colleagues (2003) found that deception is significantly harder to maintain in synchronous verbal communication than in asynchronous written communication because written responses can be edited and polished while verbal responses must be generated in real time. The 'unrequested reference' technique -- calling someone the candidate did not nominate -- draws on research by Aamodt, Bryan, and Whitcomb (1993) demonstrating that non-nominated references provide significantly more balanced evaluations (r = 0.24) than nominated references (r = 0.08) for predicting job performance.