Last updated: August 2026
Most reference checks are a courtesy call that confirms a decision already made. The offer is drafted, the referee is a name the candidate supplied, and the questions are soft enough that nothing can go wrong. Ten minutes later the box is ticked and nobody has learned anything.
Run properly, a reference check is one of the cheapest screens in the whole hiring process. It costs a phone call, and it is the only stage where you hear about the candidate from someone who has watched them work for years rather than perform for an hour. I run them to a fixed structure on every placement, and the structure is what does the work. This post is how to get real information out of yours.
Why do most reference checks surface nothing?
Because they are improvised, and improvised calls drift into pleasant conversation. The referee says the candidate is great, the hiring manager agrees the candidate seems great, the call warms up, and every question that follows gets softer. Chat produces chat.
The selection problem runs underneath it. Referees are nominated by the candidate, so you are calling the two people on earth most likely to speak well of them. The useful information sits in the texture of the answers: the speed, the specifics, the small hesitations. Texture only shows up when the questions are consistent enough to compare against.
A mis-hire this call could have prevented is expensive. For a rep on a $120,000 to $140,000 base you are looking at $150,000 to $300,000 by the time underperformance, lost specifications, and rehiring are counted. I have set out that maths in The True Cost of a Bad Sales Hire. A disciplined 15-minute call is cheap insurance against that number.
What makes a reference check actually work?
Two things. A fixed set of questions asked in the same order every time, and the discipline to put "would you hire them again?" last and then say nothing.
That question is the whole call. Everything before it is context for how to read the answer. When I write a reference up for a client, the rehire line is the only thing in the document set as a verbatim quote, because how it was said matters as much as what was said. A clean, fast yes closes the matter. A pause, a qualifier, an "in the right role, probably" is the referee telling you something they have decided not to say out loud. When you hear it, ask the shortest follow-up available. Why not? Then listen.
The rest is structure. Ask the question, listen, write the answer down, ask the next one.
The discipline matters more than the questions. Commentary between questions is where checks fall apart, because every reaction you offer teaches the referee what you want to hear. Stay neutral, and resist rephrasing their answers back at them. A referee left to fill their own silence will often add the sentence that matters.
Running the same structure every time gives you a second advantage: calibration. After a few checks you know what a strong answer sounds like, so a mediocre one stands out immediately. Improvised calls never build that baseline.
What areas should a reference check cover?
Six areas do most of the work. The relationship, the strengths, how the person responds to management, how they operate with colleagues and clients, where they need development, and whether the referee would hire them again.
The relationship question comes first because it weights everything after it. A direct manager across three years is a different witness from a colleague across six months, and you should hear the answer before deciding how much the rest of the call means. Push for the direct manager wherever you can. They saw the work set, the work done and the gap between the two, which a peer usually did not.
The strengths question is a quality test as much as a content test. A referee with real knowledge gives you specifics: named accounts, a territory turned around, the way the person prepared for presentations. Generic praise, "great guy, works hard", tells you the referee either did not observe closely or is being careful.
The development question is the honesty test. Nobody credible has no development areas. A referee who cannot name one is protecting the candidate, and that itself is information. A referee who names one thoughtfully has just told you the reference is real, and usually handed you a useful onboarding note as well.
The rehire question goes last because it summarises everything above it, and because it is the one answer worth hearing after the referee has already talked themselves warm.
How do you handle a referee who gives you nothing?
Probe once, with a request for an example, then move on. "Could you give me an example of that?" converts a vague answer into a specific one more often than any other sentence in the call. If the second answer is also fog, note it and continue. Two probes on the same question turns the call into an interrogation and shuts the referee down.
For the over-praiser, the same tool works harder. A referee describing a flawless candidate should be asked for the example every time. Specifics separate genuine enthusiasm from loyalty, and a superlative that cannot produce a single concrete story has told you what it was worth.
How do you verify specific claims from the interview?
Save them for the end, after the structured questions are done. If the candidate told you they grew a territory 30 per cent, managed five reports, or landed a named account, close the call with "they mentioned X, can you give me some context on that?"
Held to the end, these probes are clean verification. Scattered through the call, they break the structure and signal exactly which claims you doubt. In sales hiring these checks matter more than in most functions, because sales interviews select for exactly the persuasive polish that makes claims sound bigger than they were. The interview itself needs its own defences, and I have written about those in the interview questions post. The reference check is where a claimed number gets tested against a witness.
What do you do with a lukewarm reference?
Treat it as a flag to investigate, and calibrate for the referee first. Some people are measured about everyone they have ever worked with, and a careful referee's "solid, reliable, did the job" can be honest praise in their register.
The useful move is a second data point. Ask the candidate for another referee, ideally a direct manager from the relevant period, and run the identical structure. Two lukewarm references from two independent witnesses is a pattern. One, from a naturally dry referee, may be nothing. If the pattern holds, raise the specific concern with the candidate directly and watch how they handle it. A good candidate engages with the concern. A defensive rewrite of history is its own answer.
What a lukewarm reference should never do is get overridden in silence because the offer was already drafted and everyone is tired. That is the exact moment the check exists for, and the cost of ignoring it is the subject of half the mis-hire stories I hear. What good looks like in this market is worth being fussy about, and I have described it in What Makes a Great Building Products Sales Rep. If the numbers help the decision, the Cost of a Bad Hire Calculator is there.
What do you do with a negative reference?
Look at who is giving it before you look at what they said. A direct manager naming a specific concern is evidence. A former colleague running someone down is a different thing entirely, and you hear it more often than people expect. Office politics, a promotion that went the other way, an old falling out. None of it shows up in the reference, all of it shapes the answers.
So take it with a pinch of salt until you know where it came from. That does not mean dismissing it. It means finding out whether the person had the vantage point to know, and whether they had a reason to shade it. Then get a direct manager on the phone and run the identical structure. If two independent people who were well placed to see the work say the same thing, you have a pattern worth acting on. One hostile voice on its own is not.
The candidate deserves the same treatment you would want. I have seen good people nearly lose roles over a reference that turned out to be a grudge, and the only reason it did not land was that somebody checked the source.
Can you trust what a reference tells you?
Not completely, and anyone who tells you otherwise is selling something. A reference is one person's account of a working relationship that has ended, filtered through how it ended, how the business was travelling at the time, whether the referee was any good at managing, and what has happened to both of them since. There are too many factors in play for one call to be the answer.
Motive is part of that. Most referees are straight with you, but not all of them are, and the ones who are not rarely announce it. I run references on every placement and I still treat them as one input among several. A reference confirms or complicates what the interview told you. It does not replace it and it should not carry a decision on its own. Where it earns its money is at the margins: the candidate you are already unsure about, the claim you could not test, the second opinion on something the interview raised and never settled.
What should a written reference actually look like?
A record, not a recommendation. When I write one up it is first person, it covers every area in the same order every time, and it says what the referee said. No flags, no contradiction call-outs, no verdict from me at the bottom. If the referee could not name a development area, that is what goes in the document, rather than a manufactured weakness to make it look balanced.
The client is capable of reading it. My job is to run the call properly and hand over what was actually said, in a form they can compare against the next one.
How do you build reference check questions for a specific role?
Everything above is the general structure. The questions that surface the most are the ones written for the actual role and the actual referee, because a spec-focused BDM and a trade counter rep fail in completely different ways, and a former manager can answer things a former colleague cannot.
If you use ChatGPT, Claude or Gemini, this prompt does that part for you. Fill in the four brackets and paste it in. It gives you a question set for that specific call, plus what a weak answer sounds like on each one so you know what you are listening for.
You are helping me prepare for a reference check on a sales hire in the
Australian building products industry.
THE ROLE
Job title: [job title]
Products and market: [what they sell and who they sell to]
What this person has to do well to succeed: [the two or three things
that actually determine success in this role]
THE REFEREE
Who they are to the candidate: [direct manager / skip-level manager /
peer / internal stakeholder / client]
How long they worked together and when: [duration and rough dates]
FROM THE INTERVIEW
Claims I want tested: [any specific numbers, accounts, or achievements
the candidate mentioned]
Concerns I want probed: [anything that felt thin, or leave blank]
WHAT I WANT BACK
1. Eight to ten questions in the order I should ask them, written in
plain Australian English, open rather than leading.
2. Under each question, one line on what a strong answer sounds like and
one line on what a weak or evasive answer sounds like.
3. Weight the questions towards what this referee is actually positioned
to know. Do not ask a peer to assess management response.
4. End with the two or three follow-up probes most worth having ready if
an answer comes back vague.
Keep it to one page. No preamble, no summary at the end. Just the
questions and the listening notes.Two things to hold onto when you use the output. Ask the questions in the order it gives you and do not improvise mid-call, because a fixed order is what lets you compare one candidate against the next. And keep the rehire question last no matter what the model suggests, since it is the summary of everything that came before it.
The prompt writes the questions. The discipline in the call is still yours.