Reference Calls That Reveal How Someone Handles Ambiguity

The hiring manager closed the interview loop with a live ambiguity exercise that went well. Two days later she dialled a former colleague and read from the corporate script: On a scale of one to five, how adaptable is he? The referee paused, inventing a number that felt polite. Nothing in that pause resembled the incomplete warehouse interface both people had lived through six months earlier.

That gap is where most reference packs fail. The form asks for a trait score. The work asks whether someone can name an unknown, seek evidence, and change course when the feed arrives late or the owner of a boundary is missing. A useful call follows the work, not the form.

The Script Asks for Scores Nobody Measured

Standard reference packs still lean on adjectives and ratings: adaptable, collaborative, resilient, culture fit. Referees who want to help will answer. They will not usually say they never kept such a score.

Reference checking earns its keep when it gathers past behaviour from people who actually observed the work. Meta-analytic summaries treat reference checks as a modest predictor relative to stronger selection methods; they are not a substitute for a structured interview loop. The call still matters. It can confirm — or puncture — claims the candidate made under pressure. It only does that if the questions invite incidents rather than compliments.

A one-to-five adaptability score compresses unlike situations into one invented unit. The late interface, the missing product owner, and the contested data definition are not the same fog. Treating them as one trait wastes the only advantage a referee has: they were there.

Asking whether someone "can handle ambiguity" is the wrong question. Referees cannot score a trait they never measured. They can recount a shared unclear situation.

What a Referee Can Honestly Know

A referee can speak to shared work, shared timelines, and shared incomplete information. That is a narrower claim than most scripts imply, and it is a stronger one.

Government assessment guidance treats reference checking as an evaluation of past job performance collected from supervisors, peers, and subordinates who knew and worked with the applicant. Structured public-service practice goes further: ask for facts, relevant incidents, and behavioural examples, not opinions about general ability. Acas puts the same duty in quieter language for UK employers: if an opinion is offered, evidence should support it, and callers should ask only for what they need.

Peers often saw more of the fog than a distant sponsor did. A former manager who sat in the same stand-ups can narrate the first forty-eight hours after a half-specified interface landed. A programme sponsor who only saw the go-live date can speak to outcomes, not to how the candidate handled the unknown on Tuesday afternoon. Prefer the person who shared the contested boundary.

The best referee is not always the former manager. A peer who shared a contested boundary often saw how the person behaved when the map was incomplete.

If the referee never shared an ambiguous situation with the candidate, end that line of questioning. Record that the behaviour was not observed rather than forcing a score into the silence.

Questions Referees Can Actually Answer

Replace trait prompts with shared-situation prompts. Keep the same core questions across finalists so the calls remain comparable and structured.

Picture a senior analyst candidate whose last programme included a retail stock feed that arrived incomplete three nights in a week. The order platform and warehouse connector disagreed about which field closed the receiving step. Both teams stayed polite. Nobody owned the blank.

Ask the referee who lived that week something like this:

• Shared fog: "Tell me about a time the two of you faced incomplete information on a delivery — a late feed, a half-specified interface, or a contested owner. What was missing?"
• First forty-eight hours: "In the first two days, what did they ask for, and what did they stop assuming?"
• Evidence: "What would have changed their mind about the approach they took?"
• Course correction: "When better information arrived, what did they revise — and what did they leave alone?"

Those questions are answerable because they point at a scene both people occupied. "How adaptable are they?" is not answerable without invention.

Follow-ups should stay short and factual. Canada's structured guide suggests probing a generalisation with "Can you give me a specific example?" and separating group credit with "What was the applicant's personal role in the events?" — both probes for behavioural evidence. That is enough. Do not coach the referee toward the adjective you hoped to hear.

Skip the global trait bank, the culture-fit slogan, and the naked "Would you rehire them?" until an episode sits underneath. Rehire without an episode is a popularity poll.

Listening for Ambiguity Handling — Not Loyalty

The call fails when warmth is mistaken for signal. A referee who likes the candidate will offer praise. Praise is not a decision sequence.

Listen for three markers inside the episode:

• Naming the unknown — they can say what was missing without pretending certainty they did not have.
• Evidence-seeking — they asked a named person for a named artefact, or they stopped a local assumption until a signal existed.
• Course correction — when the map improved, they changed a rule, a sequence, or an owner, without turning the review into blame theatre.

Put two answers about the same missed receiving step side by side. The courtesy answer: "She's very adaptable and great under pressure." The diagnostic answer: "On Wednesday the feed lacked the warehouse acceptance field. She refused to mark stock complete from allocation alone, asked the integration lead for an explicit accepted/rejected/timed_out contract, and rewrote the transition rule once that state existed." Only the second answer lets you hire for the job you actually run.

A strong reference that offers only adjectives is weak evidence. Praise without episodes cannot distinguish performance from politeness.

Notice what the diagnostic answer does not do. It does not declare the candidate brilliant. It does not rate adaptability. It narrates a decision under incomplete information — the same class of work your brownfield programme will demand in the first month.

If the referee freezes at the word "failure," reframe. Ask about a late requirement, a contested owner, or a half-specified interface instead. Ambiguity need not be prosecuted as a miss; it only needs to be a shared incomplete map.

If opinions arrive without evidence, treat them as incomplete — the same standard Acas sets for fair references. Ask once for an example. If none comes, thank the referee and move on.

Anticipate the objection that this sounds cold. It is the opposite. Specificity protects people. It keeps ordinary mistakes inside a learning story and keeps serious conduct out of suggestive adjectives. You are not trapping a former colleague; you are refusing to invent a score on their behalf.

When the Call Is Done — Or Wrong

One clear episode with all three listen markers is often enough to confirm what the interview already showed. A long checklist of unrelated traits rarely improves the decision.

Thin signal means switch referees, not dig harder into the same silence. SHRM's practical advice matches that instinct: when an employer policy blocks meaningful detail, ask the candidate for another name rather than reading the blockage as guilt. Prefer someone who worked with the candidate recently on daily delivery work over a distant admirer.

Stop the call when the referee can only speak to reputation, politics, or outcomes they never watched being made. That is not a failed reference; it is an honest limit. The programme sponsor who celebrated the go-live can still be worth five minutes for delivery reputation — then you still need a peer for the fog.

Structure the next call the same way. Structure is what improves usefulness across selection methods, reference checks included.

Keep dignity intact. Do not force people to prosecute former colleagues. Do not smuggle discrimination risk into "culture" questions that wander off the job. Ask for what you need, listen for behaviour, and leave the rest alone.

Ask what was shared when the map was incomplete.

Listen for how the unknown was named, tested, and revised — not for how warmly it was praised.