The Interviewer's Craft
面试官的手艺
大多数管理者以为面试能力随着晋升自动到账,证据却说:随意闲聊式的面试对绩效的预测只比抛硬币略好。面试是一门手艺:先定义胜任力、用统一问题与评分标准建立结构;用行为式提问代替假设式提问并层层追问;警惕晕轮效应与『文化契合』式偏见;管住自己的说话比例、容忍沉默;分清危险信号与紧张噪音;面试后先独立打分再集体复盘。别忘了面试同时也是营销——最强的候选人在面试你,而被拒的候选人会把体验讲给整个行业听。
当前浏览器暂不支持语音朗读
Most managers assume that interviewing is a skill they acquired automatically, somewhere between their promotion and their first hire. The evidence says otherwise. Unstructured interviews — the friendly, meandering chats most of us conduct — predict job performance only slightly better than chance, while a single bad hire costs months of salary, a team's momentum and, not rarely, the resignation of the colleague who absorbed the damage. Interviewing is a craft, with techniques that can be learned, and the managers who learn them compound the advantage with every seat they fill.
Begin by noticing that an interview runs in both directions. While you assess the candidate, the candidate is assessing you: your preparation, your questions, the way you speak about the team you lead. In a tight talent market, every interview is marketing. The strongest candidates have options, rejected candidates talk, and a dismissive hour with an unprepared interviewer travels much further than any employer-branding campaign. An interviewer who has clearly read the CV, builds genuine rapport in the first minutes, and answers hard questions about the role honestly is doing brand work of the most persuasive kind.
Structure is the least glamorous upgrade and the most powerful one. Before you open a single CV, define the four or five competencies the role genuinely requires, design questions that test each of them, and ask every candidate the same core set. Pair the questions with a simple rubric, written in advance, that describes what a weak, an acceptable and an excellent answer looks like. Structure feels mechanical precisely because it removes the improvisation in which bias thrives — and it is the only way answers from Tuesday's candidate can be fairly compared with Friday's.
Prefer behavior to hypothesis. "How would you handle a difficult stakeholder?" invites a fluent little essay that measures nothing but eloquence. "Tell me about a time you disagreed with a stakeholder — walk me through it" invites evidence. Then probe, gently and relentlessly: What exactly did you do? What did the other person say? What happened next? What would you do differently? Hypothetical questions reward the candidates who talk best; behavioral questions reward the ones who have actually done the work. The difference only surfaces under follow-up, which is why the probe, not the opening question, is the real instrument.
The great enemy in the room is your own mind. The halo effect arrives in the first ninety seconds — a confident handshake, a shared alma mater, an easy joke — and quietly converts the rest of the hour into a search for confirmation. Likeable is not the same as competent, and "culture fit" is too often a polite name for "reminds me of myself", which is how teams end up hiring the same person eight times. The antidote is unglamorous: take notes on what was said rather than how it felt, score each competency against the rubric immediately, and hold your verdict open until the evidence is actually in front of you.
Watch your talk ratio. An interviewer in love with his own company can spend forty minutes pitching the vision and then wonder why every candidate seemed so agreeable. Aim to speak for roughly twenty percent of the hour, and let silence do some of the interviewing for you: when a candidate pauses after a hard question, resist the urge to rescue them with a hint. The pause is where the real answer is being assembled, and candidates who can think in front of you are showing you something no reference call ever will.
Learn to separate red flags from noise. Nervousness is noise: strong engineers stumble over words, and excellent marketers blank on a date under pressure. Signal looks different: an answer that stays vague after two probes, achievements narrated entirely in "we" with no retrievable "I", casual disrespect toward the coordinator who scheduled the call, every previous employer described as the villain. When you notice one, do not pounce and do not excuse it — note it, and test it again from another angle later in the hour. A pattern confirmed twice is information; a twitch observed once is weather.
The interview is not over when the candidate leaves the room. Hold the debrief the same day, but collect written scores before anyone speaks, so the most senior or the loudest voice does not become the verdict. Insist on evidence over adjectives — "she rebuilt the funnel dashboard in her second week at her last job" outranks "great energy" every time. Over several cycles, calibrate: compare interviewers' scores with how hires actually performed, and let the flattering questions die. Then close the loop fast, with a decision in days and courtesy in both directions — because in hiring, as in marketing, the audience you lose still talks to the audience you want.