Run a structured interview, asking every candidate the same questions and scoring them with the same guide.
When you hire through conversation, you tend to pick the person who was best at the conversation. In an unstructured interview the talk moves from topic to topic, no two candidates hear the same questions, and you leave with an overall feeling about each person. That feeling then becomes the decision. Mostly, the format tests how good someone is at unstructured interviews.
Some people do very well in this format: confident, talkative candidates who are similar to the interviewer. Others tend to score lower whatever their ability to do the work. This includes quieter people, people from different backgrounds, and people answering in their second or third language. Researchers in personnel psychology have studied this for decades. Their findings include a widely cited 1998 review by Frank Schmidt and John Hunter, and they show that structured interviews predict job performance better than unstructured ones.
A structured interview asks every candidate the same questions in the same order. Each question is linked to a skill in the role profile, and every answer is scored the same way. Because everyone answers the same questions about the same skills, you end up comparing them on the job they would do, and how the chat went drops out of the comparison.
Wei Ling needed someone for a new role running a client account. Over two days she met three candidates and talked with each one about their career, their previous job and their weekends. By the end she felt certain about the second candidate, Shu Fen. Shu Fen was easy to talk to, laughed at Wei Ling's jokes and had studied at NTU, as Wei Ling had. Aziz, one of the other two, was quiet and came across as less sure of himself.
Daniel then asked her what made Shu Fen better than the other two at running a client account, and Wei Ling had no answer. When she thought about it, she saw that she had asked each candidate different questions, her notes were thin, and her judgement came mostly from how each conversation had felt.
Go back to the role profile from Lesson 7.1, "Define the role before you write the ad", and write out the skills it calls for. Each question you ask should test one of those skills.
There are two kinds of question you can use. A past behaviour question asks what the person has actually done. For example: describe a time a client was angry about something that wasn't your fault, what you did about it, and how it turned out. A situational question gives them a realistic case from the job and asks what they would do. For example: at 5pm a client emails to say their shipment is missing and they need an answer tonight, and your manager is on leave. What do you do? Past behaviour questions work well for candidates with relevant experience. Situational questions suit people early in their careers or moving from another field. Many interviewers use some of each.
Plan your follow-up questions in advance as well. Keep them neutral and use much the same ones with everyone: "What did you do next?", "What was the result?", "What would you do differently?" They help you hear the whole answer while the candidate still decides which way it goes. On the day, ask every candidate the same questions in the same order.
For each question, write down in advance what a weak, an adequate and a strong answer looks like. This written description is the scoring guide. It turns your overall impression into a judgement that someone else could follow. For the angry-client question, Wei Ling wrote:
Weak: blames the client or the company, or describes no specific action. Adequate: describes listening, apologising for the inconvenience and getting the immediate problem fixed that day. Strong: does everything in the adequate answer, and also describes following up to make sure the problem stayed fixed and changing something so it wouldn't happen again.
Record your scores as soon as each interview ends, before you talk about the candidate with anyone. When there are several interviewers, each one scores alone first and the group compares afterwards. If you discuss the candidate first, the most confident person in the room shapes everyone's view, and the scores lose their value.
Talking about the work and doing the work are different things. Where you can, add a work sample: a short task that looks like the real job. It should be short, realistic and the same for every candidate. Mark it against a written guide, the same way you mark the questions. For Wei Ling's role, it could be three client emails to answer plus a small spreadsheet of shipment data with two errors to find. Keep the time it asks for reasonable. If the task takes more than an hour or two of a candidate's time, pay them for it. Long unpaid tasks are unfair to candidates, particularly those with jobs and families, and they discourage good people from applying.
Wei Ling brought all three candidates back for a new final round. It had five structured questions, a scoring guide and a 45-minute work sample (invented figures). Shu Fen scored well on rapport. However, she found 0 of 2 spreadsheet errors, and her answer to the angry-client question was weak. Aziz had the highest score on both the questions and the work sample. Wei Ling could now give Daniel her reasons in two sentences, backed by evidence.
Base your own decision on the scores, and be ready to explain it with evidence to anyone who asks.
The questions themselves are the foundation of all this, and they come from the role profile. Write yours next, along with the guide that will tell you what a good answer sounds like.
Write five interview questions, each linked to a skill in your role profile. For each one, describe a weak answer, an adequate answer and a strong answer.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).