# What information is most useful when deciding between candidates? **The most useful information when deciding between candidates** is specific evidence of reasoning, learning, and collaboration tied to the role's outcomes. Strong evidence shows how a person handled real constraints, changed a view when facts changed, and worked with other people when ownership or time was unclear. Prestige, confidence, speed, similar hobbies, and interviewer enthusiasm are weak information. They can make a conversation memorable without showing how the candidate will do the work. ## Start with the work you need to predict Useful information depends on the role. A team hiring someone to stabilize a payment service needs evidence about diagnosis, risk choices, and production ownership. A team hiring an engineer to guide a migration needs evidence about sequencing, influence, and learning across unfamiliar systems. The same impressive story can be relevant to one role and nearly empty for another. Write three or four outcomes before choosing questions. For each outcome, ask what behavior would make success more likely. If the person must reduce repeated incidents, you may need to hear how they separate symptoms from causes, make a safe change with incomplete data, and help others use the learning. This turns interviewing into evidence collection rather than general admiration. Keep the prediction modest. An interview cannot prove future success. It can provide samples that are more or less connected to the job. Your task is to collect enough relevant, comparable detail to make uncertainty visible. A confident story should not feel certain merely because it was easy to hear. ## Reasoning under constraints Ask for a piece of work where the obvious solution was not available. The database could not be replaced, the release date had a real customer reason, or a dependent team controlled the interface. Listen for how the candidate identified the binding constraint, generated options, and explained tradeoffs. The exact technology matters less than whether the reasoning can be followed. Follow the moment when new evidence appeared. Did a metric contradict the first theory? Did an operator point out a failure mode? Strong reasoning is not stubborn consistency. It includes noticing when an assumption is wrong and updating the plan without rewriting history. Ask what the candidate changed and what remained uncertain. Specific detail matters. A polished answer about balancing quality and speed may contain no usable evidence. Ask what quality risk existed, who accepted it, how it was monitored, and what happened later. You are not trying to catch the candidate. You are giving them room to show the actual shape of judgment. ## Learning tells you how range grows No candidate arrives knowing every internal tool, product rule, and relationship. Evidence of learning helps you judge what unfamiliarity is safe. Ask about a difficult area they entered, how they found reliable sources, which early belief changed, and how they knew they were becoming useful. Look for a method rather than a heroic claim about learning fast. Failure can offer strong learning evidence when the candidate names their own decision, the consequence, and the change that followed. A story where every problem came from a weak teammate reveals little about self correction. At the same time, do not reward dramatic confession by itself. The useful part is the link between new understanding and later behavior. Consider whether your process lets people demonstrate learning. A narrow trivia test mostly rewards prior exposure and quick recall. A small unfamiliar problem with room for questions can show how a person builds a model. Use only exercises that resemble a real demand of the role, and tell candidates what kind of reasoning you want to observe. ## Collaboration under real pressure Collaboration evidence is not whether the candidate seems friendly. Ask about a disagreement where responsibilities, expertise, or deadlines pulled in different directions. Listen for how they represented another person's concern, made authority visible, and protected the work relationship while still advocating for a technical need. Pleasant agreement is not the only successful pattern. Constraints reveal habits. During an incident, did the candidate keep product and support partners informed without flooding them with internals? During a design dispute, did they create a test or decision record? When another engineer struggled, did they take over silently or help the person regain ownership? These details connect treatment of people to engineering outcomes. Beware stories where collaboration means persuading everyone to accept the candidate's original plan. Influence can include changing one's own plan, narrowing a claim, or escalating a decision to the right owner. Ask what the other person needed and what the candidate would do differently. Respect becomes visible in choices, not in a label. ## Useful and weak signals | Useful information | Weak information | Why the difference matters | | --- | --- | --- | | Reasoning through a named constraint | Prestigious employer | Context is not judgment | | Updating a view with new facts | Fast answers | Speed can hide shallow thinking | | Learning method and later application | Confident delivery | Style is not growth | | Collaboration during disagreement | Similar hobbies or background | Familiarity creates bias | | Specific actions and consequences | Interviewer enthusiasm | Feeling needs an evidence trail | Weak information is not always meaningless. A previous employer can explain the scale somebody has seen, and response speed can matter in a role with urgent coordination. The error is treating a proxy as proof. Follow the signal into specific work. What scale did the person personally handle? What did the fast response allow others to do safely? Interviewer enthusiasm deserves particular care because it spreads easily. “I would love to work with them” may reflect an excellent evidence based conversation, or it may reflect familiarity. Ask what the candidate did that created the view and which role outcome it supports. If the answer remains general, record the warmth but do not let it carry the decision. ## Comparable questions and independent notes Use the same core question for candidates when it covers an essential outcome. Follow up can differ because experience differs, but the starting point should give each person a fair route to relevant evidence. Comparable does not mean robotic. It means you can explain why the information belongs in the same category. Have interviewers write notes before discussing the candidate. Include observed actions, relevant context, and the requirement informed. This protects memory from the first forceful opinion. It also exposes when two interviewers interpreted the same answer differently, which is useful information about the question or the role definition. Notes should distinguish quotation, observation, and interpretation. “Candidate asked whether retries were idempotent” is an observation. “Candidate has strong production judgment” is an interpretation that needs more support. Both can appear, but keeping the trail visible makes later comparison more accurate and feedback to the process more useful. ## How should you handle missing evidence? First ask whether the candidate had a fair opportunity. If no interview explored learning, the absence of a learning story is a process gap. Do not turn it into a negative. Decide whether the requirement is important enough for a focused follow up, or whether other evidence already answers it. More interviewing is not automatically fairer, so keep any addition narrow. Negative evidence is different. If a candidate receives a clear prompt to discuss a mistake and repeatedly shifts all responsibility to others, that behavior may matter. Record what was asked, what they said, and why ownership matters for the role. Avoid diagnosing personality. A single response still needs proportionate weight and context. Conflicting evidence deserves inspection rather than arithmetic. One interviewer may see careful reasoning while another sees weak prioritization. They could be observing different constraints. Compare the examples, question quality, and role relevance. The aim here is to understand what information you possess, not to conduct the entire decision meeting or average scores. ## When two candidates look close - Return to the role outcomes. - List the strongest evidence for each outcome. - Mark missing information separately from concerns. - Remove prestige and familiarity from the comparison. - Check whether questions were comparable. - Identify the smallest material uncertainty. - Use one focused follow up only if needed. - Accept the uncertainty that remains. Close candidates often expose an unclear priority. One may show stronger production depth while another shows stronger cross team learning. The answer depends on the work, current team support, and which gap can reasonably be learned. State that tradeoff directly. Do not invent a global ranking of who is the better engineer. A focused follow up should resolve one material question, not create another full loop. Share the context, ask a comparable question, and document the result. If the candidates remain close, make a timely choice using the evidence you have. Hiring information can improve judgment, but it cannot remove every risk. Responsible managers know when collection has reached its limit. ## Information should survive a careful retelling Test a signal by explaining it without the candidate's name, employer, accent, or presentation style. Can another manager understand what happened, which constraint mattered, and why it informs the role? If the claim collapses into “they seemed sharp,” you have an impression that needs a trail. If the account preserves actions and consequences, it can support a reasoned comparison. Ask interviewers what evidence would have changed their interpretation. Someone who says nothing could change it may be defending an identity judgment rather than evaluating a sample. A thoughtful answer might name a missing example, a clearer ownership boundary, or a second case under different constraints. This question reveals both the strength of the evidence and the interviewer's uncertainty. Keep the final record proportionate. A candidate does not need to be extraordinary in every area. They need enough evidence for the outcomes the role requires, with gaps the team can responsibly support. That standard is less dramatic than choosing the most impressive story, but it is more connected to the work that begins after acceptance. ## What are common questions? ### What evidence is most useful between candidates? Use specific examples of reasoning under constraints, learning from new facts, and collaboration during real engineering work tied to the role outcomes. ### Why are confidence and speed weak signals? They may reflect communication style or practice rather than judgment, learning, or the ability to produce safe results in the role. ### Should every candidate receive identical questions? Use comparable core questions for essential outcomes, then adapt follow up to understand the specific experience each candidate presents. ### What is the difference between missing and negative evidence? Missing evidence means the process did not reveal an area. Negative evidence is a relevant observed response that raises a role concern. ### What should you do when two candidates are close? Return to role outcomes, remove weak proxies, identify the smallest material uncertainty, and use one focused follow up only when it can change the choice. Related: [becoming a hiring manager](https://isilta.com/blog/becoming-a-hiring-manager/), [moving from interviewer to hiring manager](https://isilta.com/blog/from-interviewer-to-hiring-manager/), [hiring as a team lead](https://isilta.com/blog/hiring-as-a-team-lead/).