How do you prepare for an interview?

The most common way to prepare is to find a list of questions and build the answers in advance. That rests on a single assumption: that everyone is asked the same questions. In a system where the questions are prepared in advance, that assumption does not hold, because every question there has somewhere to anchor: a criterion of the posting, or a claim in your CV.

The second wrong place is the length of the answer. On the scale, the only difference between adequate and strong is a concrete anchor: a lived example, a number, the reasoning behind a decision you made, or the tool you used. A short answer carrying one of those can count as strong, while a long answer carrying none stays general.

This guide describes the screen of a company running its interviews with questions prepared in advance. The definition of the method sits on a separate page on the employer side: what a structured interview is. Do not assume every company you apply to works this way; the interview module is open on certain plans and most conversations still run end to end in human hands.

• Where the question comes from: a criterion of the position, or a claim in your CV.
• What is measured: not the length of the answer, but the concrete anchor inside it.
• The decision: the system produces a suggestion, and a person makes the call to advance or reject.

The requirement list you see on a posting page is plain text. The same list is not plain on the hiring side: every line carries a weight and some are marked required. The public page deliberately drops those two fields and prints only the text of the line. The list you read and the record the other side uses are the same record, only displayed differently.

Under full weighting, the weights of the lines have to add up to exactly 100. A line marked required counts as a genuine barrier; when it is not met, it trips the hard gate. The lines are not equal, but there is no way to read which one is heavy from the posting page.

What to do is not turn the list into a guessing game. Write a single evidence sentence for every requirement: on which job, when, what you did, and what came out of it. Note the lines you have nothing for somewhere separate; step four is about exactly those.

• The posting criteria are turned into a checklist produced once per posting and applied identically to every candidate.
• Each line looks at a single fact that can be answered yes or no: not is it good, but is it there.
• The weight and the required flag live only in the panel and never appear on the posting page.

The quality condition for generating questions is written down: every question has to rest on a specific criterion of the position or a specific claim in the candidate CV. Generic questions of the tell me about yourself kind, questions gluing two topics together with an and, and questions that close with a yes or a no are not produced. Under the question you face there is usually a line you wrote yourself.

The same condition carries the most useful steer for preparation: topics that appear in the criteria but look thinly evidenced in the profile are treated as the most valuable questions. In other words, the claim you covered in a single line and never expanded on is the most likely question. You will be asked about the project you described least, not the one you described most.

The pool knows three categories and what each of them looks for is written separately. The rounds have separate jobs too: the first round verifies the basic competency and genuine interest in the position, the second digs under the topics that stayed general in the first and asks for a concrete example, a number and the reasoning behind a decision, and the third closes the remaining risks such as level of responsibility and expectation fit. What is asked in the second round is not a new topic, it is the detail of the same one.

• Technical: the depth of the technologies, tools and projects you wrote in your CV; not memorised facts but application and judgement.
• Behavioural: questions resting on something that happened; not a hypothetical scenario but an example you actually lived.
• Culture fit: which kind of team you are productive in, how you resolve disagreement, how you take feedback.

In answer grading the model produces no score, no percentage and no number. All it can give is one of four call words; the arithmetic is done in code. The definitions of the four calls are not left to guesswork either, they are written in the prompt, and the table below is their exact counterpart.

The weights are fixed: 1, 0.67, 0.33 and 0. The summary is the average of those weights, rounded out of 100. The distance between them says where preparation should go: moving one answer from adequate to strong gains 0.33 units, and all it takes is putting a real anchor inside the answer.

The reasoning is built from inside the answer too: which sentence of yours led to the call is written down, nothing absent from the answer is assumed, and the reasoning is at most two sentences. The consequence is blunt: what you did not say is not counted in your favour. If you do not say the example in your head out loud, that example does not exist.

The answer summary is the weight average of every question in that round. An unanswered question counts as zero and is not removed from the total question count. The reasoning in the code comment is one sentence: if it were removed, a candidate could lift their summary by staying silent. Brushing a question aside does not protect the summary, it pulls it down.

This rule is the opposite of the one on the CV side, and that is where confusion happens most. In the CV evidence rubric an item with no evidence does not count as half a point, it leaves the denominator; the reasoning there is the sentence the absence of evidence is not a shortcoming of the candidate. In an interview, absence is a choice: the question was asked and no answer came.

The consequence: an answer that addresses the question but stays general counts as adequate and carries 0.67, while giving no answer at all carries 0. On a topic you do not know, the thing to do is not to fall silent but to stay with the question, describe the real experience you do have, and say clearly how far your knowledge goes. Build that into a sentence before the conversation and keep it ready.

• The summary is the weight average of every question in the round, rounded out of 100.
• An I do not know on its own counts as unanswered and carries zero.
• Text unrelated to the question also sits on the unanswered side.

The invitation email carries a fixed set of information, and most candidates close it without reading the bottom half. Every invitation holds the date and time, the duration of the conversation, whether it is in person or online (Google Meet), and the note passed on to you. An in person conversation adds the address, a Google Maps link and an embedded map image; an online one adds the Meet link. At the very bottom sits the this date does not suit me link.

Reminders go out in two windows: once when 2 to 24 hours remain, and once when 2 hours remain. If the two windows overlap, the 2 hour one takes priority, and each window fires only once per interview. The same reminder goes to you and to the interviewer. A reminder not arriving does not mean the conversation was cancelled, so put the calendar entry in yourself from the invitation email.

The this date does not suit me link in the email does not cancel the interview, it only tells the company. The page it opens holds one sentence of context, an optional explanation box and a single button. The explanation you leave appears in the email sent to the company; if you write nothing, the fact that there was no explanation is what goes.

On the candidate side you can meet interview questions in two places, and they are not the same list. The first is the Interview section of your own CV report, the second is the question pool in the panel of the company you applied to. You never see the second one.

The section in the CV report gives every question in three parts: the question itself, why it is asked, and what a good answer should contain. The categories split four ways: Technical, Experience, Behavioural, General. Generation asks for at most four questions and the report mapper carries at most five. On the free result this section is among the locked ones; the only visible thing is how many questions were prepared.

The pool on the panel side knows three categories: Technical, Behavioural, Culture Fit. If the CV analysis of your application ran, the first round of the pool is filled once with the questions from it, and those rows are written into the Technical category. That is why the labels of the two lists do not map one to one: two surfaces of the same engine, each using its own label set.

At the end of the round a single summary comes out of the answers, and that summary keeps bands of its own: 75 and above strong, 50 and above a second round. These are not the decision bands of the position, they are the thresholds of the answer scale itself.

What comes out is not a decision, it is a suggestion. Every suggestion closes with the same sentence, and that sentence is not typed by hand, it comes from the code: this is a suggestion, and the application status changes only with your approval.

This is where preparation gets confused the most. Let us separate what combines with what under three headings.

The six items below are what the five steps should leave in your hands the day before the conversation. None of them depends on which system the other side uses: an answer with a concrete anchor prepared serves you just as well in front of a human interviewer.

• You have an evidence sentence for every requirement in the posting: on which job, when, what you did, and what came out of it.
• Behind every claim covered in a single line of your CV stands an example you can tell.
• Every example carries a number or the reasoning behind a decision you made; with neither, the example stays general.
• You have built, in advance, the sentence you will use to say how far your knowledge goes on a topic you do not know.
• You have put the date, duration, conversation type and address from the invitation email into your calendar.
• If the date does not suit you, you used the link in the email without waiting for the day of the conversation, and wrote a workable time window in the explanation box.