Evaluated text
The candidate writes an answer in their own words. This works best with an expected answer.
What Brondem evaluates
Brondem compares the candidate answer with the question, the expected answer, and the interview settings. The result is shown as a score out of 100, evaluation components, a short overview, strengths and weaknesses, and completion indicators.
1Question
The prompt should check one skill or one work scenario.
2Expected answer
The reference shows which points should be considered during evaluation.
3Candidate answer
The system evaluates the answer content and compares it with expectations.
4Candidate card
HR sees the score, question details, and completion indicators.
How to write a question and expected answer
Expected answer is the basis for evaluating a question
The expected answer sets evaluation criteria and helps compare a candidate's answer with the requirements of the question.
Strong question
Checks one skill, includes role context, and asks for an explanation or practical example.
Weak question
Too broad: "tell us everything", "rate yourself", or "what do you know about development".
Strong expected answer
Contains required points, accepted variants, signs of a strong answer, and typical mistakes.
Weak expected answer
Uses general phrases without criteria, for example "the candidate should understand the topic".
Question
Explain how you would check support operator quality after the first week on the line.
Expected answer
The candidate should name measurable criteria: first response speed, resolution completeness, communication tone, repeat contacts, script quality, and escalation quality. A strong answer includes a review example: dialog checks, sample scoring, error review with a mentor, and a short improvement plan.
Candidate answer
"I would review response time, quality of the solution, communication tone, and repeat contacts. Then I would sample several dialogs, give feedback, and agree on an improvement plan."
| What to add to expected answer | How the system uses it |
|---|---|
| Role level | The same answer may be enough for a beginner and weak for an experienced candidate. |
| Required points | The system can see which elements should be present in the answer. |
| Accepted wording | The candidate does not need to repeat the reference word for word. |
| Typical mistakes | It is easier to explain why the answer received a lower score. |
| Example of a strong answer | It makes evaluation more consistent and easier to verify. |
Which questions can be used in an interview
Evaluated test
Options can have one correct answer, several correct answers, or an important order. The score is based on correct choices.
Statistical
These questions do not affect the final score. They are useful for questionnaires, segmentation, and interview statistics.
| Type | Example | Affects score |
|---|---|---|
| Evaluated text | "Describe how you would handle a conflict with a client" | Yes |
| Single-choice test | One correct option among several | Yes |
| Multiple-choice test | The candidate should select all correct options | Yes |
| Ordered test | Correct sequence of actions: A to B to C | Yes |
| Statistical | City, work format, expected salary, schedule readiness | No |
Statistical questions are not included in the overall question table. Statistics can be reviewed for a candidate and for the interview as a whole.
What makes up the final score
The score shows answer quality inside a specific interview. For text questions, Brondem splits the answer into three components: correctness, relevance, and completeness. For test questions, scoring depends on the selected correct options, multiple choices, or correct order. The final score is a priority-weighted average of evaluated questions; statistical questions are excluded.
Correctness Correctness - whether the answer is factually correct and matches expected criteria.
Relevance Relevance - whether the candidate answered the question asked.
Completeness Completeness - whether the important elements were covered.
Relevance Relevance - whether the candidate answered the question asked.
Completeness Completeness - whether the important elements were covered.
example 88%
Checks how factually correct the answer is. When an expected answer exists, the system compares the candidate with the recruiter's benchmark: required concepts, examples, constraints, and typical mistakes.
example 99%
Checks whether the candidate answered the exact question instead of replacing it with general reasoning or another topic.
example 96%
Checks whether key elements are covered. Completeness is not the same as text length.
How to read the recommendation
70–100
Invite
The answers are worth considering for the next stage
40–69
Consider
Manual answer review is needed
0–39
Reject
Read the answers manually before making a decision
The AI-report recommendation helps identify what to review next, but it is not a final hiring decision. Misconduct and Risk signal do not recalculate the recommendation; they can change the manual-review guidance. They do not change the knowledge score, and the final decision is made by HR or the employer.
How to read Misconduct and Risk signal
Misconduct and Risk signal are two separate metrics, distinct from knowledge scoring. Their percentage values show signal intensity, not a statistical probability of a violation. Risk signal can be high for unusual behavior even when individual signs have limited evidentiary strength; Misconduct rises more conservatively with stronger combinations of signs.
Misconduct
A conservative measure of sufficiently strong technical and behavioral signs that may indicate non-independent completion or an attempt to influence evaluation.
Risk signal
A more sensitive manual-review indicator. It considers a broader range of technical, behavioral, timing, and contextual anomalies and shows how closely a completion should be reviewed.
Misconduct and Risk signal help recruiters identify answers and sessions that deserve closer review. They are not proof of a violation.
Across both channels, risk analysis still operates for bot completions, but Bot and Mini App use their actual available channel data. Partial coverage does not mean low or zero risk; when a complete numeric assessment is unavailable, the interface shows that status instead of a favorable 0%.
| What is visible in the card | How to read it |
|---|---|
| High score and low risks | The candidate can be reviewed first in the list. |
| High score and visible risk signals | Open the highlighted answers before moving forward. |
| Medium score and strong key answers | It may be useful to compare with other candidates for the same skill. |
| Low score and incomplete answers | The candidate likely does not meet the vacancy requirements. |
Where HR reviews the result after completion
Question details
Each question shows the prompt, expected answer, candidate answer, score, and explanation.
Candidate comparison
Candidates can be compared by score, key answers, and completion indicators.
History
The system keeps the history of a candidate and the full completion history inside an interview.
Export
Data can be exported for one candidate or the whole interview in CSV, Excel, PDF, and JSON formats.
Reminders
The team can return to important actions at the right time and avoid losing candidates.
Source links
Separate links help show where the candidate came from: site, channel, partner, or campaign.
Telegram bot and mini app
Mini App. It gives candidates a convenient way to complete an interview.
Telegram Bot. It is an alternative channel for completing an interview.
Regardless of the channel, knowledge evaluation remains the same and the final decision is made by HR or the employer.
Which limits should be considered
Questionsshort wording for one skill or one work scenario
Expected answerevaluation reference, not a text for literal repetition
Candidate answercontent matters more than length; artificial expansion does not improve quality
Question countdepends on the plan and company settings
Reportscounted separately from candidate interview completions
Repeated completionone candidate should not repeat the same interview
Files and exportsavailable capabilities depend on the plan
Mini appavailability depends on the plan and company settings
Candidate card review order
1. Open the overall score
It quickly shows the candidate's position in the flow.
2. Check key questions
Start with questions that matter most for the role.
3. Compare answers with expectations
The priority set by HR helps make review easier.
4. Check completion indicators
Risk signals help decide which answers need extra review.
Short answers about metrics
Why is an expected answer needed?
It sets evaluation criteria. Without it, evaluation is possible, but more often requires manual clarification.
Does the candidate need to repeat the reference literally?
No. The meaning, correct concepts, constraints, and practical logic matter.
Do statistical questions affect the score?
No. They are used for questionnaires, filters, and interview statistics.
Does the mini app change evaluation?
No. The Mini App does not change knowledge or content scoring.
Build an interview with clear criteria
The score sits next to the recommendation, and the items below show score components and separate completion indicators.