Read the delivery numbers first, then the weakest moment and ignore the score until you have a second take to compare it with. A single score tells you little. The change between two takes of the same question tells you whether your fix worked.
What is in a report
| Part | What it is | Trial | Paid |
|---|---|---|---|
| Score | An integer from 0 to 100 and a two-sentence summary. | Yes | Yes |
| Coaching points | Specific feedback tagged structure, content, delivery or keywords, referring to your own words. | 3 | 5 to 7 |
| Delivery numbers | Pace, filler words and long pauses, measured from your recording. | No | Yes |
| Role-specific feedback | The strongest gap in the answer and what an interviewer for the role would probe. | No | Yes |
| Weakest moment | The place the answer lost credibility, quoted, with the reason. | No | Yes |
| Rewrite | Your weakest sentence rewritten as a stronger one. | No | Yes |
| Missed keywords | Terms interviewers expect for the role that your answer did not use. | No | Yes |
| Follow-up question | The sharpest question an interviewer would ask next. You can practice it straight away. | No | Yes |
| Ideal answer | A model answer to the same question, to read next to your own. | No | Premium and single report |
Every account gets 2 trial answers a month. See the plans.
How the score is decided
Your recording is transcribed and a language model reads the transcript against the question and the role. It is told how to read a score:
| Score | Meaning |
|---|---|
| 90 to 100 | Would pass a Bar Raiser |
| 70 to 89 | Solid but probing would expose gaps |
| 50 to 69 | Hand-waving detected |
| Below 50 | Significant structural or depth failure |
On paid plans the model is also told to check four things, which is why paid feedback is harder to satisfy:
Individual ownership
Saying "we" for actions you took is flagged. The interviewer wants to know what you did.Specific metrics
A claim like "improved performance significantly" is challenged. It wants a baseline and an outcome.Story structure
Whether the situation, the action and a measurable result are all present, or one of them is skipped or buried.Rehearsed language
Stock phrases are separated from specific storytelling. Technical questions are checked for trade-offs and depth.
The score is the model's judgment of your words. It has not been validated against hiring outcomes and two runs on the same answer can differ a little.
Reading the delivery numbers
Pace
Words per minute. The tool treats 100 to 160 as in range and highlights anything outside it.Filler words
Counts "um", "uh", "you know" wherever they occur. Words like "so", "like" and "right" count only when the transcript sets them off with a comma, so ordinary uses are not counted. The transcript's punctuation is automatic, so a few will be missed or miscounted.Long pauses
Silences of 2 seconds or more between words. One or two to think is normal. Several can mean you lost the thread.
An example report
An illustrative report for a Software Engineer answer. The numbers are made up to show the shape of a report, not a real person's result.
Content
Add one number to your main example, such as time saved, error rate or volume handled. A figure is easier to remember than "a big improvement".
Keywords
You did not mention "database indexing" or "caching". Interviewers for this role listen for both.
Delivery
You said "um" 3 times, mostly at the start of sentences. A short silent pause works better.
- Pace
- 159 words per minute, upper end of the ideal range
- Filler words
- 4: "um" 3, "like" 1
- Long pauses
- 3
What a report cannot tell you
Whether you will get the job
The score is not a prediction. Interviewers weigh things a transcript does not hold.How a particular employer will judge you
The rubric is general. A company can value different things from the ones the model looks for.Whether the feedback is right
It is a model's reading of your words. Treat the specific, quoted points as suggestions to test on the next take, not as facts.
Common questions
Can I compare a trial score with a paid score?
Not directly. Trial feedback comes from a plain coaching prompt and paid feedback from a stricter one that plays a rigorous interviewer who challenges vague claims and shared credit. The same answer can score differently, so compare scores within one plan.
What does a score of 70 mean?
On the scale the model is given, 70 to 89 means solid but probing would expose gaps. A 70 is at the bottom of that band: a reasonable answer that follow-up questions would start to pull apart.
Does a higher score predict getting the job?
There is no evidence for that. The score is a language model's judgment of your transcript against a rubric. It has not been checked against hiring outcomes and the practice data we hold cannot show it.
Does the report look at my video?
It is built from what you said and when you said it: the transcript and the timing of your words. Your audio is transcribed and discarded and recordings are never stored. Read the privacy policy.
Keep going
How to practice for a video interview
The routine that turns a report into a better next take.
Practicing for an interview with AI
What employers allow at each stage, from their own published rules.