How to demo a product by showing it handle a wrong answer

Last updated 2026-09-22 · 8 min read

What is the short version?

You have a long screen recording of your product working normally. The thirty seconds worth posting are the ones where a user gets something wrong and the product handles it well, and that is the stretch Vizard Agent is looking for.

1. Tell Vizard Agent which wrong answer or failure to feature.
2. Give it the arc: mistake, hint, reconsideration, understanding.
3. Say the recording stays the main picture, with no generic imagery.

What do you need before you start?

The recording and the moment you want. Vizard Agent maps the screen states itself, but naming the misconception — the one your users always have — tells it which of the many exchanges in the recording is the one that sells.

• The screen recording. Unedited is fine.
• The failure to feature. The common wrong answer.
• The brand. Typeface, colour, logo.
• The platform. It decides the shape and the length.
• Whether there is speech. Or only interface sound.

What do you type into Vizard Agent?

Describe the journey rather than the feature. "Shows how it responds when a student gives the tongue-map answer" names a specific moment in the recording and a specific claim about the product, and both of those are checkable.

Prompt

Variants worth knowing:

• "The common but incorrect answer." Which failure, exactly.
• "Mistake, hint, reconsideration." The arc, in three beats.
• "No generic imagery." Keeps the demo a demo.

What does Vizard Agent actually do?

Here is the Vizard Agent sequence on a tutoring product's screen recording cut down into a short demo. Finding the four beats inside a long recording is the actual job here, and the edit that follows from them is very short indeed.

1. Checks the supplied logo.
2. Checks the recording's timing and extracts a visual overview.
3. Checks the available motion and caption treatments before building anything.
4. Reviews the overview to find the question, the misconception, the hinting and the correction.
5. Extracts larger checkpoints from the long recording to select the story beats.
6. Maps the exact on-screen progression from those checkpoints.
7. Locates the major screen-state changes in the recording.
8. Checks whether the source has usable speech or only interface sound.
9. Zooms into the key moments so the edit shows the student's actual answer and the product's hint.
10. Reads the opening question and first answer at full size.
11. Reads the tutor's first hint so the edit can emphasise its exact wording.
12. Reads the final explanation and the reconsideration at full size, and checks the closing state.

Steps ten to twelve are the difference between a demo and a screen recording with music. The words on screen are the product, so they are read at full size and the edit is built to make each of them legible at the moment it matters.

Step eight decides the whole treatment. A recording with narration can carry itself; one with only interface sound needs captions, and knowing which you have before building saves rebuilding it.

Step nine matters because screen recordings are wide and feeds are not. Zooming into the exchange keeps the words readable at the size the video will actually be watched.

What does the result look like?

Half a minute in which a real user gets something wrong, the product responds, and the user arrives at the right answer — in the product's actual interface, with the important wording legible, in your brand's typeface and colour.

When does this not work well?

Some products have no visible failure to show. If the interaction is fast, if the product's value is in the result rather than the exchange, or if the recording shows real user data, a different demo is needed.

• No visible exchange. The product simply returns a result.
• Real user data on screen. It has to be replaced or blurred first.
• Very long interactions. Thirty seconds cannot carry them.
• Interfaces with tiny text. Unreadable at feed size even zoomed.
• Failures that look bad. Some mistakes make the product look broken.

How do you fix a result that came back wrong?

Name the beat that is missing or that has been misread in the cut. Vizard Agent can re-cut around a single beat — the hint, say, or the moment of reconsideration — without rebuilding the rest of the demo around it.

• "The hint is cut too short." That beat is extended to its full wording.
• "You cannot read the answer." The zoom is tightened at that moment.
• "It starts too late." The question is included so the mistake makes sense.
• "The branding is wrong." The typeface and colour are applied from your assets.

How does Vizard Agent compare to doing it yourself?

By hand you record a clean run, because that is what a demo is supposed to be. It is also what every competitor's demo looks like, and it answers none of the questions a buyer actually has about what happens when things go sideways.

Common questions

Will it look negative to show a mistake? No. In Vizard Agent's cut the mistake is the user's and the recovery is the product's.

Can it blur real data? Yes, and Vizard Agent will flag anything identifiable it finds.

What if there is no narration? Vizard Agent captions the exchange instead, using the on-screen wording.

Can it add a voiceover? Yes, though on a text-based product the screen usually speaks for itself.

How long should a demo like this be? Thirty to forty-five seconds for a feed; tell Vizard Agent where it runs.

Can it use our fonts and colours? Yes. Give Vizard Agent the assets and it applies them throughout.

What aspect ratio for LinkedIn? Vizard Agent builds 4:5 or 1:1 for the feed, and will advise which.

Can it make several versions? Yes — one per misconception is a strong series.

Will the cursor be visible? Yes, and Vizard Agent can emphasise it where it helps.

Can it speed up the boring parts? Yes, though Vizard Agent keeps the hint itself at full speed.

What about sound effects? Light ones on the state changes work; Vizard Agent keeps them subtle.

Can it end on a call to action? Yes, as a card in your brand style.

Does it need the whole recording? Yes — Vizard Agent selects the beats from it rather than from a clip you chose.

How do I check it? Watch it muted at phone size; if the wording is unreadable, the zoom is wrong.