Brett Lindenberg
You know your audience understood your video when they can use its message: they answer a question about the key point correctly or explain the idea in their own words. Watch time cannot show this, because a viewer can play every second and still miss the point. This is Part 2 of a three-part series on whether your video content is working. Part 1 checked whether viewers saw the key moments. Here you write down what understanding looks like, ask for it at the moment the point is made, read the answers, and fix the part of the video that the wrong answers point to.
Why watch data and "Did that make sense?" fall short
A watched percentage tells you a viewer was present. It says nothing about what they took away. A product demo can hold a buyer for every minute and still leave them unsure how your product differs from the one they use now.
Asking viewers whether they understood does not fix this. In a 2002 paper in Cognitive Science, Leonid Rozenblit and Frank Keil showed that people believe they understand complex things in far more depth than they do, and called it the illusion of explanatory depth. The illusion was far stronger for knowledge of how things work than for facts, procedures, or stories, and strongest when the mechanism was visible. A product demo that shows a feature working on screen fits that description. A viewer who clicks "Yes, that made sense" may be sure and still be wrong. Ask them to use the idea instead.
A question in the video also helps the viewer pay attention. In a 2013 study in PNAS, Harvard researchers split a 21-minute statistics lecture into four parts. Students who answered test questions between the parts reported mind wandering at 19% of check-ins, compared with 41% for students who got no tests, and they scored higher on the last part of the lecture. These were college students in a lab, so expect the size of the effect to differ for your audience.
Step 1: Write down what understanding looks like
Start with the key moments you named in Part 1. For each one, finish this sentence: "A viewer who understood this can ___." The blank should be something a viewer does, such as choose or explain. "Knows about the feature" is too vague to measure.
| Video | The message | A viewer who understood can |
|---|---|---|
| Product demo | The product removes the weekly manual export | Pick, from four tasks, the one the product does for them |
| Comparison video | Plan B fits teams that need approvals | Choose the right plan for a team described in two sentences |
| Customer onboarding | Connect your data source before you invite your team | Put three setup steps in the right order |
| Safety training | Lock out the machine before you clear a jam | Choose the first action in a short jam scenario |
Scenario questions like the last two are better evidence than recall questions. A viewer can repeat a phrase they just heard without understanding it. Applying it to a new case is harder to fake.
Step 2: Ask at the moment the point is made
Put the question right after the explanation, at the point where viewers have enough context to answer. In Mindstamp, a question pauses the video when it appears and continues after the viewer answers. A question at the end tests memory of the whole video. A question at the moment tests that one point, and the answer tells you which part of the video to fix.
Match the question type to the evidence you need
Mindstamp has 13 question types. These are the ones that fit a comprehension check:
| Evidence you need | Question type | Scored |
|---|---|---|
| The viewer recognizes the right idea or applies it to a scenario | Multiple Choice (up to 5 answers) or Dropdown Select | Yes |
| The viewer knows every part of a rule | Choose All That Apply (exact set, no partial credit) | Yes |
| The viewer can explain it in their own words | Free Response, Voice Response, or Video Response (up to 120 seconds) | Free Response only. Voice and video answers are for you to review. |
| The viewer can point to the right place on screen | Drawing Response | No |
| How sure the viewer is | Rating (1 to 5 stars), asked after a scored question | No |
| Which idea landed for the audience, when there is no right answer | Display Poll, which then shows a chart of how all viewers answered | No |
A confidence rating is most useful next to a scored question. A viewer who answers wrong and rates their confidence at five stars holds a misconception, which is a bigger problem than a guess. See interactive video question types for all 13 types and how to create an interactive video quiz for the setup steps.
Write wrong answers that match real mistakes
Each wrong answer should be a mistake a real viewer could make, such as following the old process or skipping a step people often skip. When many viewers pick the same wrong answer, you learn which mistake the video failed to correct. A wrong answer that no viewer would choose adds nothing to the report.
Let viewers tell you what is unclear
Two features collect comprehension signals without a quiz:
- A button with the Collect Feedback click action shows your prompt and a box for a short response. Label it something like "Something unclear? Tell us." The full list is in the guide to click actions.
- Genie AI answers viewer questions inside the player, using only the video's transcript and the notes you add. The Genie AI tab lists every viewer question with Genie's answer. Filter it to Unanswered to see what viewers wanted to know that the video does not cover. Genie is on supported plans.
Step 3: Read the answers
In Mindstamp, open Reporting. Each page answers a different question:
- Questions shows how all viewers answered each question. Start here.
- Views has one row for each viewing session, with its participation and correctness scores.
- Interactions lists each answer with its prompt and value, and plays back voice and video responses.
Participation is the share of questions a viewer answered. Correctness is the share of their answers that were correct, and it counts every attempt. A viewer who answers wrong and then right scores 50% on that question. Read the two scores separately: low participation means viewers skipped the questions, and low correctness means they answered and got them wrong. The guide to Mindstamp reports defines each figure.
Viewers are anonymous until you identify them. To see which people understood, identify viewers with lead capture, a personalized link, or URL parameters, as the guide to viewer identity explains.
What the pattern of answers tells you
| What you see | What it usually means | What to change |
|---|---|---|
| Most viewers answer correctly | The point landed | Keep it. Move on to Part 3 and check whether viewers acted |
| Many viewers pick the same wrong answer | The video leaves one specific misconception in place | Find the sentence that causes it. Rewrite it, or add on-screen text or a drawing at that moment |
| Wrong answers are spread evenly | The question is unclear, or the video never made the point | Reword the question, then check that the explanation comes before it |
| Many viewers skip the question | It feels optional, or it comes at a bad time | Move it closer to the explanation. If the answer matters, turn off Allow skipping |
| Correct answers with low confidence ratings | Viewers guessed right | Add a worked example before the question |
| Unanswered Genie questions on the same topic | Viewers need information the video leaves out | Cover it in the next version of the video, or add it to Genie's notes |
Decide which measure you will act on before you open the report, and wait for enough views before you change the video. A report with many columns makes it easy to find a number that agrees with what you already believed. Change one thing at a time so you know what made the difference. To practice reading this kind of evidence, try the knowledge gap diagnosis example, where learners replay one step and many miss the question after it.
Step 4: Help viewers who missed it, inside the video
A wrong answer can do more than mark a score. In Mindstamp, each answer can run an action. Choose one by how much the point matters:
- Show Message displays your explanation, and the video continues when the viewer closes it. Use it when the point is useful but not essential.
- Change Time jumps back to the explanation, so the viewer watches it again and meets the question again.
- Repeat Until Correct shows Try Again and the same question until the viewer gets it right. It needs a scored question type, with Allow skipping and Time Limit turned off and Show Correctness on.
- Play Video Clip, on a button, pauses the video and plays a short clip, such as a second example for viewers who want more detail.
- Switch Video opens a separate review video. A button with Switch to Previous Video at its end returns the viewer to the video they came from.
To let the viewer choose whether to review, store the answer in a variable and show a "Try again" button only after a wrong answer. Our try again button guide has the steps. Variables and conditional logic are on supported plans. Answer actions and Repeat Until Correct need neither.
Fit the response to the audience. In required training, send a wrong answer back to the lesson. In a marketing or sales video, offer an optional "See an example" button and let the viewer keep going, since they chose to watch and can leave at any time.
Common questions
How many questions should one video have?
Ask one question for each key moment you named in Step 1. Each question then has a reason the viewer can see, and each answer tells you about one part of the video. Questions added only to raise interaction counts make the report harder to read.
Will questions make viewers leave?
A question asks for effort, so put it where answering gives the viewer something back, such as feedback or a path that fits them. In our own data, across 8.07 million views of 18,315 Mindstamp videos (2018 to October 2026), videos where viewers took more than three actions per view averaged 51.8% watched. Videos where viewers took no action averaged 43.9%. These videos differ in many ways, including length, audience, and purpose, so treat this as a pattern, not proof that questions keep viewers watching.
Can I check understanding without a quiz?
Yes. A Display Poll on the main idea shows which idea landed, and Genie's list of viewer questions shows what the video left out. In a branching video, the path a viewer chooses also shows what they took from the part before it. For a full feedback form, see how to create a video survey. For training programs, our list of sample evaluation questions covers questions to ask after the course.
Can AI write the questions?
In Mindstamp, Generate questions on the Transcript page suggests quiz questions from the video's transcript. You still need to check that each draft tests the point from Step 1 and that its wrong answers match real mistakes. See how to turn any video into a quiz using AI.
The three questions in this series
- Did your audience see the most important content?
- Did your audience understand the message? (this post)
- Did viewers take the action you wanted?
Understanding is the link between the other two. If viewers reach a key moment but answer the question about it wrong, the fix is in the video. If they answer right but do not act, the fix is usually in the offer, which Part 3 covers.
Start with one video. Pick its most important moment, write the sentence "A viewer who understood this can ___," and add one question that asks the viewer to do exactly that. As views come in, open the Questions report and look for the most popular wrong answer.
Sources and further reading
- Rozenblit, L., and Keil, F. (2002). The misunderstood limits of folk science: an illusion of explanatory depth. Cognitive Science, 26(5), 521 to 562.
- Szpunar, K. K., Khan, N. Y., and Schacter, D. L. (2013). Interpolated memory tests reduce mind wandering and improve learning of online lectures. Proceedings of the National Academy of Sciences, 110(16), 6313 to 6317.
