Assessing learning with interactive video: how to check understanding and retention

Why completion does not show learning

A completion record shows that the video reached its end or a set checkpoint. It does not show what the learner can do. If the seek bar is on, a learner can drag it to the end and reach the completion point without watching. In Mindstamp, the watched percentage counts only the seconds a learner actually played, so the report shows that gap, and the settings that stop skipping ahead close it.

Evidence of learning is something the learner does with the content. The Kirkpatrick model calls this Level 2, Learning: the degree to which people acquire the intended knowledge, skills, attitude, confidence, and commitment. Kirkpatrick Partners notes that pre- and post-tests are the traditional measure, and that role plays and teach-backs suit some content better. Inside a video, those become questions, scenario choices, and recorded answers. For other assessment types, such as diagnostic pre-tests and portfolios, see our guide to learning assessment strategies. For the case for interactive training in general, read five reasons interactive video works for training.

Match each check to the learning objective

Start from the verb in the objective. "List the steps" needs a different check from "decide what to do" or "explain it to a customer." The table shows which in-video check fits each kind of objective and what you get back.

In-video checks by learning objective, with Mindstamp scoring and reporting (checked October 2026)

The learner should be able to	Check inside the video	Scored automatically?	What the report shows
Recall a fact, value, or step order	Multiple choice, dropdown, number, or date question	Yes	Each answer, right or wrong, and every attempt
Apply a rule to a case	A short case in the video, then a multiple choice or choose-all question	Yes (choose-all needs the exact set)	The answer, including which wrong option the learner chose
Decide what to do in a situation	Branching scenario: each choice plays its consequence	Yes, when the choices are a scored question	The answer and the path the learner took
Explain why	Free response or voice response question	Free response only, against the accepted answers you list	The typed or recorded explanation
Spot something on screen	A hotspot the learner must click, or a drawing response	No	Each click with its time, or the learner's drawing
Perform a spoken skill, such as a pitch	Video or voice response, up to 120 seconds	No	A recording to review with a rubric
Report confidence or intent to apply	Likert scale or rating question	No	The spread of answers across learners

Six of the 13 Mindstamp question types can be scored: free response, multiple choice, choose all that apply, dropdown, number, and date. The others record the response without a score. That suits explanations and skills, which a person should judge. To see how the choice of question changes what you learn, try the question configuration clinic, a short example that compares prompts for the same objective.

Write questions that test understanding

A question that repeats a sentence from the script tests whether the learner heard it. A question that asks the learner to use the idea in a new case tests whether they understood it. Compare two checks on a discount policy:

• Recall: "What is the largest discount a rep can offer without approval?"
• Application: "A buyer on a two-year contract asks for 18% off. Policy allows 10% off for terms of one year or longer. What do you do first?"

The University of Waterloo's Centre for Teaching Excellence publishes rules for designing multiple-choice questions that apply directly to checks inside a video:

• Put the whole problem in the question. A learner who knows the material should be able to answer before reading the options.
• Use three to five options. Waterloo notes that research finds three-option questions about as effective as four or five, mostly because good wrong answers are hard to write. A Mindstamp multiple choice question takes up to 5 answers.
• Build the wrong answers from real mistakes. When each wrong option is an error learners actually make, the answer tells you which mistake to fix.
• Avoid negative wording such as "Which of these is NOT ...". Learners who know the material still miss negatively worded questions.

Write feedback for every answer, including the right one. In a 2008 study, Butler and Roediger found that feedback after a multiple-choice test raised the share of correct answers on a later test and cut how often learners repeated the wrong options they had seen (Memory & Cognition, 2008). Keep each explanation to the reason the answer is right or wrong, then return the learner to the lesson.

To draft questions from a transcript, try our free video knowledge check generator, which also suggests where each check belongs, or the training quiz generator. Review every drafted question before you use it.

Place each check right after the segment it tests

Put each question where the learner has just seen the idea. The explanation is still fresh, and a wrong answer can send them straight back to it. Research on recorded lectures supports this. In a 2013 Harvard study, students watched a 21-minute statistics lecture split into four segments of about 5.5 minutes each. In the second of two experiments, students who took a short test after each segment reported mind wandering on 19% of attention probes, compared with 39% for students who restudied the material and 41% for students who did neither. The tests also improved learning (Szpunar, Khan, and Schacter, PNAS 2013).

A quiz saved for the end also misses the learners who stop early. Across 8.07 million views of 18,315 Mindstamp videos (2018 to October 2026), the average view played 46.5% of the video. These are mostly training, marketing, and sales videos made by our customers, and many factors differ between them, so treat the figure as a pattern, not a benchmark for all video.

For required training, make the checks impossible to skip. In Mindstamp, turn off Allow Seeking, turn off Allow skipping on each question that must be answered, and turn on Prevent passing required interactions so nobody can drag past a question. Let the content set the spacing: one check at the end of each key idea, wherever it falls on the timeline.

Decide what a wrong answer does, and what the score means

Each answer can trigger an action. Choose the one that fits the purpose of the check:

• Show a message that explains the answer, then continue. This suits practice questions.
• Turn on Repeat Until Correct. The learner sees Try Again and answers the same question until they get it right, without being shown the answer.
• Send a wrong answer back to the explanation with Change Time, so the learner watches the segment again and then meets the question again.
• Send a wrong answer to a short remedial lesson, either a later section of the same video or another video with Switch Video. The warehouse safety skill router example sends each answer to a different level of support before a follow-up question.

These choices change what the score measures, so decide which number is your evidence before launch. The same score appears in Mindstamp reports and in the SCORM package your LMS receives, and conditional logic can read it while the video plays. Two settings in the video's Settings control it:

• Questions in score. With All questions, the default for videos created since October 2026, every scored question counts, so a question the learner skipped or never reached counts as wrong. Conditional questions do not count. With Answered questions only, the score covers just the questions the learner answered.
• Answer that counts. This picks the first answer (the default) or the most recent one when a learner answers a question more than once. With Repeat Until Correct turned on, the most recent answer is always right, so every learner who finishes scores 100%.

In a branching video, All questions also counts the questions on paths a learner did not take, which lowers the score of a learner who chose well. Use Answered questions only for a branching check. Our help article on passing scores and retries has the details.

First-attempt correctness tells you what learners knew when they met the question. Final correctness tells you whether the feedback fixed it. The Interactions report keeps every attempt as its own row, so you can calculate both. There is no single pass mark for a video. You set a minimum score where it is used: on a completion certificate, on completion emails, or on the completion report. On Pro and higher plans, conditional logic can read the running score and show a different ending to a learner below your pass mark.

Assess decisions with a branching scenario

Many jobs need judgment more than recall. A branching scenario tests judgment: the video sets up a realistic moment, pauses, offers several plausible responses, and plays the consequence of the one the learner picks. The discount demo at the top of this page works this way. Each response opens a different result scene, and the recap shows the route you took.

To get a score as well as a path, build the decision as a multiple choice question. Mark the best response as correct and give each answer its own Change Time action to the matching consequence scene. The report then holds the score, the chosen answer, and the route. A few rules make scenario checks worth the effort:

• Make every option something a real employee might do. An option nobody would pick tells you nothing.
• Play the consequence first, then the coaching that explains it.
• End with a recap that names the principle, and let learners replay to try another choice.
• Keep the first scenario to one decision. Add a second decision after you see how learners handle the first.

For worked examples, try the role-based sales practice case, which changes the scenario by job role. Our guide to scenario-based training covers how to write the scenes, the free branching scenario generator drafts a decision tree, and the video branching page shows the setup.

Check what learners notice with hotspots

Some objectives are about attention: spotting a hazard, finding a setting on screen, or noticing a buying signal. A hotspot turns that into a task. Place a hotspot over the hazard, show it for a short window, and set it to require a click, so the video waits at the end of the window until the learner finds it. Mindstamp records each click with its time, so you can see who found the target while the scene played and who needed the pause.

Clicks are not marked right or wrong. If you need a score, follow the task with a scored question, or ask for a drawing response ("circle the hazard") and review the drawings. In sales training, the same pattern works for buyer cues: put a hotspot on the buyer while they ask about rollout dates and see which reps click it. Our help article on requiring a click before the video continues covers the setup, and the hotspots page shows more uses.

Assess spoken skills with recorded answers

When the objective is a conversation, such as a pitch, an apology to a customer, or a safety briefing, ask learners to show it. A Mindstamp video or voice response question lets the learner record an answer of up to 120 seconds in the browser, with nothing to install. The recording is saved with the learner's record, and you play it back from the Interactions report. These answers are not scored automatically, so review them against a short rubric. Our free rubric generator drafts one from your criteria. A recording of a face or voice is personal data, so tell learners who will watch it and why.

How to verify that knowledge lasts

A score taken minutes after a lesson is a weak guide to what learners keep. In Roediger and Karpicke's 2006 experiments, students who restudied a passage did better on a test 5 minutes later, but students who had practiced recalling it did better on tests 2 days and 1 week later (Psychological Science, 2006). A check at 5 minutes would have picked the wrong study method. To verify retention, measure again after a delay:

1. Write a short follow-up check with new questions on the same objectives. Use new cases, so learners must apply the rule again instead of remembering the earlier answer. Turn off Repeat Until Correct on this check, so each score reflects one attempt.
2. Time it to how long the knowledge must last. In a 2008 study of more than 1,350 people, Cepeda and colleagues found that the best gap before a review grew with the length of retention needed: about 20% to 40% of the interval for a test one week later, and about 5% to 10% for a test one year later (Psychological Science, 2008). If learners must remember a procedure for a year, that points to a follow-up roughly 2.5 to 5 weeks after the first training.
3. Identify learners the same way both times. Use the same email address or employee ID, through lead capture, a personalized link, or your LMS, so the completion report can group both views under one person. Anonymous views on different devices cannot be joined later.
4. Compare the results. A completion report across both videos gives one row per learner with a score for each video. For detail by objective, export the views with their answers and compare the questions that test the same objective.

The follow-up is also practice, because each recall attempt strengthens memory. Our guide to knowledge retention strategies covers spacing and other ways to make training stick.

How to read the results and fix the training

Open the results after the first group of learners finishes. In Mindstamp, Reporting > Questions shows how all learners answered each question, and Reporting > Views shows each session with watched percentage, participation, and correctness. Read them at three levels.

Each question

The share of learners who answer a question correctly is its difficulty. The University of Washington's item analysis guide classes a question as hard when 50% or fewer answer it correctly. It also gives the ideal levels for separating strong and weak learners: about 77% correct for a three-option question, 74% for four options, and 70% for five. Those targets come from tests built to rank people. A training check usually aims for most learners to get it right after the lesson, so use the numbers to spot problems, not as goals.

• When many learners miss a question, look at the segment before it first. The explanation is the usual cause.
• When one wrong option draws many answers, you have found a shared misconception. Name it in the lesson and in the feedback.
• When strong learners miss a question more often than weak ones, the answer key may be wrong or two answers may be defensible. The Washington guide says a negative discrimination value may mean the item was mis-keyed. To check, export the views with their answers and compare how your top and bottom scorers answered.

Each learner

Participation and correctness answer different questions. Participation is the share of questions a learner answered. Correctness is the share of answers that were right. High participation with low correctness means the learner engaged and still misunderstood, which calls for coaching. Low participation usually means the learner left early, so check the watched percentage and the furthest point they reached.

Each section of the video

On Core and higher plans, advanced reporting shows where learners drop off. A steep drop inside one segment suggests it runs long or loses the thread. Our guide to measuring training effectiveness covers what to measure beyond the video, such as behavior on the job.

What to look for in a platform for in-video learner checks

If you are choosing software, these features decide whether a training video can produce assessment evidence for a large group:

• Scored question types, feedback for each answer, and a record of every attempt.
• Required answers, and a way to stop learners from dragging past questions.
• Answer actions that replay a segment or open a scenario branch, set up in an editor without re-editing the video.
• Learner identity without new accounts, from an LMS launch, an email address, or an employee ID.
• Exports with one row per learner across a course, plus every answer to every question.
• Delivery through your LMS, usually as a SCORM package.
• Pricing you can predict as the audience grows.

Mindstamp covers this list on supported plans. It adds interactions on top of video you already have, from an upload or a host such as Vimeo, Wistia, or Kaltura, so nobody needs video editing skills to build a check. It exports SCORM 1.2 and SCORM 2004 (3rd and 4th edition) packages. Your LMS receives completion and a score, and the individual answers stay in Mindstamp reports. Plans have no per-viewer charge. They bill by videos, admin seats, and streaming minutes, so plan for the minutes your learners will watch. Before a large rollout, also know what it does not do: Mindstamp is not an LMS, it does not send xAPI or cmi5 statements, and it does not publish a WCAG conformance level or a VPAT. Our overview of how Mindstamp works for employee training covers the setup, and our comparison of video assessment tools covers other options.

Common questions

Which video training platform offers built-in quizzes and learner checks to verify knowledge retention?

Look for an interactive video platform that puts scored questions inside the video, records every attempt for each learner, and reports one learner across several videos, so you can compare the first training with a later follow-up check. Mindstamp does this. Questions pause the video, answers are scored and saved against the learner, and the completion report puts each learner's score for every video in one row. To compare other platforms, see our guide to interactive video software for corporate training.

How can an L&D team scale interactive training video with branching and quizzes to 500 or more employees?

The build effort depends on the number of modules, not the number of learners. Start from recordings you already have, because the interactions sit on top of the video and nobody has to re-edit it. Use the same pattern in every module, for example a check after each key segment and one scenario decision, so new modules go faster. Deliver through your LMS with a SCORM package, or share links that pass each employee's email or ID, and put each course in one playlist or group for reporting. Plan streaming minutes too: 500 employees who each watch a 12-minute module use about 6,000 minutes. Check plan capacity on the pricing page. For live and self-paced formats, see our guide to interactive online training.

How is an in-video check different from an LMS quiz?

An LMS quiz usually comes after the video and records a score in the gradebook. An in-video check comes at the moment of teaching, can send a wrong answer back to the explanation, and records the answer with its place in the video and the path the learner took. With a SCORM package you get both: the LMS records completion and a score, and Mindstamp keeps the answer-level detail. The LMS and video evidence example shows which record answers which audit question.

What is an interactive assessment?

An interactive assessment asks the learner to act inside the content, by answering, choosing, clicking, or recording, and responds to what they do with feedback or a different path. In a video, it is the difference between a quiz page after the lesson and a question that stops the lesson, checks the answer, and decides what the learner sees next. When the purpose is to adjust teaching while learning is still happening, it is a form of formative assessment.

Can in-video checks support compliance training records?

Yes, if every learner is identified. Mindstamp can export a completion report with one row per learner and a score for each video, or a views export with each answer. A completion certificate can require a watched percentage and a passing score, and each certificate has a public verification link. Our help article on exporting training records for an audit lists the columns. Your LMS or HR system usually stays the official record, so decide early which system an auditor will see.

Start with one module

Pick one training video with a clear objective. Add a scored question after each key segment, one scenario decision, and a confidence question at the end. Share it with a small pilot group, read the question breakdown, fix the weakest segment, and schedule a short follow-up check for a few weeks later. You can build it during Mindstamp's 14-day free trial, which needs no payment card. Our step-by-step guide to creating an interactive video quiz walks through the setup.