Custom Software Development

Beyond the Video Lesson: Designing Practice, Feedback, and Progress

A video can explain a skill. The activity around it is where learners use it, find out what to fix, and try again.

The 30-second version
  • Watching a lesson doesn't show that a learner can use it. The activity after the video is where that evidence comes from, and it's where most of a learning platform's software lives.
  • Start with an observable objective, then design the second attempt: what the learner sees after a wrong answer, what they try next, and what gets recorded.
  • Feedback isn't automatically good. In a review of 607 feedback effects, more than a third made performance worse. Feedback about the task helps; feedback about the person tends not to.
  • Completion is a record, not a result. Track hinted and independent success separately, and check skills on new examples, because what feels like progress during practice often isn't.
  • Before building anything, answer the seven questions in our Activity Blueprint. Often an existing LMS can host the activity; the custom work is the activity itself.

Direct answer: To design interactive learning activities, start with an observable learning objective. Choose a task that lets learners demonstrate it, plan specific feedback for each kind of mistake, and give learners a chance to use that feedback on a new attempt. Then decide what evidence of progress to record, such as hinted versus independent success, and check whether your existing platform can support the experience or needs to be extended.

Here's a situation we see a lot. An education business has great lessons. The videos are well produced, the curriculum works in a classroom, and families or schools are paying for it. Then someone asks a simple question: what should happen after the learner finishes watching?

The usual answer is a quiz. Sometimes that's exactly right. But "add a quiz" skips the decisions that make practice work: what the learner is actually doing, what they hear back when they miss, and what happens next. Those decisions are also where most of the software gets built.

At Virgo Development, we've built interactive learning platforms for My Music Workshop and Miss Humblebee's Academy. This article goes inside one activity. When Your Curriculum Becomes Software covers whether to build, and How to Plan a Custom Learning Platform traces one lesson across your whole business. Here, we'll take one lesson and design the activity that comes after it, step by step.

Getting learners to that activity matters, too. In one large study of edX courses, the share of learners who went on to try the practice problem dropped steadily as videos got longer.

Fewer Learners Reached the Practice After Longer Videos Share of edX video sessions followed by an attempt at the next practice problem, by video length 0% 50% 75% 56% 48% 43% 41% 31% Shortest 0–3 min 2nd group 3rd group 4th group Longest Based on 6.9 million video sessions in four edX courses. The study shows a pattern, not proof that length caused it. For design: treat the step after the video as its own experience, not a link at the bottom of the page.
Figure 1 — Practice attempts after videos of increasing length, grouped into five roughly equal-sized length buckets. Source: Guo, P. J., Kim, J., & Rubin, R. (2014), "How Video Production Affects Student Engagement: An Empirical Study of MOOC Videos," Proceedings of the First ACM Conference on Learning @ Scale, 41–50. The paper states the shortest bucket (0–3 minutes) and the five percentages; the other bucket ranges are shown only in its chart, so they're labeled by order here.

Those are university-level online courses, not a children's app, and the numbers describe a pattern rather than a cause. But the design point carries over: practice doesn't happen just because it's available. It needs the same care as the lesson. Every well-designed activity we've worked on follows the same loop: objective → practice → feedback → retry → progress. Let's build one.

01 — The objective

What Should the Learner Be Able to Do?

"Understand patterns" is a fine curriculum goal, but software can't observe understanding. It can only observe what a learner does. So the first design step is to rewrite the objective as an action you could watch:

After this activity, the learner should be able to ___, with ___ level of support.

For our running example, a lesson on repeating patterns for early learners, that becomes: "The learner can extend an AB or ABB pattern by choosing the next shape, without a hint." Now everyone can see what success looks like, including the developers.

This step also sorts out what the software can judge on its own and what needs a person. Choosing the next shape is easy to check automatically. Explaining why the pattern repeats, out loud, to a parent, is better left to a human, and the platform's job there is to prompt the conversation, not to grade it.

02 — The worked example

How Do You Turn a Lesson into an Interactive Learning Activity?

Here's one short lesson turned into an activity, with the decision made at each step and what it asks of the software. This is an illustrative design, not a client's shipped feature.

Design elementIllustrative decisionWhat the software has to support
LessonA two-minute video: circle, square, circle, square… "What comes next?"Short video segments tied to a skill, not one long recording
ObjectiveExtend an AB or ABB pattern without a hintActivities tagged by skill, so results roll up to the skill, not just the lesson
Worked exampleThe narrator solves one pattern, thinking out loudA "watch me" step the curriculum team can author and reorder
PracticeFive new patterns; the learner picks the next shape from three optionsA pool of items per pattern type, drawn so no two learners get an identical set
FeedbackSpecific to the mistake (see the next section)Each wrong answer classified by kind, with feedback written for each kind
RetryA new pattern of the same type, with a hint ready if neededRules for when to repeat, when to hint, and when to move on
ProgressIndependent success recorded separately from hinted successAn attempt record that stores support level, not just right or wrong

Notice what the practice step is not about. Tapping a shape is just the input method. The learning is in the decision: the learner has to spot the repeating unit and apply it to a pattern they haven't seen. That's also why the practice uses new patterns instead of replaying the one from the video. Recognizing an answer you've already seen is a lot easier than producing one.

The worked-example step is deliberate, too. The U.S. Institute of Education Sciences' practice guide on organizing instruction recommends alternating worked examples with problems learners solve themselves, and rates quizzing to retrieve what was learned among its best-supported strategies. In software terms, both are just steps in the activity's sequence, which is why that sequence needs to be something your curriculum team can edit.

03 — Feedback and the second attempt

Meaningful Feedback Is More Than Right or Wrong

It's tempting to treat feedback as a solved problem: green check, red X, done. The research says feedback deserves more care than that, because it doesn't always help.

Feedback Helped on Average. More Than a Third of the Time, It Backfired. 607 effects of feedback interventions on performance, from 131 studies Improved performance, or no change Made performance worse ≈ 62% 38%+ "over one-third" of effects Tended to help Feedback that keeps attention on the task: what was wrong and what to try next Tended to hurt Feedback that shifts attention to the person: praise, comparison, or a verdict on ability For design: write feedback about the work, not the worker, and give the learner something specific to do next.
Figure 2 — How often feedback interventions helped or hurt performance. Source: Kluger, A. N., & DeNisi, A. (1996), "The Effects of Feedback Interventions on Performance," Psychological Bulletin, 119(2), 254–284. The abstract reports an average effect of d = 0.41 and "over 1/3" negative effects; 38% is the share reported in summaries of the paper (e.g., The Open University). The "help vs. hurt" pattern summarizes the authors' feedback intervention theory. The studies cover many kinds of tasks and settings, not only schools.

Education research tells the same two-sided story. The Education Endowment Foundation rates feedback as one of its highest-impact, lowest-cost approaches, about six extra months of progress over a year on average across 155 studies. It also notes that digital feedback has typically shown smaller effects, around four months, and that feedback works best when it gives specific information about how to improve. The format doesn't do the work. The design does.

So instead of one "wrong" state, our pattern activity plans for each kind of response:

Learner responseWhat it may meanIllustrative feedbackWhat happens next
Correct, first tryLikely spotted the repeat"Circle, square, square. You found it."A new pattern; a different type later
A shape from the pattern, wrong spotSees the shapes, missed the repeating unit"That shape is in the pattern. Look at what comes after each circle."Try again with the repeating unit highlighted
A shape not in the patternMay not understand the task yetReplay the short spoken instruction and the worked exampleA supported item, then a fresh one
No responseStuck, unsure what to do, or interface troubleA gentle spoken prompt after a pauseLogged separately; not counted as wrong
Rapid guessingTapping until something worksA short pause: "Let's say the pattern together."Success recorded as supported, not independent

That last column is the second attempt, and it's the part most often left out. The EEF's guidance on feedback stresses giving learners a real chance to act on it. In practice, that means a new item of the same kind, not just a second click on the one they missed. Otherwise "correct" may only mean the learner eliminated the wrong options.

Hints follow the same logic. A simple hint ladder might go: highlight the repeating unit, then say the pattern aloud with the learner, then show the answer with an explanation and move to a fresh pattern. Each step gives a bit more help, and the platform records how much help was needed. Timing is a design choice as well: a child mid-task usually benefits from a quick nudge, while a longer project might get feedback afterward. There's no single rule that fits every activity.

04 — Evidence of progress

Track Learning Progress, Not Just Lesson Completion

"Lesson complete" is the easiest thing for a platform to record and the easiest thing for a parent to misread. It's a useful record, but each piece of data only proves so much. Here's the framework we use, which we call the Evidence Ladder:

Recorded eventWhat it tells youWhat it doesn't prove on its own
Video completedThe lesson reached its completion conditionThe learner can apply the skill
Activity attemptedThe learner participatedAny response was accurate or independent
Correct after a hintThe learner succeeded with supportThey can do it without help
Correct on a new pattern, no hintIndependent performance carried over to a new exampleThey'll remember it next week
Correct on a later checkRetention, on that checkComplete mastery of the subject

This is a planning framework, not a universal assessment standard. Its job is to stop a platform from calling a learner "proficient" because they finished something. The higher rungs need new examples and a later check, because how well practice feels like it's going is a poor guide to what's been learned.

What Felt Like Progress Wasn't Learning to recognize 12 painters' styles, then naming the painter of paintings learners hadn't seen CORRECT ON NEW PAINTINGS Blocked by painter 35% Mixed together 61% SAME LEARNERS 78% scored better when examples were mixed together 78% said blocked examples worked as well or better For design: mix pattern types in practice, and measure progress on new examples, not on how practice felt.
Figure 3 — Blocked vs. mixed (interleaved) examples, and learners' own judgments. Source: Kornell, N., & Bjork, R. A. (2008), "Learning Concepts and Categories: Is Spacing the 'Enemy of Induction'?" Psychological Science, 19(6), 585–592, Experiment 1a (120 undergraduates; percentages correct are from the first test block). Later studies, including with children and older adults, have found similar results.

Recognizing a painter's style is pattern recognition, which is exactly what our activity teaches. So the design takes two cues from it. Once learners have the basics, practice mixes AB and ABB patterns together instead of running all of one type first. And progress is judged by results on new patterns, not by a smooth streak of right answers on familiar ones.

In the software, that means each attempt saves more than true or false. A simplified, illustrative attempt record might look like this:

attempt
  learner          child profile (under a family account)
  activity         extend-the-pattern
  item             ABB-07          // which pattern, not just which lesson
  skill            patterns.extend.ABB
  result           wrong_position  // the kind of answer, not just right/wrong
  hints_used       1
  attempt_number   2
  support          hinted          // hinted | independent
  seen_before      false           // a new example?

From records like that, each person can get the view they need. The learner sees what's next. A parent sees growth on a skill, not a row of green checks. A teacher sees which learners needed hints on ABB patterns this week. An administrator sees which activities produce the most "no response" events, which is often a sign of an interface problem rather than a learning one.

05 — Real learners

Design for the Students Who Will Actually Use It

All of this assumes the learner understands what the activity is asking them to do. That's an assumption worth testing early, with the people who will actually use it.

Engineer Ian Anderson learned this firsthand on Miss Humblebee's Academy, a learning platform for children ages three to six. Before releasing changes to a puzzle activity, he asked his five-year-old son to try it. Ian expected a click on a puzzle piece, then a drag. His son skipped the click and dragged the piece straight from the sidebar, which the activity wasn't built to handle. That one test changed the build before paying customers saw it. The full story is in The Path the Engineer Expected vs. the Path the User Took.

For an activity like ours, testing with real learners answers practical questions:

  • Reading: Can a pre-reader follow the instructions? Spoken instructions and a replay button may matter more than any text.
  • Input: Do learners tap, drag, or both? Supporting both is usually cheaper than teaching one.
  • Cues: Does anything depend on color or sound alone? Shapes that differ in form, not just color, work for more learners.
  • Devices: Does it work on the hand-me-down tablet the family actually uses?
  • Errors: When learners miss, is it the skill or the interface? Your "no response" and "rapid guessing" data will help tell them apart.

06 — Requirements

What Does Your Learning Platform Need to Support?

Once one activity is designed this way, the platform requirements are much clearer. Here's the checklist we use before building any activity:

The Activity Blueprint

Answer these seven questions for one activity before development starts.

  • The learning objective: what the learner should be able to do, and with how much support
  • The learner's action: the decision or performance the activity requires, not just the input method
  • The evidence of success: what counts as correct, and on which examples
  • The feedback for each meaningful response: including partly right, no response, and guessing
  • The retry or next step: a new item, a hint, a worked example, or moving on
  • The information saved for progress: attempts, hints, support level, and new-example results
  • What the curriculum team must be able to change: items, feedback wording, hints, and order, without a developer

Those answers turn into a short list of platform capabilities: reusable activity templates, editable feedback and hint content, saved attempts and results, skills linked to activities, views for learners, parents, and teachers, and a way to preview and test an activity before it's published.

The last item on the blueprint is the one that decides whether a platform scales. My Music Workshop's platform was built so the founder's team can create and manage content themselves, and on Miss Humblebee's Academy we built a Lesson Builder that lets educators assemble lessons by skill, theme, or learner need. If every new pattern or feedback message requires a developer, the activity will stop growing the day the project ends. Our free Learning Platform Planning Worksheet includes space to map activities like this one.

07 — Extend or build

Do Interactive Learning Activities Require a Custom LMS?

Usually not a whole one. Start with the platform you already have and ask which blueprint answers it can't handle. A mature LMS often manages enrollment, delivery, and records well. The gaps tend to show up in the activity itself: classifying kinds of wrong answers, hint ladders, support-level tracking, age-appropriate interactions, or a parent view built around skills.

Those gaps can often be filled by extending the platform or building the activity as a separate experience that connects to it, using standards like LTI, SCORM, or xAPI where results need to flow back. A fully custom platform makes sense when the activities are the product and nothing available supports the learning loop you need. When Your Curriculum Becomes Software walks through that decision in detail.

08 — Common questions

Quick Answers

How do you design interactive learning activities?

Start with an observable objective, such as "extend a pattern without a hint." Choose an action that lets learners demonstrate it, write specific feedback for each kind of mistake, give learners a new attempt to use that feedback, and decide what evidence of progress to save, such as hinted versus independent success.

What makes a learning activity truly interactive?

The learner has to make a decision or perform a skill, and the activity responds to what they did. Clicking "next" or dragging an item is only an input method. An activity is interactive in a useful way when the learner's choice changes the feedback, the next item, and the record of their progress.

How is useful feedback different from a correct-or-incorrect message?

Useful feedback responds to the kind of mistake and tells the learner what to try next, such as "That shape is in the pattern. Look at what comes after each circle." A plain "incorrect" tells them only that they missed. Feedback works best when it focuses on the task, not the person, and is followed by a chance to try again.

What is the difference between lesson completion and learning progress?

Completion shows that a learner reached the end of a lesson. Progress shows what they can do: whether they succeeded with or without a hint, on new examples, and again at a later check. A platform should record these separately rather than treating a completed lesson as a learned skill.

Do interactive learning activities require a custom LMS?

Not usually. Many programs can keep their existing LMS for enrollment, delivery, and records, and extend it or connect a custom activity to it. A fully custom platform makes sense when the activities are the product and no existing platform supports the practice, feedback, and progress tracking the curriculum needs.

09 — Let's talk

Your Curriculum Is Ready. What's the Next Step?

The goal isn't more animations, quizzes, or games. It's software that supports the way your learners actually learn. You already have the lessons, the expertise, and the vision. Bring us one lesson. We'll map what learners do, what feedback they receive, and what progress your team needs to see, then identify what your existing platform can support and what may need to be built.

Schedule a free 30-minute Discovery Call with me and the team, or explore Learning Experience Development to see how we take an activity from blueprint to launch.

JL

Jamie Lords ↗ LinkedIn is CEO of Virgo Development and teaches Business and Marketing at Southern New Hampshire University, including helping design the marketing courses taught there. Virgo Development builds custom interactive learning platforms, including My Music Workshop and Miss Humblebee's Academy; see the My Music Workshop case study. Sources: Guo, Kim & Rubin (2014), ACM Learning @ Scale; Kluger & DeNisi (1996), Psychological Bulletin 119(2); Kornell & Bjork (2008), Psychological Science 19(6); Education Endowment Foundation, Teaching and Learning Toolkit: Feedback and the guidance report Teacher Feedback to Improve Pupil Learning; Pashler et al. (2007), Organizing Instruction and Study to Improve Student Learning, IES Practice Guide (NCER 2007-2004). The Activity Blueprint and Evidence Ladder are Virgo Development's working frameworks, not industry standards. The pattern activity, feedback messages, and attempt record are illustrative, not descriptions of any client's system. Research findings are design guidance, not proof that a particular platform improves learning. Miss Humblebee's Academy and My Music Workshop details come from Virgo Development's published engineer spotlights, case study, and each platform's public site. Compiled November 2026.

Share this article Share on LinkedIn