TL;DR: Give novices a complete example that exposes the reasoning, ask them to explain the decisive step, remove guidance gradually, and then require an independent attempt with feedback. Revisit the principle after a delay using a problem whose surface details have changed. Worked examples and retrieval practice serve different purposes: one makes a new procedure available for understanding; the other asks learners to reconstruct knowledge without the answer in view. The lesson below is an original teaching design informed by research, rather than a tested program or a promise of a particular learning gain.
The lesson's destination is a decision, not a formula on a slide
Imagine a short course for people who read product reports but have uneven mathematical preparation. Its learning objective is specific: given two subgroup rates and their denominators, learners should decide whether the overall rate can be calculated and, when it can, calculate it correctly. They should also explain why averaging the displayed percentages can be misleading.
This is a better destination than “understand weighted averages,” because the instructor can observe the action. A learner may recite the formula while still using the wrong denominator in a report. Another may calculate correctly using counts without remembering the phrase weighted average. The second learner has accomplished more of the intended task. The assessment should reward the decision and reasoning, not familiarity with the course's vocabulary.
The classroom, learners, and reporting situations described here are hypothetical. The numerical examples are deliberately small so the arithmetic can be checked without software. The central difficulty is selecting the right representation of the data. Adding a spreadsheet interface too early would introduce another procedure and make it harder to see whether the learner understood the principle.
The lesson has four performances: interpret a fully worked case, complete a partially worked case, solve a new case without guidance, and recognize when a later case lacks enough information. These performances are related, but success on one does not automatically establish success on the next. The sequence makes each transition visible to both instructor and learner.
What the research supports, and what this lesson adds
Sweller and Cooper's 1985 research on algebra examined learning through worked examples relative to conventional problem solving. It helped establish the value of presenting solutions during initial skill acquisition. The result is a reason to take example study seriously, not a rule that every subject or every learner should spend the same amount of time watching demonstrations. Their original paper contains the methods and boundaries of the experiments.
Renkl's theory of example-based learning discusses how learners can move beyond superficial inspection, including explanations of underlying principles and changes in guidance as knowledge develops. The practical implication used here is to ask about the reason for a step before asking for more repetitions of its arithmetic. Our particular reporting lesson and its progression criteria are instructional proposals.
Karpicke and Roediger's 2008 study used foreign-language vocabulary pairs and compared continued study with continued retrieval after an item had first been recalled. Continued retrieval benefited performance on the delayed test. That finding supports including retrieval after initial success; it does not directly measure the effects of this weighted-average lesson. The authors' paper makes the distinction between immediate acquisition and later recall especially clear.
Together, these sources motivate a sequence with two different kinds of difficulty. At first, remove unnecessary search so learners can inspect a meaningful solution. Later, remove the solution so they must reconstruct the principle. A course that keeps the answer visible throughout never reaches the second demand. A course that starts with unguided problems can fail to supply the knowledge the learner is supposed to retrieve.
Before instruction: find out which prerequisite is missing
The opening task takes only a few minutes. Ask learners what sixty percent of ten is, what eighty percent of thirty is, and what a denominator represents in a completion rate. Collect the answers before showing the lesson's central problem. These questions identify whether multiplication, percentages, or interpretation of a rate needs attention.
The point is not to sort the class into strong and weak people. A learner who calculates percentages accurately may still mistake the denominator. Another who understands the denominator may need an arithmetic refresher. Those learners require different help. A single total score would collapse the distinction and encourage the instructor to reteach material that some people already understand.
Provide a brief prerequisite repair when needed. Represent sixty percent of ten as six out of ten, then connect the count to the percentage. A small drawing of ten boxes can make the correspondence inspectable. Stop when the learner can move between the count and rate. This repair should not become a long detour into every way percentages appear in everyday life.
For learners already comfortable with the prerequisite, offer an explanation task: “Why does the number of people matter when we combine two rates?” This keeps them engaged in the same conceptual problem without requiring them to wait through identical arithmetic practice. It also gives the instructor an early signal of whether apparent expertise includes the relevant reasoning.
Demonstration one: combine counts before combining rates
Show this report: ten people entered onboarding route A, and sixty percent completed it. Thirty people entered route B, and eighty percent completed it. Ask for the completion rate across all forty people. Keep the report visible beside the worked solution so learners do not have to remember the numbers while following the explanation.
The demonstration proceeds through meaning before notation. In route A, six people completed onboarding. In route B, twenty-four people completed onboarding. Together, thirty of forty people completed it. The overall completion rate is seventy-five percent. Each calculation corresponds to a count of people, so the final numerator and denominator describe the same combined population.
Now contrast the tempting shortcut. Averaging sixty and eighty gives seventy percent. That calculation treats the two routes as equally important units. But the original question asks about people, and route B contains three times as many people as route A. The shortcut answers a different aggregation question, even though both calculations begin with the same two percentages.
The instructor's annotation should identify the decisive choice: “We return to counts because completion is counted across people.” Then write the equivalent weighted calculation in words: one quarter of the people have a sixty percent completion rate, and three quarters have an eighty percent completion rate. One quarter of sixty plus three quarters of eighty equals seventy-five.
Do not introduce several competing formulas on this page. The count representation explains the operation; the weighted representation connects it to the term used in later reports. Adding normalized weights, summation notation, spreadsheet functions, and a programming implementation at once would obscure the same decision under multiple surfaces. Those representations can be added when the learner needs them for a task.
End the demonstration with a reasonableness check. The combined rate should lie between sixty and eighty because both groups contribute people and use the same definition of completion. It should be closer to eighty because more people belong to route B. These checks cannot prove every answer correct, but they can reject some errors before the learner accepts an output.
The explanation prompt that reveals a superficial shortcut
Ask learners to answer: “What did we count, and why did we not average the displayed rates?” Do this before giving another set of numbers. A learner who says only “because the groups have different sizes” has identified a condition. Ask them to connect that condition to the people represented by each percentage. The answer should explain how different sizes change each group's contribution.
A learner may say that weighting is required because route B is more successful. That is a different misconception. Weight follows the denominator, not the attractiveness of the outcome. You can expose the error by reversing the rates while keeping group sizes fixed. The larger group still receives the larger weight, even when its completion rate is lower.
Let learners compare two explanations in pairs, then ask each person to write their own short version. Pair discussion supplies alternative language; individual writing reveals whether one person simply adopted the other's sentence. The instructor needs enough visibility to identify a mistaken principle before removing guidance. Length is not the criterion: a precise sentence can be better than an elaborate paragraph.
Demonstration two: remove the final step, then the choice of representation
The next case has twenty people in route A with a fifty percent completion rate and eighty people in route B with a seventy-five percent completion rate. Supply the converted counts: ten completions in A and sixty in B. Leave the combined numerator, denominator, and overall rate for the learner. The intended answer is seventy completions out of one hundred people, or seventy percent.
This partial example asks the learner to finish a meaningful operation while leaving the harder setup visible. If someone divides by two because there are two routes, the error now appears clearly at the aggregation step. Restore the explanation of what the denominator counts. Giving another complete demonstration of every multiplication would address a difficulty the learner may not have.
In the following case, remove the converted counts as well. Present twenty people with a forty percent completion rate and thirty people with a sixty percent completion rate. Ask learners to construct the counts, combine them, and check the answer. Eight plus eighteen gives twenty-six completions among fifty people, so the overall rate is fifty-two percent.
The progression should follow evidence rather than the slide number. If several learners still combine rates without denominators, give them a contrast case and an explanation opportunity before removing more support. If others can already justify the method, move them to a new problem. Guidance that helps one learner may be unnecessary for another, even within a small class.
One proposed progression criterion is that the learner completes two differently sized examples and explains the denominator without prompting. This is a local classroom rule, not a validated threshold from the cited papers. It is useful because it makes the instructor's decision explicit. A different course can select a different criterion and observe whether that criterion predicts later independence.
Remove an explanation before removing a number
Fading need not mean erasing the final line of every example. You can retain all calculations and ask learners to supply the missing reason for a step. “Why does this line divide by fifty?” tests conceptual reconstruction while reducing arithmetic demand. This is particularly useful when arithmetic mistakes are masking an otherwise sound understanding.
The opposite version retains the explanations but removes the calculations. That tests whether the learner can execute the described operations. Comparing the two performances helps separate procedural difficulty from conceptual difficulty. The distinction matters because a learner who cannot select a denominator needs different feedback from someone who multiplies twenty by forty percent incorrectly.
Independent retrieval: close the demonstration and attempt a new case
For the first independent attempt, remove access to the worked examples. The learner receives a fresh problem with no highlighted denominator and no intermediate steps. Fifteen participants in one group have an eighty percent completion rate; five participants in another have a forty percent completion rate. Ask for the combined rate and a brief explanation of the method.
The answer is fourteen completions among twenty participants, or seventy percent. The equal average of the displayed rates would be sixty percent. Ask learners to commit to an answer before receiving feedback. If the solution appears immediately beside the prompt, they can recognize it without reconstructing the path. Recognition may feel fluent, but it does not reveal the same performance.
Keep the attempt low stakes. An error provides information for the next teaching move; it should not punish the learner for entering the stage designed to reveal gaps. Invite the learner to mark the point where they became unsure. This makes feedback more specific and can distinguish an omitted calculation from a misconception that began at the first step.
After feedback, give another attempt with changed values. Simply rereading the correction leaves the corrected answer available in short-term context. A new problem asks whether the learner can apply the repaired principle. The instructor can then decide whether to continue independent practice or restore a partial example for the step that remains unstable.
Retrieval here includes both the method and its reason. A learner can memorize “multiply, add, divide” without knowing when that sequence applies. Require them to identify what the numerator and denominator represent. This explanation is especially important before transfer problems, where the same surface percentages may belong to different populations or different kinds of measurements.
Feedback clinic: treat different wrong answers differently
Consider four submissions to the independent task. One learner writes sixty percent. Another writes fourteen percent. A third writes seventy percent with no explanation. A fourth writes seventy percent and identifies fourteen completions among twenty participants. These responses should not receive identical feedback, even though a simple grading system might divide them into correct and incorrect.
The sixty percent answer suggests equal weighting. Ask the learner to represent the two groups as fifteen and five people, then show how many completions each supplies. The fourteen percent answer may indicate a lost denominator or a confusion between a count and a rate. Ask them what fourteen represents before supplying another formula. Diagnosis begins from their actual operation.
The unexplained seventy percent answer needs a probe, not immediate suspicion. Ask how the learner would handle a case with equal group sizes. Their response may reveal a sound method, a lucky calculation, or a memorized answer pattern. The explained seventy percent answer can move to a changed context. Correctness alone does not tell the instructor what kind of support remains useful.
Feedback should identify the decision to repair and give the learner something to do. “Remember weighted averages” names a topic. “Write each group's completed-person count, then choose a denominator that contains everyone in those counts” directs an operation. Good feedback is specific enough to act on while leaving the learner responsible for carrying out the correction.
Avoid turning the clinic into a catalogue of every possible mistake. Address errors that appear in the class or plausibly arise from the present representation. A long warning sheet can load the learner with irrelevant exceptions. Keep a record of observed errors for course revision, and use that record to decide which contrast belongs in the next iteration of the lesson.
Delayed transfer: change the surface and include an impossible problem
At a later session, give learners a report about average support tickets rather than completion percentages. Ten customers average two tickets each, and thirty customers average four tickets each. Ask for the average across all forty customers. Twenty plus one hundred twenty equals one hundred forty tickets; divided by forty, the combined average is three and a half tickets per customer.
The mathematical relationship is familiar, but the surface is different. Learners must recognize that each group average summarizes a count spread across customers. The relevant denominator is still people, although the numerator is now tickets. If they can only handle problems labeled completion rate, the earlier lesson has not yet supported this kind of transfer.
Then supply two completion rates with no group sizes. Ask whether the combined rate can be calculated. The correct response is that the information is insufficient unless an additional relationship, such as equal group sizes, is established. A learner who requests the missing denominators is exercising the intended decision skill. Do not penalize them for refusing to manufacture a number.
A second impossible case mixes denominators: one rate is completed sessions divided by started sessions, while another is completed people divided by enrolled people. Even if both percentages and both denominator counts are known, combining them directly would mix unlike quantities. Ask the learner to identify the mismatch and request a common definition. This tests interpretation rather than arithmetic alone.
Finally, introduce a table with an extra irrelevant column, such as the names of the routes or their launch dates. The learner must select what matters without being told that every number belongs in a formula. Keep the irrelevant information modest; transfer should challenge selection of the principle, not become a reading puzzle designed to ambush the class.
The learner's handout preserves the attempt, not just the answer
Divide a practice page into three areas: the first attempt, the feedback received, and a second attempt on a fresh case. Ask learners to keep the first response rather than erase it. The page then records a change in reasoning. A corrected final answer alone hides whether the learner initially selected the wrong denominator or merely made a multiplication error.
On the instructor's version, label each problem by its purpose: prerequisite check, representation choice, execution, or transfer. Keep those labels off the learner's initial prompt when they would reveal the operation. This allows the instructor to select a replacement task with the same instructional purpose without giving away the conceptual category.
The delayed page should also include a confidence estimate before feedback. Use it to start a discussion about how the learner judges readiness, rather than as another grade. If someone feels certain on every task but misses the missing-denominator case, the next teaching move can include checking what information an answer requires. That is a specific calibration problem, not a general lack of motivation.
Assess durability without confusing it with speed
For the delayed assessment, collect three kinds of evidence: the answer, the representation chosen, and the explanation of when the method applies. Record whether the learner had access to notes or calculators. The use of a calculator may be appropriate for the real task, but it should not conceal which part of the reasoning the assessment is intended to observe.
Time limits need a purpose. If the workplace task requires rapid interpretation during a meeting, a later timed exercise can be relevant. The first assessment of understanding need not reward the fastest person. Otherwise, an instructor may mistake arithmetic fluency for superior conceptual command and miss a slower learner who selects the correct denominator consistently.
Compare immediate and delayed performance at the level of the skill. A person may still calculate accurately but forget to check whether the denominators match. Another may retain the conceptual distinction but make arithmetic errors. Those patterns call for different revision. A course completion percentage cannot show them, and neither can a satisfaction survey asking whether the lesson felt clear.
For course improvement, inspect which transition produced trouble. Did learners understand the demonstration but fail when counts disappeared? Did independent success vanish after a delay? Did they retain the familiar calculation but fail the insufficient-information case? Each pattern points toward a different instructional change. Add support or practice at the specific transition, rather than simply extending every lesson.
Adapt the sequence to a technical course without copying its arithmetic
A programming course can use the same distinction between available guidance and reconstructed reasoning. A complete example might trace a small function through an input and output. A faded example could leave one branch unfinished. Independent practice could require a new function with the same underlying structure. Delayed transfer could alter the surrounding data model while preserving the conceptual demand.
The instructor should first identify which knowledge transfers. It might be choosing an invariant, distinguishing a value from its representation, or selecting a test that exposes a boundary. Reusing the same code with renamed variables may test recall of a template more than recognition of the principle. Changing everything at once can make a transfer failure impossible to interpret.
For an expert learner, a complete example may be unnecessary. A short diagnostic problem can establish whether they already possess the relevant structure. Move them to independent work when their explanation supports that choice. The sequence is a way to manage guidance according to knowledge; it is not a ritual that everyone must perform from the first slide.
The lesson's central design question remains concrete: what should the learner be able to reconstruct when the example is gone? Choose that performance, expose the reasoning needed for it, and observe the transitions carefully. A polished explanation earns its place when it helps the learner reach independent use, and independent use earns its place when it survives a changed case and a later attempt.