How Education Works | Instructional Design — How Goals, Evidence, Sequencing and Revision Become a Learnable Course

Read this as a design manual, not a promise of results. The worked courses, learner responses and review scenarios below are original illustrations, not reports of measured eduKate outcomes. Research and institutional guidance are linked where used. Suggested lesson sequences and review rules are proposals to adapt and test, not universally validated prescriptions.

Reading route: The design job · Goals and evidence · Building the sequence · Worked mathematics example · Worked argument example · Review and implementation · Sources and further reading.

1. Instructional design connects the destination to the learner’s actual work

A curriculum identifies worthwhile territory. Teaching is the interaction through which learners encounter and work with it. Assessment supplies evidence about what learners can do. Instructional design connects these decisions before, during and after delivery. Its central question is: what must happen between the learner’s starting point and the capability we intend to develop?

Carnegie Mellon University’s Eberly Center describes the importance of aligning learning objectives, assessment and instructional activities. The useful principle is that the task used to judge learning should correspond to what was intended and what learners had opportunities to practise. This does not require every lesson to look like an examination. It requires the intellectual demands to be connected. Source: Eberly Center, assessment alignment.

Consider a course advertised as teaching students to evaluate information. If students only underline facts and copy summaries, evaluation remains largely unpractised. Adding an evaluation question to the final test does not repair that omission. A designer needs to create occasions for learners to compare claims, inspect evidence, identify uncertainty and defend a judgement before their independent performance is assessed.

2. Start with a capability, not a content container

Requests often arrive as containers: make a worksheet, build an online module, prepare a workshop, cover Chapter Four. The container may be necessary, but it does not yet explain the educational job. Before building it, translate the request into something a learner should be able to do.

“Understand percentages” is broad. “Identify the reference quantity in an unfamiliar percentage-change problem and justify the calculation” is more informative. It names a decision and a context. “Improve writing” is similarly broad. “Use relevant evidence to support a claim and explain how that evidence supports it” gives the designer a clearer target.

Not every valuable goal can be reduced to one short performance statement. Curiosity, ethical judgement and appreciation may require varied evidence across time. The point is not to force every educational aim into a narrow checklist. It is to prevent a vague aspiration from concealing the absence of any plan for how learning will be recognised.

For each major goal, write two sentences: “The learner should become able to…” and “We would regard the following as insufficient evidence…” The second sentence is particularly useful. It stops recognition, copying, attendance or a polished group product from silently standing in for the capability that matters.

3. Design the evidence before designing the presentation

Suppose the goal is independent method selection. An exercise headed “Use Method B” cannot establish that learners can select Method B. It establishes something narrower: they can attempt the method after the selection has been supplied. That evidence may still be useful during instruction, but its meaning should remain explicit.

Design three evidence situations. First, a supported attempt that reveals whether the learner can follow the reasoning. Second, an independent attempt that removes the crucial prompt. Third, a changed situation that tests whether the learner recognises the same relationship without the familiar surface cues. These are proposed design categories, not a mandatory three-test formula.

For each situation, specify the allowed support. A formula sheet, calculator, dictionary, model answer, peer discussion or teacher hint may be appropriate. The question is whether it performs the very operation being assessed. A calculator may be harmless when the target is choosing a valid model; it changes the task when the target is fluent mental calculation.

Record what an incorrect response would tell you. A wrong final number may follow correct reasoning with an arithmetic slip. A correct answer may follow an invalid rule that happened to work. Evidence becomes richer when the task includes a short explanation, comparison or error analysis rather than relying only on the final product.

4. Inspect the starting point without turning it into an identity

A design needs a starting estimate, not a permanent label. Instead of describing a class as “weak,” identify the particular knowledge and procedures required by the next task. Which terms must be understood? Which notation must be readable? Which smaller decisions need to be sufficiently secure?

For a percentage-change lesson, the starting check might separate subtraction, fraction interpretation, conversion to percentages and the meaning of an original quantity. For an argument lesson, it might separate understanding the source, identifying its claim and connecting evidence to a conclusion. Each check should be short enough to inform teaching without consuming the entire lesson.

A prerequisite is not simply something that appeared earlier in a textbook. It is something the learner needs for the present reasoning to work. Ask what would break if that knowledge were absent. This question often reveals that some supposed prerequisites are optional background while a small, easily overlooked distinction is essential.

Use the initial evidence to offer routes. One learner may need a brief repair example. Another may be ready for a contrast problem. A third may already understand the target and need a more demanding application. The design should allow movement between routes when new evidence appears.

5. Analyse the task at the level of decisions

A task analysis describes the operations hidden inside a competent performance. “Write an explanation” contains several: identify the question, select relevant information, decide what relationship matters, express it clearly and check that the answer addresses the question. A learner can fail at any of these points.

Do not analyse only the visible actions. Opening a document and typing a paragraph are actions; selecting the right evidence is a decision. Good design makes the important decisions available for teaching. An expert example should therefore show why a piece of information was used, why another was rejected, and what would change the conclusion.

A useful planning note has four fields: decision required, information needed, likely error, and evidence that distinguishes understanding from imitation. For example, “Choose the denominator” becomes more teachable when paired with “the quantity used as the reference,” “choosing the larger number automatically,” and “explaining the choice in a reversed comparison.”

Task analysis should stop before it becomes an unmanageable catalogue. Describe the decisions that affect the learning goal. The purpose is to improve a lesson, not produce a document so intricate that no teacher can use it.

6. Build a sequence in which each experience has a job

The Institute of Education Sciences’ 2007 practice guide recommends, among other approaches, distributing learning over time, alternating worked solutions with problem solving, connecting concrete and abstract representations, and using quizzes to revisit important content. Its recommendations carry different evidence ratings; they should not be treated as equally certain universal rules. Source: IES, Organizing Instruction and Study to Improve Student Learning.

For a particular course, translate principles into named jobs. An opening question can expose a distinction. A worked example can reveal a decision. A partially completed task can give the learner control of that decision. An independent task can test whether it is now internally available. A later task can check access after the first lesson is no longer fresh.

This is not an argument that every lesson must follow one sequence. A knowledgeable group may begin with a difficult problem before examining alternative solutions. A novice group may need more explicit preparation. What matters is that the designer can explain why the next activity belongs at that point for those learners.

Remove activities whose only justification is variety. Keep variety that creates a useful contrast, new representation or different demand. A lesson may look less spectacular after this edit while becoming much clearer about what the learner is expected to notice and do.

7. Separate explanation, rehearsal and independent performance

During explanation, the teacher may legitimately carry much of the reasoning. During rehearsal, responsibility is shared. During independent performance, the learner should carry the part of the task that the design intends to assess. Confusing these conditions creates misleading success.

Annotate a proposed lesson with a simple question: who makes the next important decision? If every significant choice is made by the teacher, an apparently active lesson may still provide little independent practice. Students can fill many blanks while the structure of the answer remains externally supplied.

Support removal should be specific. Remove the chapter label, not the readable font. Remove the worked calculation, not the learner’s necessary assistive technology. Remove the teacher’s suggested conclusion, not the source material that any competent person would consult. Independence means ownership of the target reasoning, not artificial deprivation of ordinary tools.

When support is removed, expect useful information rather than guaranteed success. Difficulty may reveal a design gap. Restore only the support needed to address that gap, then create another opportunity to test the learner’s own decision.

8. Make feedback an event in the design

The Education Endowment Foundation’s feedback guidance warns that feedback is not automatically beneficial and that producing it consumes time. It emphasises the principles of effective feedback rather than assuming written or spoken delivery is inherently superior. Source: EEF, Teacher Feedback to Improve Pupil Learning.

In a course plan, “give feedback” is incomplete. Add the next learner action. Will the learner revise a claim, redo a calculation, compare two explanations, identify the incorrect assumption, or try a new case? Without that action, feedback can remain a comment attached to completed work rather than part of another attempt.

For example, “Your evidence is weak” identifies a problem but gives little direction. A more useful design might require the learner to choose between two candidate pieces of evidence and explain which bears more directly on the claim. The teacher then checks the decision, not merely whether the paragraph became longer.

Budget time for this. If a lesson includes forty minutes of production and five minutes of correction after learners have mentally finished, the design has placed repair at the margin. Move a short feedback-and-retry cycle earlier, where the response can influence subsequent work.

9. Accessibility belongs in the first draft

CAST’s Universal Design for Learning Guidelines offer design considerations across engagement, representation, and action and expression. They include clarifying language, supporting access to information and enabling appropriate ways to respond. Use these as questions for examining barriers, not as a guarantee that every possible format improves every task. Source: CAST UDL Guidelines.

Ask whether a difficulty is central to the objective or accidental. A faint diagram is not a useful challenge in mathematical reasoning. An unexplained interface is not a meaningful test of historical interpretation. Conversely, removing all reading from an assessment of independent reading changes the target rather than merely improving access.

Design alternatives with a clear reason. A printed equivalent may preserve access when connectivity is unreliable. A verbal explanation may reveal reasoning when extended handwriting is not the target. A written alternative may help a learner demonstrate an idea without being required to perform publicly. Different routes should still make the intended capability visible.

Check instructions with someone who was not involved in writing them. Ask that person to state what the task requires. If their interpretation differs, the material needs revision before learner performance is used to make claims about understanding.

10. Worked design: a unit on percentage change

Illustrative commission: learners can calculate simple percentages but sometimes choose the wrong reference quantity when explaining change. The proposed unit is not a complete mathematics syllabus. Its specific purpose is to help learners identify and justify the reference quantity in unfamiliar comparisons.

Begin with a pair of situations. A value increases from 80 to 100. In a separate situation, it decreases from 100 to 80. The numerical difference is 20 in both, but the percentage changes differ: 20 ÷ 80 × 100 = 25%, whereas 20 ÷ 100 × 100 = 20%. The useful teaching contrast is not simply that one operation is increase and the other decrease. It is that the comparison is made against a different starting quantity.

The opening response should reveal the learner’s rule. Ask: “What does the denominator represent in each calculation?” A learner who answers “the larger number” has supplied a testable misconception. A learner who says “the original quantity” but selects the wrong value may have a language or reading problem. A learner who selects correctly but calculates incorrectly needs a different repair.

For explanation, use the two quantities as named states rather than presenting a formula alone. State the original value, the change and the comparison. The teacher models the sentence: “The change is 20, measured against an original value of 80.” The formula is then a compact expression of that sentence. This makes the mathematical operation answerable to a relationship rather than a keyword.

For guided practice, supply the original and final values but leave the reference statement incomplete. The learner must finish the statement before calculating. Then reverse the situation. This preserves the arithmetic while changing the reference, creating a controlled opportunity to test whether the learner is following the relationship.

For independent practice, remove the incomplete sentence and mix the problems with a simple “percentage of a quantity” item. Now the learner must decide which relationship is being asked about. A correct formula selected only because every question on the page uses that formula is weaker evidence than choosing it within a mixed set.

For a later check, change the setting and representation. Use a small table describing a count before and after a change. Ask for a justified conclusion, not only a number. This does not magically establish broad transfer, but it supplies a more demanding observation than repeating the original example with different digits.

11. How to review the percentage unit without guessing why it worked

Suppose the first independent responses are mixed. Resist rewriting the whole unit immediately. Inspect where the reasoning diverged. A learner might subtract in the wrong direction, confuse original and final quantities, choose the correct denominator with an invalid explanation, or understand the relationship but omit the percentage conversion.

Each observation suggests a different next test. To examine the reference quantity, ask the learner to explain the comparison without doing arithmetic. To examine calculation, supply the relationship and ask for the operation. To examine reading, present the same information in simpler language while keeping the mathematical demand. These contrast tasks are design proposals for clarifying the difficulty, not clinical diagnoses.

The teacher’s review note might read: “The reversed example was useful because it exposed the larger-number rule. The independent set still gave away the topic through its title. Next version: remove that title and include one non-change problem.” That note identifies a specific design repair rather than attributing the entire result to motivation or teaching style.

Do not declare mastery from one corrected response. Keep the claim proportionate: the learner selected the reference correctly on this task under these conditions. A later independent task can strengthen or weaken that claim. The distinction between evidence and conclusion keeps the design honest.

12. Worked design: an evidence-based recommendation

Illustrative commission: students write opinions confidently but often insert evidence without explaining its relevance. The proposed task asks them to recommend whether a school library should extend opening time. All information in this example is invented for teaching and must not be presented as evidence about a real school.

Provide a short information pack with deliberately different kinds of evidence: a timetable showing current opening hours, a student survey with an uneven response, a staffing constraint and an anecdote from one student. The learning target is not “support longer opening hours.” It is to make a defensible recommendation that distinguishes evidence, limitations and practical conditions.

The first attempt is a brief recommendation written before a model is shown. Read it for decisions. Did the learner choose evidence relevant to access? Did they treat one anecdote as representative of everyone? Did they notice the staffing constraint? Did they acknowledge that survey respondents might differ from non-respondents? These observations tell the teacher what the next explanation needs to address.

The model should not be a perfect essay to memorise. Show two short recommendations. One uses several impressive facts but never explains their connection. The other uses less information, distinguishes what is known from what remains uncertain, and proposes a limited trial rather than an unlimited commitment. Ask students to identify the stronger reasoning and explain why.

For guided practice, offer a claim and ask learners to choose the most relevant evidence from the pack. Then require the missing connecting sentence: “This matters because…” The phrase is a temporary prompt, not a compulsory writing template. Its job is to make the inferential step visible. Remove it once students can perform that step without the cue.

For independent performance, change the question. Ask whether the school should create a book-reservation service instead. Some earlier information remains relevant, but the recommendation cannot simply be copied. Learners must reconsider the match between problem, evidence and proposal. The design now tests judgement rather than memory of the model’s conclusion.

For feedback, select one issue with the greatest leverage. A learner who ignores uncertainty may revise only the limitation and conclusion. A learner who chooses irrelevant evidence may sort information before rewriting. The next task should test the repaired decision in a new paragraph rather than rewarding the reproduction of the teacher’s correction.

13. Design for transfer without pretending every skill travels everywhere

Transfer is a design ambition that needs its own evidence. Solving a percentage-change problem in a new setting is a narrower claim than applying sound judgement across all domains. A course should state the distance of transfer it is attempting instead of promising a general improvement in thinking from one activity.

Specify what changes and what remains. In the percentage unit, numbers, wording and representation can change while the relationship stays stable. In the recommendation unit, the decision context can change while the need to connect evidence to a claim remains. Ask the learner to identify that stable structure explicitly.

Vary one important feature before varying many. When every feature changes at once, failure becomes difficult to interpret. A designer cannot tell whether the learner missed the principle, lacked the new background knowledge or could not understand the new representation. Controlled variation is useful because it makes the evidence less ambiguous.

Also define when the method should not be used. A learner who knows only positive examples may apply the procedure too widely. An appropriate non-example can reveal whether the learner understands the boundary rather than only the sequence of steps.

14. Make the learning journey usable outside the teacher’s head

A lesson plan that relies on one teacher remembering every explanation is difficult to maintain. Write down the educational essentials: the target, the critical distinction, the evidence tasks, likely errors and the reason for the sequence. Another teacher should be able to preserve the logic without copying the original teacher’s manner.

Separate fixed elements from flexible elements. In the percentage example, the comparison between the two reference quantities is central. The names used in the problem are not. In the library recommendation example, the need to evaluate relevance and uncertainty is central. The local setting can be adapted to something more meaningful for the class.

This distinction also helps a substitute teacher. Give a short “do not skip” note rather than a script of every spoken sentence. Include what evidence to collect before moving on. The resulting materials are not just reproducible documents; they carry enough reasoning to support informed adaptation.

15. Use technology only after the educational job is clear

When reviewing a proposed digital course, ask what would remain if the platform were replaced by paper and conversation. This thought experiment separates the educational mechanism from its delivery. A platform may still be valuable for accessibility, distribution or feedback, but those benefits should be named rather than assumed.

Design the fallback at the same time. If a learner cannot access the interactive question, can they still encounter the same relationship and demonstrate the same decision? If an automated explanation is wrong, who checks it? If a generated hint gives away the answer, does the subsequent score still count as independent evidence?

For AI-assisted preparation, a sensible boundary is to treat generated examples and answer keys as drafts requiring subject review. Do not make a tool’s fluent wording the evidence for its correctness. The designer remains responsible for checking the actual task, the solution and the conditions under which the solution holds.

For learner use, distinguish supported practice from independent demonstration. Both can belong in a course. Problems arise when the record does not reveal which occurred and a supported result is later interpreted as unaided capability.

16. Test the first version before treating it as a finished product

A practical first-version review can be small. Ask a colleague to work through the material without your explanation. Check whether the instructions make sense, whether the answer key is correct, whether the sequence introduces something before relying on it, and whether the final task actually tests the stated objective.

During an initial classroom use, observe points of friction. Where do learners ask procedural questions? Where do they copy without deciding? Which example is repeatedly misunderstood? Which resource is inaccessible? Collect enough evidence to improve the design, without turning every learner interaction into a surveillance record.

A trial lesson is not automatically an experiment establishing effectiveness. Teacher explanation, student familiarity and the unusual attention paid to the first use can all complicate interpretation. Report what was observed and what will change. Reserve stronger causal claims for evidence capable of supporting them.

Before wider use, distinguish errors that block release from refinements that can wait. An incorrect answer key or missing essential instruction is a blocker. A slightly awkward heading may be a refinement. A course should not be released merely because most of its checklist is complete when one remaining problem undermines the learning claim.

17. Evaluate learning, implementation and burden separately

Three questions should remain separate in a review. Did the intended teaching take place? What could learners do afterwards? What did the design cost in time, effort and resources? A course can be faithfully delivered without producing the intended learning. It can also produce promising results under conditions too demanding to sustain.

Keep a small evidence set for each question. Implementation evidence might include whether the independent attempt and feedback cycle occurred. Learning evidence might include the explanation on a changed task. Burden evidence might include teacher preparation time and the accessibility problems encountered. Do not allow a convenient completion statistic to answer all three questions.

Look for displaced work. An intervention that adds useful practice may still take time from another necessary topic. A complex marking system may generate more comments while reducing the time available for teaching. The design review should account for these trade-offs instead of assuming every addition is free.

18. A practical commissioning brief

Before producing the next course, write a brief that a teacher, designer and reviewer can all use. Begin with the audience and the problem: “This is for learners who can do X but cannot yet reliably do Y.” Then define the intended capability, the contexts in which it should work, and the evidence that will count.

Continue with prerequisites, available time, ordinary tools, access requirements and known constraints. State what is outside scope. A unit on choosing evidence for a recommendation should not quietly become a complete course in research methods. A unit on reference quantities should not be judged by its coverage of every percentage topic.

Finally, specify the review. Who checks subject accuracy? Who tries the instructions without prior explanation? What must learners attempt independently? What information will be collected after first use? Who owns the next revision? These decisions keep responsibility attached to the work rather than dispersed across a vague commitment to quality.

A reusable brief: audience and starting evidence; target capability; unacceptable substitutes for evidence; prerequisites; learning sequence; support and its removal; feedback followed by action; accessibility; independent and changed-context tasks; review owner; revision trigger. Keep it short enough to consult while making decisions.

19. What students and parents can ask about a learning programme

Students do not need to become instructional designers to read a course more intelligently. Ask what the current activity is preparing you to do without help. After studying an example, explain one decision that the example made. After feedback, identify the action that will make the next attempt different. Before claiming a topic is secure, attempt something that does not advertise the method.

Parents can ask similar questions without taking over instruction. What is the specific learning goal? Which part can the child now do independently? What does the teacher want the child to practise at home, and what should be left for classroom teaching? What would useful progress look like besides completing more pages?

The purpose is not to demand constant proof from a child. It is to distinguish productive support from activity that only looks educational. A family should not need a complicated dashboard to understand the central learning job.

20. Common design failures and the smallest useful repair

The course starts with a slide deck. Repair the commission first: define the decision learners should become able to make. The final task is harder in a new way. Identify the missing practice rather than blaming learners for not inferring it. The worksheet gives away every method. Add a mixed task in which method selection belongs to the learner.

Feedback appears only after the work is finished. Build a shorter feedback-and-retry cycle earlier. The strongest learners make the group product. Require an individual explanation of a critical decision. The resource is attractive but inaccessible. Remove the unnecessary barrier while preserving the intellectual target.

A pilot produces encouraging responses. Keep the claim limited to what was observed and test another context. Teachers cannot sustain preparation. Preserve the critical teaching decisions while simplifying production. The designer keeps adding content. Return to the target and identify what genuinely supports it.

These repairs are deliberately specific. “Make the course more engaging” may be desirable, but it does not tell a designer what to change tomorrow. “Remove the method label from the independent task and collect a one-sentence justification” does.

21. The final test: does the design leave the learner carrying more?

Imagine the course has ended. The slide deck is closed. The teacher is not providing the next cue. A new task appears. What part of the reasoning can the learner now initiate, explain, check and repair? That is the question around which this design manual is built.

Instructional design cannot guarantee that every learner will succeed on every occasion. It can make the intended learning more explicit, ensure that learners have relevant opportunities to practise, and create better evidence when the journey breaks. It can also make teaching more transferable between people by preserving the purpose behind the sequence.

The best course is not necessarily the one with the most content or the most elaborate interface. It is the one whose experiences, supports and evidence make sense together—and whose designers know what to inspect when they do not.

Sources, evidence boundaries and further reading

The linked sources provide the externally grounded principles used here; the worked units, commissioning brief and repair scenarios are eduKate’s original explanatory applications. They are not findings from the cited studies. Source pages were checked on 5 September 2026.

• Carnegie Mellon University, Eberly Center: alignment of objectives, assessment and instruction.
• Institute of Education Sciences: Organizing Instruction and Study to Improve Student Learning, 2007. A practice guide with recommendation-specific evidence ratings, not a single universal lesson script.
• Education Endowment Foundation: Teacher Feedback to Improve Pupil Learning, 2021.
• CAST: Universal Design for Learning Guidelines, version 3.0. A design framework; individual adaptations still require a clear instructional purpose.