The Tutor Handbook · Volume 0163 · Series ID THB-0163
Series route: The Tutor Handbook — Complete Series Index.
A tutor spends twelve minutes writing careful comments on a learner’s essay. The comments are accurate. They identify the weak evidence, the vague topic sentence, the missing link between claim and example, and a repeated sentence-level error. The learner reads the page, nods, asks one sensible question, and then puts the essay into a folder.
Nothing has gone wrong with the feedback itself. The problem is that the lesson has treated feedback as information rather than as a change in behaviour.
The Feedback-Action Loop is the tutor’s discipline of designing a near, purposeful opportunity for the learner to use feedback while the target decision is still visible, then checking whether the revised attempt shows a genuine change that can travel beyond the corrected item.
This volume owns the moment after feedback has been given. It asks what the learner must do next so that advice becomes an observable learning event rather than a well-written annotation.
Quick Answer
Good feedback does not finish when the learner understands the comment. It finishes when the learner attempts the target again, under conditions that let the tutor see whether the feedback changed what the learner can do.
That does not mean every error needs a redraft, every marked paper needs to be rewritten, or every comment needs an immediate duplicate question. The tutor should select a small number of high-value feedback targets, give the learner a task that requires the same underlying decision, reduce unnecessary support, and inspect the new response.
A useful feedback-action loop therefore has five parts: identify the target, explain the gap, require a learner response, provide a fresh or revised opportunity to act, and inspect what changed. If the learner can repeat the corrected sentence but cannot apply the principle to a new paragraph, the feedback has repaired an artefact more than a capability. If the learner can use the idea in a fresh context without the tutor restating the instruction, the evidence is stronger.
1. What This Volume Owns
The Tutor Handbook already contains volumes on feedback timing, correction load, criteria disclosure, follow-up questioning, learner voice and evidence freshness. Those volumes answer important neighbouring questions: when should feedback arrive, how much correction is usable, how visible should success criteria be, and when is later evidence independent enough to trust?
The present volume owns a different reader job: what should happen after useful feedback has already been given?
This is easy to overlook because tutors often work in an environment where feedback itself is visible work. Marking can be extensive. Explanations can be thoughtful. Parents can see comments. Learners can repeat the advice back. Yet none of those receipts proves that the learner has incorporated the feedback into later performance.
The tutor therefore needs a practical bridge between “I told them” and “they can now do it”.
2. Why Feedback Dies on the Page
Feedback often fails for ordinary logistical reasons rather than because the comment is poor. The lesson may end immediately after marking. The learner may leave with several comments but no protected time to use them. The next week may begin with a different school topic. Homework urgency may displace the planned return. The learner may assume that reading the correction counts as completing the task. A tutor may feel that revisiting the same work is inefficient when there is so much syllabus left.
There is also a subtler problem. Once the tutor has explained the correction, the learner’s next response can look successful because the route is still in working memory. If the tutor says, “Your evidence is relevant but indirect; choose the line that proves the claim most directly,” and then immediately points to the same paragraph, a correct revision may show comprehension of the tutor’s instruction more than independent evidence selection.
So feedback needs action, but the action must be designed carefully enough to tell the tutor something useful.
3. Current Evidence Puts Action Inside Feedback
The Education Endowment Foundation’s 2026 resource on feedback for 16–19 learners makes the point explicitly: feedback can support independence when learners understand next steps and have opportunities and support to act on them. The guidance recommends purposeful tasks such as redrafting, correcting errors or completing similar tasks, and warns against treating feedback as an end point.
Source: Education Endowment Foundation — Feedback, Developing Independent Learners, published 24 August 2026.
The Australian Education Research Organisation’s monitoring guidance likewise treats feedback as part of a responsive instructional cycle. Its classroom examples include learners correcting misconceptions and then attempting similar questions again so teachers can see whether the learning has been put back into context.
Source: AERO — Monitor progress: Primary and secondary, updated 23 June 2026.
These are broad teaching resources rather than a validated private-tuition protocol. The useful transfer is procedural: if feedback is meant to improve performance, the learner needs a meaningful chance to perform again.
4. The Difference Between Reading Feedback and Using It
A learner can understand a comment without being able to implement it. “Use a more precise verb” may be understood linguistically, yet the learner may still choose vague verbs when writing independently. “State the original quantity before calculating percentage change” may make sense when explained, yet disappear when a new word problem uses different nouns. “Explain the mechanism, not only the observation” may be memorable in Science, yet fail under a changed context where the causal chain is less familiar.
Feedback use therefore has at least three layers. First, the learner can recognise what the tutor means. Second, the learner can repair the original work with the comment available. Third, the learner can apply the same principle later when the comment is no longer present.
A tutor who stops at the first layer has checked comprehension of feedback. A tutor who stops at the second has checked supported revision. The strongest everyday evidence arrives when the target operation returns under a fresh or delayed condition.
5. Select the Feedback Target Before Asking for Revision
A page covered in corrections creates a false problem: the learner appears to need to act on everything. That is rarely useful.
Suppose an English paragraph has weak topic control, three punctuation mistakes, one awkward phrase, one unsupported inference and two spelling errors. If the tutor asks the learner to “fix the paragraph”, the revision bundles several decisions. A better paragraph may not reveal which feedback was understood. The learner may simply copy the tutor’s wording or correct the easiest surface errors.
Instead, choose the smallest feedback target that matters most to the learning route. For example: “Today we are checking whether you can make the evidence prove the inference.” The learner revises only the evidence–inference link. Sentence-level corrections can be handled separately or later.
This follows the same logic as diagnostic isolation elsewhere in the Handbook. When too many things change at once, the tutor loses interpretability.
6. Feedback Should Point to a Decision, Not Merely a Defect
“Too vague” identifies a problem. “Choose the noun or verb that carries the exact meaning you want the reader to infer” identifies a decision.
“Wrong method” identifies a problem. “Before calculating, identify which quantity is the original base and which is the changed amount” identifies a decision.
“Need more explanation” identifies a problem. “After stating what changed, name the process that connects the changed condition to the observed result” identifies a decision.
Actionable feedback makes the next cognitive or procedural move visible without supplying the whole answer. The learner should know what kind of work to perform, not merely that the previous answer was insufficient. This matters because the next task must be designed around the decision. A comment that names only the defect gives the tutor little basis for constructing a diagnostic return.
7. The Original-Work Repair
The simplest action opportunity is to repair the original work. This is useful when the learner needs to see the relationship between feedback and the artefact that generated it. Editing an essay paragraph, correcting an equation, rewriting a Science explanation or repairing a graph interpretation can make the feedback concrete.
But the tutor should be precise about what this proves. A successful original-work repair shows that the learner can use feedback with the original context still available. It does not automatically prove transfer. The page itself supplies cues: the familiar passage, the remembered mistake, the tutor’s markings, perhaps the model answer.
Treat the repair as the first receipt, not the final one.
8. The Parallel Return
A stronger next step is a fresh but closely matched task. The learner who corrected percentage base selection receives another percentage problem with different values and context. The learner who revised evidence selection receives a second short inference question. The learner who repaired a causal Science explanation receives a new scenario using the same mechanism.
The parallel return should preserve the target while changing enough surface detail to prevent simple copying. This is not meant to be harder. It is meant to be informative.
If the learner succeeds, the tutor has evidence that the correction is beginning to generalise. If the learner repeats the same error, the feedback may have been understood as a local instruction rather than integrated into the learner’s decision process.
9. The Delayed Return
Immediate success can be inflated by recency. The learner still remembers the exact words the tutor used. The corrected example is vivid. The answer route may feel obvious because it was just discussed.
A delayed return introduces a modest gap before checking the same target again. The correct delay depends on the learning job. It might be later in the lesson, the next session, or the next naturally occurring school task.
The point is not to make the learner forget. The point is to reduce dependence on the immediate feedback episode. When the target reappears after delay, the tutor learns whether the correction has become more available without the original comment.
10. The Changed-Condition Return
For some feedback, a changed condition matters more than a delay. A learner who can apply a writing comment only when the tutor highlights the relevant sentence is not yet fully independent. A Mathematics learner who selects the correct method only on a topical worksheet has not yet shown selection under mixed conditions. A Science learner who explains causality only when the question uses the same vocabulary as the model answer may still be bound to surface form.
The tutor therefore changes one condition that matters to the claim: remove a cue, change the representation, mix the task among neighbours, vary the context, or let the learner plan the response without the template.
The change should be targeted. If everything changes at once, failure becomes uninterpretable.
11. Constructed Case: Alicia and the Percentage Comment
This is a fictional teaching case. Alicia repeatedly calculates percentage change using the final value as the base. Her tutor writes, “For percentage change, identify the original quantity before calculating. The denominator must represent the starting amount.” Alicia reads the comment and says she understands.
The tutor does not move on. First, Alicia repairs the original question and labels “original” and “new”. Then she receives a fresh problem about a membership fee rather than a product price. She again identifies the original amount correctly.
At the next session, a mixed set includes one percentage change item among ratio and proportion questions. No labels are provided. Alicia chooses the correct base.
The three receipts mean different things. The repair shows supported correction. The fresh item shows near transfer. The delayed mixed item gives stronger evidence that method selection changed. The tutor can now reduce active monitoring without claiming that Alicia will never make the error again.
12. Constructed Case: Beatrice and the Inference Redraft
Beatrice writes an inference answer that is plausible but supported by a line that only loosely relates to the claim. Her tutor could simply point to the best quotation. Instead, the tutor asks Beatrice to compare two candidate lines and explain which one would convince a sceptical reader.
Beatrice chooses the stronger line and revises the answer. That is useful, but it is still a supported discrimination task. The tutor later gives a new short passage and asks Beatrice to write one inference with the most direct evidence she can find. There is no pair of choices this time.
Beatrice selects a relevant but indirect line again. The feedback was not useless. It revealed that Beatrice can recognise stronger evidence when alternatives are visible, but independent evidence search remains fragile. The next teaching move should target search and selection, not repeat the wording “choose direct evidence” more loudly.
13. Constructed Case: Ciara and the Science Explanation
Ciara’s Science answer states what happened but not why. The tutor gives feedback: “You have described the observation. Add the process that connects the changed condition to the result.” Ciara rewrites the answer by inserting the missing mechanism.
A weaker feedback cycle might stop there. Instead, the tutor presents a new scenario involving a different surface context but the same causal principle. Ciara explains the mechanism without being reminded to “say why”.
That fresh response is valuable because the feedback has altered the structure of her answer, not merely the wording of the original one. The tutor can now ask Ciara to self-check future explanations with a simple internal question: “Did I only say what happened, or did I explain the link?” The goal is learner-owned quality control.
14. Constructed Case: Denise and the Overloaded Marked Paper
Denise receives a Mathematics paper with fourteen annotated errors. Some are arithmetic slips. Some are notation problems. Three involve the same deeper issue: she changes methods midway through multi-step questions without checking whether the new method preserves the same quantity.
If the tutor asks Denise to correct all fourteen items, the lesson becomes a repair marathon. The learner may finish exhausted without gaining a stable rule.
The tutor clusters the errors. The high-value target is method continuity. Denise repairs one representative question and explains the decision point where she switched. She then attempts a new question designed to create the same temptation. Only after that target is stable does the tutor return to smaller presentation issues.
The feedback-action loop protects priority. A marked page is not a queue where every red mark has equal instructional weight.
15. Constructed Case: Emily and the Study-Planning Comment
Feedback is not limited to academic answers. Emily repeatedly creates revision plans that contain too many large tasks for one evening. Her tutor says, “Your plan lists goals, but it does not estimate the size of each task. Break large tasks into units you can finish in the available time.” Emily nods and rewrites the current plan.
The next week, however, she again writes “revise Chemistry” as a single block. The tutor learns that the feedback repaired a plan, not the planning operation.
So the tutor changes the action opportunity. Emily receives a realistic 90-minute window and four competing tasks. She must decide what can fit, what needs to be divided, and what should be postponed. The tutor observes the decision process rather than the neatness of the finished plan.
Action reveals whether feedback has entered the learner’s planning behaviour.
16. The Feedback-Action Card
- Target: What is the one learner decision this feedback is meant to change?
- Comprehension: Can the learner explain the feedback in their own words without merely repeating mine?
- Original repair: Will the learner revise or correct the work that generated the feedback?
- Parallel return: What fresh task can check the same target?
- Delay: Does this capability need a later return before I trust it?
- Support: Which legitimate supports remain, and which answer-giving cues should disappear?
- Receipt: What evidence would show that the feedback changed a transferable decision rather than only one artefact?
- Stop rule: When will I stop monitoring this target actively?
17. Do Not Turn Every Comment Into Another Test
The action loop can be overused. If every small correction creates a fresh verification task, tuition becomes exhausting. Learners may feel that no feedback is ever finished. Tutor workload also expands unnecessarily.
Use action opportunities in proportion to consequence. A spelling slip in ordinary practice may need only correction. A recurring misconception may need a fresh check. A prerequisite error that will affect the next unit may deserve delayed verification. A major independence claim may need changed-condition evidence.
The Error-Cost Asymmetry Gate in this series remains useful: higher-cost decisions deserve stronger receipts.
18. Do Not Confuse Compliance With Uptake
A learner can obediently rewrite work without changing understanding. They may copy the tutor’s phrase. They may follow tracked changes. They may replace the answer with a model response. The page improves, but the learner’s decision process remains hidden.
This is why a tutor should occasionally ask for explanation or use a fresh task. The goal is not to interrogate every revision. It is to distinguish artefact repair from capability change when that distinction matters.
A beautifully corrected page is evidence about the page. It becomes evidence about the learner only through the conditions under which the learner produced the correction.
19. Feedback Tone Matters, but Tone Is Not the Whole Mechanism
Learners are more likely to engage with feedback when it is clear, focused and psychologically usable. The EEF notes that motivation, confidence and trust can influence how learners respond. A torrent of criticism can reduce uptake even when every comment is technically correct.
But warmth alone cannot substitute for action structure. “Great effort—just keep working on evidence” is pleasant but vague. “Your inference is plausible. On the next question, choose the line that most directly proves it, then explain why that line is stronger than your first choice” gives the learner a usable next move.
The best feedback-action loop combines humane delivery with an explicit performance opportunity.
20. Group Feedback Needs Individual Receipts
In three-learner tuition, one learner’s mistake can trigger a shared explanation. That can be efficient. But after group feedback, the tutor still needs to know whether each learner can use the correction.
If Alicia makes the error and the tutor explains it to everyone, Beatrice and Ciara may understand while listening. Or they may simply follow Alicia’s corrected answer.
A brief private first attempt on a similar task can protect individual evidence. The group can then discuss differences. This does not require three separate lessons. It requires distinguishing shared instruction from individual receipt.
21. Feedback and Access Supports
A feedback-action loop should not remove legitimate access supports merely to make the check feel “independent”. If a learner routinely uses approved text-to-speech because decoding print is not the target, keep that access condition stable. If a formula sheet is legitimately part of the learning or assessment environment, the tutor need not remove it when checking a different capability.
Remove or fade only assistance that performs the target operation. The Accommodation Baseline and Access-Support Boundary volumes remain the owners of those distinctions. The present article simply applies them to feedback: the action task should test the thing the feedback was meant to change, not unrelated access barriers.
22. Feedback About Process Versus Feedback About Product
Product feedback says something about the current artefact: this paragraph lacks evidence, this equation has a sign error, this graph label is missing. Process feedback points to a repeatable decision: select direct evidence, check sign when moving terms, label axes before interpreting the trend.
The action loop is most powerful when it can move from product to process. That does not mean feedback should always be abstract. Beginners may need concrete correction first. The tutor can repair the product, then name the repeatable decision and check it on another example.
The learner eventually needs a rule that can travel.
23. Feedback and Metacognitive Independence
The final goal is not a learner who waits for better feedback. It is a learner who begins generating some of the same checks independently.
After several cycles, Alicia may ask herself, “What is the base?” Beatrice may ask, “Which line actually proves this?” Ciara may ask, “Have I explained the link?” Emily may ask, “Can this task fit the time I have?”
This is where feedback starts becoming internal scaffolding. The tutor should make the self-check visible enough to transfer, then gradually stop prompting it every time. The learner’s own monitoring becomes part of the return state.
24. Feedback and Tutor Workload
More feedback is not necessarily better feedback. A tutor who writes a paragraph of commentary on every page may consume the time needed to observe learners using any of it.
A more sustainable pattern is selective depth. Mark enough to locate the important pattern. Discuss the highest-value target. Build one action opportunity. Record whether the learner could use it. Smaller surface errors can be corrected economically or clustered for later attention.
This protects the tutor’s capacity for live observation. It also helps the learner recognise that feedback has hierarchy: some comments protect a prerequisite or recurring decision; others simply polish one piece of work.
25. Feedback and Parent Expectations
Parents can reasonably expect marked work to contain guidance. The tutor should also help them see why a page with fewer comments may sometimes represent stronger teaching.
Her marked work is improving, but the important change is that she now selects direct evidence on fresh passages without me pointing to the line. I am reducing the evidence prompt and checking whether that holds next week.
Or:
He can correct sign errors once they are marked, but he is not yet detecting them independently. I am adding one self-check before final answers rather than increasing the amount of marking.
This tells the family what the feedback is changing.
26. Feedback and Learner Voice
A learner may say that feedback is helpful because it is clear, encouraging or detailed. That matters, but it answers a different question from whether the feedback changes performance.
The Learner-Voice Interpretation Gate remains the owner of that distinction. In the present loop, learner voice can help the tutor improve usability: Which comment was hard to understand? Which felt impossible to act on? Which arrived too late? Which made the next step clearer?
The tutor combines those experience signals with direct evidence from later attempts.
27. When the Action Attempt Fails
A failed return task does not prove that the feedback was bad. Several possibilities remain.
The learner may have understood the rule but lacked prerequisite knowledge. The new task may have added too much difficulty. The tutor may have removed support too quickly. The feedback target may have been too broad. The learner may have repaired the wrong part of the process. Or the original diagnosis may have been mistaken.
Treat the failed return as information. Use one discriminating question before repeating the same feedback. The correct next move may be reteaching, narrowing the target, restoring a scaffold, or revising the diagnosis.
28. When the Action Attempt Succeeds
Success should also be scoped. One fresh success may justify moving forward while keeping light monitoring. Repeated success across changed conditions may justify fading a prompt. A high-cost claim such as “this weakness is fully resolved” deserves stronger evidence than a low-cost decision such as “try the next mixed set”.
Do not turn the feedback-action loop into a ceremonial mastery test. Its purpose is to support sensible route decisions.
29. Research Limits
Feedback research is broad and heterogeneous. Effects vary by task, age, subject, feedback type, timing, source and what learners are asked to do afterwards. The EEF’s 2026 16–19 resource notes evidence gaps, including limited direct evidence on verbal feedback in that age phase and less research on group approaches.
A 2025 meta-analysis in Educational Psychology Review on instructional support in peer-feedback processes found that support can improve aspects of feedback provision and learning outcomes, while also noting that more rigorous work is needed on feedback reception. That work concerns peer feedback rather than private tutoring, so it should not be treated as direct validation of this Handbook protocol.
Source: Enhancing the Peer-Feedback Process Through Instructional Support: A Meta-Analysis.
The AERO examples are school-based demonstrations, not controlled trials of private tutoring. The Handbook therefore does not claim that one fixed feedback-action protocol has been validated for three-student Singapore tuition. The recommendation is an evidence-informed design principle: if feedback is intended to change later performance, build a reasonable opportunity to observe whether the learner can act on it.
30. Common Failure Modes
- The marked-page endpoint: returning feedback and moving on without any opportunity to respond.
- The correction marathon: asking learners to repair every error regardless of instructional priority.
- The copy-back illusion: treating a tutor-directed rewrite as independent learning.
- The duplicate-task trap: checking with an item so similar that memory of the correction supplies the answer.
- The over-delay problem: waiting so long that the original target is lost inside unrelated new content.
- The support collapse: removing legitimate access supports rather than only answer-giving help.
- The group-receipt error: assuming a shared explanation means every learner can use the feedback.
- The feedback identity label: commenting on personal traits rather than the task decision.
- The verification treadmill: turning every comment into another test.
31. The Thirty-Second Feedback-Action Gate
What exact learner decision is this feedback meant to change, what will the learner do with the feedback now, what fresh or delayed opportunity will require the same decision again, and what support can remain without performing the target for them?
If the tutor can answer those questions, feedback is no longer merely commentary.
32. The Independence Direction
At the start, the tutor may choose the target, explain the gap, design the return task and remind the learner to check. Later, the learner should increasingly perform those functions.
They begin to identify which feedback matters most. They can translate a comment into a next action. They can decide whether a correction has been tested only on the original item or on fresh work. They can notice when they are repeating an old error without waiting for the tutor to circle it.
That is the real payoff. A learner who merely receives excellent feedback remains dependent on excellent feedback. A learner who learns how to act, check and transfer becomes harder to trap in the same mistake twice.
Evidence and Connected Reading
- Education Endowment Foundation — Feedback, Developing Independent Learners (24 August 2026)
- Australian Education Research Organisation — Monitor progress
- AERO — Monitor progress: Primary and secondary (updated 23 June 2026)
- Educational Psychology Review — Enhancing the Peer-Feedback Process Through Instructional Support: A Meta-Analysis (2025)
- The Tutor Handbook — Complete Series Index
Final Compression
Feedback is not finished when the tutor has said the right thing. It is not finished when the learner has nodded. It is not necessarily finished when the original answer has been corrected.
A useful feedback cycle gives the learner a chance to act on a defined target, then looks for the same decision on work that is fresh enough to mean something. The tutor keeps legitimate access, removes only support that performs the target, and stops checking once the learner has produced enough independent receipts for the next decision.
The important question is not “Did I give feedback?” It is “What changed when the learner had to act again?”
That is the Feedback-Action Loop.
That is Tutor Handbook Volume 0163.
