The Tutor Handbook Vol No.0156 | Active-Ingredient Adaptation Gate

The Tutor Handbook · Volume 0156 · Series ID THB-0156

Series route: The Tutor Handbook — Complete Series Index.

A tutor adopts an evidence-informed routine because it has a clear educational purpose. Then the learner arrives with a different pace, different prior knowledge and a different way of showing difficulty.

The tutor does what good tutors are supposed to do: adapt.

But adaptation can improve fit or destroy function. A retrieval check becomes an open-notes review. A scaffold-fade routine keeps the scaffold because the learner looks more comfortable with it. A diagnostic probe becomes so supportive that it no longer distinguishes the competing explanations. A structured feedback cycle is shortened until the learner receives the correction but never has to use it.

The surface looks personalised. The mechanism may have disappeared.

The Active-Ingredient Adaptation Gate asks what a tutoring routine must preserve to perform its intended educational job, what can change safely around that core, and what evidence will show whether the adaptation still works for this learner.

The point is not fidelity for its own sake. The point is to adapt with enough causal understanding that personalisation does not become accidental dismantling.

Quick Answer

Before adapting an evidence-informed routine, identify its intended outcome, the mechanism that is supposed to produce that outcome and the smallest set of load-bearing features that must remain present. Separate those features from delivery details that can change for language, age, access, time, group size or learner preference. Make one bounded adaptation at a time when possible, preserve legitimate accommodations, and then check whether the target learner action still occurs. If the routine no longer produces the behaviour or evidence it was chosen for, the adaptation has crossed the boundary from tailoring into replacement.

1. What This Volume Owns

This volume owns the boundary between adaptation and mechanism loss. It does not own whether structured material may be overridden at all; The Material Override Gate decides when current learner evidence justifies departing from a planned material route. It does not own whether an intervention was actually delivered as intended; The Implementation Fidelity Check handles that distinction after delivery.

The present question comes between them: once adaptation is justified, which parts can move and which parts must stay?

2. Why “Use It With Fidelity” Is Not Enough

Strict fidelity sounds safe until the routine meets a real learner. A younger learner may need different language. A learner using an accommodation may need a different access channel. A three-student group may need a different turn-taking structure from a one-to-one setting. A session may be shorter than the original programme assumed. The school may use different notation. A learner may have mastered one component already.

Refusing all adaptation can make an evidence-informed approach less valid for the actual context. But adapting everything can make the original evidence irrelevant because the delivered approach is no longer meaningfully the same.

Good implementation therefore needs a tighter question: faithful to what?

3. The EEF Idea of Active Ingredients

The Education Endowment Foundation’s implementation guidance uses the idea of active ingredients: elements closely connected to an intervention’s theory or mechanism of change. Its guidance argues for being faithful to what matters while allowing intelligent adaptation where the mechanism does not depend on exact surface delivery.

That is especially useful in tutoring because tutoring is inherently responsive. A tutor cannot simply reproduce a classroom programme at smaller scale and call the result personalised.

The practical task is to convert “active ingredient” from an implementation phrase into a tutoring decision. What learner operation is the routine trying to produce? Which feature makes that operation unavoidable? Which features merely make delivery easier?

4. Start With the Educational Job

Never identify an active ingredient by asking which step appears in the manual. Start with the educational job.

If the routine is a retrieval check, the job may be to require the learner to bring target knowledge to mind without the answer being visible. If the routine is a diagnostic probe, the job may be to distinguish two competing explanations for an error. If the routine is scaffold fading, the job may be to transfer part of an operation from external support to learner control. If the routine is feedback, the job may be to produce a better next performance rather than merely deliver information.

Once the job is explicit, adaptation becomes easier to judge. Anything that performs the target operation for the learner is now suspicious, even if the lesson feels smoother.

5. Separate Mechanism From Wrapper

A tutoring routine usually contains both mechanism and wrapper.

Mechanism: the learner action or informational condition that is supposed to create the educational effect.
Wrapper: the wording, format, timing convention, visual design, example context or organisational structure used to deliver it.

A retrieval prompt may be spoken, written, digital or on paper. The wrapper changes. The active ingredient remains that the answer is not supplied before retrieval. A diagnostic question may use football, shopping or geometry as a context. The context changes. The active ingredient remains that different plausible learner states produce different responses.

Tutors become more adaptive when they can preserve the mechanism while changing the wrapper.

6. Some “Details” Are Actually Load-Bearing

The danger is assuming that every implementation detail is cosmetic. Sometimes frequency, delay, sequence or support condition is part of the mechanism.

If a delayed return is necessary to test retention, removing the delay changes the evidence. If a worked example is meant to reduce unnecessary search before independent practice, replacing it with unguided discovery changes the learning route. If a scaffold is supposed to fade, keeping it permanently changes the target from independent performance to supported performance.

Therefore, the tutor should ask not “Can I change this detail?” but “What causal role does this detail play?”

7. Adaptation Should Solve a Named Fit Problem

Adaptation becomes arbitrary when the reason for changing the routine is vague. “This works better for her” is not enough. Better in what way?

The language is too complex for the target skill being assessed.
The learner has already mastered one component, creating unnecessary repetition.
A legitimate access support is needed to reach the target task.
The group size changes how turn-taking must work.
The school uses a different but equivalent representation that the learner must recognise.
The original task difficulty sits outside the learner’s current measurement range.
The session length requires redistribution without reducing total meaningful practice.

When the fit problem is named, the tutor can change the feature that causes it rather than redesigning the whole routine.

8. Do Not Personalise by Removing Productive Difficulty

Some adaptations make learning feel easier by removing the very operation that needs to become stronger.

A learner dislikes mixed practice, so the tutor returns permanently to labelled topics. A learner finds retrieval effortful, so notes remain open. A learner hesitates before explanation, so the tutor supplies sentence starters that contain most of the causal chain. A learner becomes frustrated by checking, so the tutor marks every step immediately.

Comfort can be useful when unnecessary barriers are reduced. It becomes misleading when the barrier removed is the target learning operation itself.

9. Do Not Confuse Accessibility With Answer-Giving

Legitimate access supports must be preserved. The active-ingredient gate should never become an excuse to strip accommodations in the name of “pure” performance.

The correct question is whether the support performs the target cognitive or academic operation. Text-to-speech may improve access when decoding is not the target. Enlarged print may be irrelevant to the construct. Extra processing time may be an established access condition. A formula sheet may be legitimate when the target is interpretation rather than recall.

The Access-Support Boundary remains the owner for that distinction. Here the point is that adaptation should remove irrelevant barriers while preserving the mechanism that the routine needs.

10. Planned and Contingent Adaptation Are Different

AERO’s scaffold-practice guidance distinguishes planned scaffolding from support introduced responsively during instruction. That distinction maps well onto adaptation.

A planned adaptation is designed before the session because the tutor already knows something stable about access, prior knowledge, curriculum language or group format. A contingent adaptation occurs because new learner evidence appears during the lesson.

Both can be appropriate. But contingent adaptation deserves an especially clear record of what triggered it, because in-the-moment help is where tutors most easily over-assist.

11. The Tight–Loose Map

Before using a routine, write a small “tight–loose” map.

Tight: features that preserve the mechanism or validity of the evidence.
Loose: features that can change without altering the educational job.
Unknown: features whose importance is unclear and therefore should be changed cautiously.

For a fresh-item verification check, “learner has not seen the exact item” may be tight. Whether the item appears on paper or screen may be loose. Whether the check occurs after two days or five may be unknown depending on the capability and decision.

The map prevents the tutor from treating every specification as sacred or every specification as disposable.

12. Constructed Case: Alicia and Open-Notes Retrieval

This is a fictional teaching case. Alicia becomes anxious when asked to retrieve formula relationships without notes. The tutor wants to make the routine supportive and allows the formula sheet to remain visible.

The session becomes calmer. Accuracy rises. But the educational job was to discover whether Alicia could retrieve the relationship independently. The adaptation changed the construct of the check.

A better adaptation preserves access without supplying the target answer. The tutor shortens the set, uses familiar notation, explains that the check is low stakes and allows a second supported phase after the independent attempt. Emotional support changes; retrieval conditions do not.

13. Constructed Case: Beatrice and the Sentence Frame

Beatrice struggles to connect evidence to inference. A sentence frame helps: “This suggests ___ because ___.” The active ingredient is not the exact phrase. It is that Beatrice must state the relation between evidence and claim.

The tutor later adapts the frame into a visual arrow: evidence → meaning → inference. This preserves the mechanism while changing the wrapper.

A different adaptation would fail: filling in the “because” clause for Beatrice. That would produce a polished answer while the relational reasoning remains tutor-owned.

14. Constructed Case: Ciara and Three-Student Turn-Taking

Ciara’s Science practice routine was designed one-to-one: learner answers, tutor probes, learner revises. In a three-student class, using the sequence literally creates long idle periods.

The tutor adapts by giving all three learners a private first response, then rotating probes while the others compare their own reasoning against a checklist. The exact interaction structure changes.

The active ingredients remain: independent first evidence, a probe that reveals reasoning and a learner-owned revision. Group efficiency improves without losing the mechanism.

15. Constructed Case: Denise and the Shortened Feedback Loop

Denise already identifies most of her algebra errors independently. The full feedback routine—identify error, explain criterion, model repair, guided repair, fresh item—has become excessive.

The tutor adapts by removing stages Denise no longer needs. She identifies the error and criterion herself, repairs it, then completes one fresh item. The routine is shorter but more demanding.

This is intelligent adaptation because the feature removed was temporary scaffolding, not the core requirement that feedback change later performance.

16. Constructed Case: Emily and Familiar Contexts

Emily finds an abstract word problem difficult partly because the vocabulary is unfamiliar. The tutor changes the story context to something Emily understands while preserving the mathematical relationship.

This can be an appropriate early adaptation if the goal is to reveal the mathematics rather than vocabulary knowledge. But the tutor later returns to unfamiliar contexts because school problems will not always be personalised.

Adaptation can therefore be temporary. A wrapper change may expose the target mechanism first, followed by deliberate restoration of the broader condition once the learner is ready.

17. Adapt One Variable When the Consequence Matters

When an adaptation is large or the evidence base is uncertain, changing one important feature at a time makes interpretation easier.

If the tutor simultaneously changes session length, task difficulty, scaffold, feedback timing and grouping, later improvement or decline is hard to interpret. The route may still need all those changes, but the programme should acknowledge that causal attribution becomes weak.

A bounded trial can be more informative: change the representation first while holding support constant; or change support while holding task family stable.

18. Adaptation Needs an Outcome Signature

Before changing the routine, say what successful adaptation should look like.

If we simplify the language but preserve the reasoning demand, the learner should become more accurate at identifying the relationship without receiving extra clues about the method.

This outcome signature protects against adaptations that improve only surface fluency. If the learner works faster because the tutor supplied more of the route, the adaptation may feel successful while the target capability remains unchanged.

19. Monitor the Mechanism Before the Final Outcome

Long-term marks may take time to move. Implementation evidence can appear sooner.

If the routine is designed to increase independent method selection, monitor whether the learner is actually selecting methods with less cueing. If it is designed to improve error correction, monitor whether feedback leads to learner-generated repair. If it is designed to improve transfer, check whether the knowledge appears under a changed representation.

The EEF implementation framework distinguishes implementation outcomes from final outcomes for exactly this reason: we need to know whether the intended practice is happening before interpreting the final result.

20. A Routine Can Fail Even When Delivered Faithfully

The active-ingredient gate should not become a defence mechanism for favourite interventions. Sometimes the tutor preserves the hypothesised mechanism and the learner still does not improve.

Then the programme must keep alternative explanations alive. Perhaps the mechanism was wrong for this learner state. Perhaps the outcome was measured poorly. Perhaps dosage was insufficient. Perhaps the routine depends on a prerequisite not yet present. Perhaps the evidence base does not transfer cleanly to this subject or age.

Fidelity tells us what was tried. It does not guarantee success.

21. A Routine Can Succeed After Significant Surface Adaptation

The reverse is also important. If the mechanism is understood, a tutor can change language, example context, delivery medium, group arrangement or pacing while preserving the essential learner operation.

This is a mark of professional expertise. The tutor is not attached to ritual. They can explain what must remain invariant and what can move around it.

The goal of training, therefore, should not be “never adapt”. It should be “adapt from a model of how the practice is supposed to work”.

22. The Active-Ingredient Adaptation Card

Routine: What evidence-informed practice are we using?
Educational job: What learner action or evidence condition is it supposed to produce?
Mechanism: Why should this practice help?
Tight features: Which elements are load-bearing?
Loose features: Which elements are delivery choices?
Unknowns: Which features may matter but are not yet clear?
Fit problem: What specific learner or context issue justifies adaptation?
Access: Are legitimate accommodations preserved?
Adaptation: What exactly will change?
Outcome signature: What should happen if the adaptation works as intended?
Check: What fresh evidence will show whether the mechanism survived?

23. When Materials and Mechanisms Disagree

Structured materials may contain a sequence that is useful on average but poorly fitted to a particular learner’s current state. The Material Override Gate may justify deviation.

But the tutor should distinguish between skipping redundant wrapper and skipping active ingredient. If a programme uses cumulative review because later tasks depend on retrieval of earlier content, removing all review to save time can weaken the programme. If the same review uses an example context inappropriate for the learner, changing the context is less risky.

Override the material with a causal reason, not a preference.

24. When Learner Preference Conflicts With the Mechanism

Learner preference matters. It can improve participation, agency and willingness to persist. It does not automatically determine the instructional mechanism.

A learner may prefer rereading to retrieval, labelled topics to mixed practice or permanent hints to independent attempts. The tutor should listen to what the preference reveals—perhaps the task is too hard, the support fade is too abrupt or the learner does not understand the purpose—without assuming the preferred condition best supports the target capability.

Preference can change the wrapper and pacing. It should not quietly redefine the learning objective.

25. When Parent Requests Conflict With the Mechanism

Parents may understandably ask for more model answers, more notes, more worksheets or more direct help because these produce visible outputs. The tutor should explain when a requested adaptation would remove the operation the learner needs to own.

I can give her another worked model, but the evidence says she can already follow a model. The next learning job is choosing the method without the label. I’m going to keep that part unsupported and make the surrounding language clearer instead.

This is not inflexibility. It is transparent protection of the active ingredient.

26. Three-Student Tutorials Need Layered Adaptation

A small group may share one routine while needing different wrappers. Alicia may answer orally, Beatrice may benefit from a short written planning pause, and Ciara may use an approved visual support. The active ingredient can remain common: each learner produces an independent first response before discussion.

Group teaching becomes more coherent when the tutor knows what is common at mechanism level and what can differ at delivery level. Without that distinction, differentiation can fragment the class into three unrelated lessons.

27. AI and Digital Tools Make Active Ingredients More Important

Technology can automate parts of a routine so smoothly that the tutor stops seeing which operation has moved from learner to tool. An AI system can generate hints, explanations, examples and feedback. A platform can choose item difficulty and decide when to advance.

The tutor should ask who is performing the active ingredient. If the educational job is learner explanation but the tool generates the explanatory chain, the routine has changed. If the job is practice selection and the platform chooses appropriately while the learner still solves independently, the tool may support rather than replace the target operation.

The existing AI Material Verification Gate owns accuracy and safety of generated materials. This volume asks the different question of whether tool use preserves the educational mechanism.

28. Research Transfer Requires the Same Discipline

An evidence-informed practice from a classroom, another country or another age group rarely arrives with every implementation detail validated for Singapore small-group tuition. The tutor should preserve the evidence boundary.

Use research to identify plausible mechanisms and conditions. Then ask which features were central in the original evidence and which are local delivery choices. Adapt cautiously where context differs and collect learner-level evidence before broadening the claim.

Research-informed adaptation is neither copy-and-paste implementation nor free invention.

29. Research Foundation and Limits

The EEF’s third-edition implementation guidance, published in April 2024, emphasises behaviours, contextual factors and a structured implementation process. Its detailed implementation report discusses fidelity alongside intelligent adaptation and recommends specifying active ingredients so educators know where close adherence matters and where adaptation is possible. AERO’s scaffold-practice guidance, updated in May 2026, similarly distinguishes planned and responsive scaffolds and recommends fading support as proficiency develops.

These sources are school-facing rather than validations of a specific three-student tuition protocol. The application here is therefore research-informed professional judgement. This article does not claim that the tight–loose map is a validated measure or that one list of active ingredients fits every intervention.

30. Common Failure Modes

Ritual fidelity: copying every surface step without understanding purpose.
Personalisation by comfort: removing productive difficulty because the learner dislikes it.
Adaptation without a fit problem: changing routines from preference rather than evidence.
Access confused with assistance: removing legitimate accommodations or, in the other direction, giving answer-level help as though it were access support.
Unknown feature treated as loose: changing a condition whose causal role is not understood.
Everything changes at once: making later evidence uninterpretable.
Outcome without mechanism: celebrating higher immediate accuracy when the learner performed less of the target operation.
Fidelity as guarantee: assuming a well-delivered routine must work.
Favourite-method protection: using active ingredients to defend a routine against disconfirming learner evidence.

31. The Thirty-Second Adaptation Gate

What exact learner or context problem am I adapting for, what is this routine supposed to make the learner do, which feature makes that operation happen, what can change without supplying the answer or changing the construct, and what fresh evidence will tell me whether the mechanism survived?

If the tutor can answer that, adaptation becomes professional design rather than improvisation.

32. The Independence Direction

The learner should eventually inherit some of this reasoning. Mature students can learn to ask why a study routine works rather than treating it as superstition.

“I can shorten this practice set, but I still need to retrieve before checking.” “I can change the context, but I still need to choose the method without a label.” “I can use text-to-speech because decoding is not the target here, but I still need to construct the explanation myself.”

That is deeper than compliance. It is the beginning of learner-owned implementation judgement.

Evidence and Connected Reading

Final Compression

Adaptation is not the opposite of fidelity. Good adaptation depends on knowing what fidelity is for.

Name the educational job. Identify the mechanism. Mark the load-bearing features. Preserve access. Change the wrapper when fit requires it. Keep the target operation with the learner. Check the mechanism before celebrating the outcome. Change course when evidence says the routine is not working.

The strongest tutor is neither the one who follows every routine exactly nor the one who personalises everything. It is the one who knows which parts may bend because they understand which parts must not disappear.

That is the Active-Ingredient Adaptation Gate.

That is Tutor Handbook Volume 0156.

Small Group Tutorials

The Tutor Handbook Vol No.0156 | The Active-Ingredient Adaptation Gate — How a Tutor Changes an Evidence-Informed Routine for One Learner Without Removing the Parts That Make the Routine Work