The Tutor Handbook Vol No.0079 | The Implementation Fidelity Check — How a Tutor Decides Whether a Learning Route Failed or Was Never Actually Run as Intended

The Tutor Handbook · Volume 0079 · Series ID THB-0079

The Tutor Handbook: Complete series index.

The tutor tried the new route.

It did not work.

Or did it?

The plan called for three short mixed-practice sessions across a week. Only one happened. The plan called for the learner to attempt before receiving a method cue. The tutor supplied hints almost immediately because the lesson was running late. The plan called for a later independent return. Homework was checked through a parent-supported attempt instead. The plan called for one narrow feedback target. Three different corrections were layered onto the same task.

The learner’s result is disappointing. The tutor says the intervention failed.

That conclusion may be premature because the route that failed is not the route that was designed.

The Implementation Fidelity Check asks whether the active ingredients of a learning route were actually delivered with enough of the intended dose, sequence, support boundary, practice condition and learner participation for the tutor to judge the intervention itself rather than a distorted version of it.

This does not mean tutors should follow a script mechanically or refuse to adapt. Good teaching is responsive. A learner may reveal that a planned example is too difficult, that the support level is wrong, that school has changed the immediate demand, or that a task needs to be shortened. Intelligent adaptation is part of professional judgement.

The problem is different: when the tutor changes so many defining elements that the intervention’s mechanism no longer operates, then later describes the outcome as evidence about the original intervention.

Quick answer

Before deciding that a learning route worked or failed, identify its active ingredients and check whether those ingredients were actually present.

For a tutoring intervention, active ingredients may include the type of learner attempt, the timing and amount of support, the practice spacing, the requirement to choose rather than imitate, the feedback target, the delayed return, the expected learner explanation, or the conditions under which performance is sampled.

Then ask five questions. Was the route delivered often enough? Was it delivered in the intended order? Were the critical supports present or absent as designed? Did the learner actually engage in the target operation? Did unplanned changes alter the mechanism? If the answer to these questions is mostly no, the responsible conclusion is not “the intervention failed.” It is “we do not yet know what the intended intervention would have done because implementation drift changed the treatment.”

If implementation was sufficiently faithful and the predicted receipts still did not appear, then the failure is more informative. The route may need to be changed, abandoned, rediagnosed or replaced.

1. What this volume owns

This volume owns the question that comes immediately before a route is judged.

Volume 0054, The Drift Check, asks whether an adopted routine has quietly changed over time. This volume is related but narrower and more evaluative. It asks whether the route was delivered closely enough to its intended mechanism that a result can fairly be attributed to the route.

Volume 0052, The Trial Run, owns bounded testing of a new route. The Implementation Fidelity Check asks whether the trial actually tested what it claimed to test.

Volume 0051, The Intervention Cost, owns side effects. Fidelity is different. A route can be delivered faithfully and still create costs; a route can also look ineffective because it was never delivered as intended.

2. Design is not delivery

A beautiful learning plan can exist on paper and never exist in the learner’s experience.

The tutor writes:

Attempt independently → receive one discriminating hint if needed → complete a fresh item → return after three days → mix with neighbouring methods.

What happens in practice?

The learner hesitates. The tutor immediately names the method. The fresh item is skipped because the lesson is ending. Three days later the planned return is displaced by urgent homework. The next week the learner completes blocked practice under the chapter heading rather than a mixed set.

The written plan and lived intervention are now different objects.

3. Failure of implementation is not failure of mechanism

This distinction matters because different failures require different repairs.

If the mechanism is wrong, choose a different intervention. If the mechanism is plausible but the route was implemented inconsistently, fix delivery before replacing the idea. If the route was delivered faithfully and still failed, the result becomes valuable evidence against the current hypothesis. If fidelity cannot be established, reduce the certainty of the conclusion.

Education research routinely separates intervention design from implementation. The Education Endowment Foundation’s current implementation guidance distinguishes active ingredients that should be protected from aspects that can be adapted intelligently. Its implementation report describes fidelity as including structural elements such as dosage and training as well as dynamic elements such as whether key teaching strategies were actually present. That is a school-scale framework, not a validation of this tutoring protocol, but the distinction translates cleanly.

4. First identify the active ingredient

A tutor cannot check fidelity if the route is described only as “do more practice,” “use feedback,” “work on confidence” or “teach differently.”

The active ingredient is the part of the route believed to cause the intended change.

For mixed practice, the active ingredient may be the requirement to discriminate between methods rather than execute a method that has already been announced. For scaffold fading, the active ingredient may be the learner performing a previously supported decision with less answer-giving help. For delayed retrieval, the active ingredient includes time passing before the learner reconstructs knowledge. For feedback, the active ingredient may be a specific information gap followed by an opportunity to act on the feedback. For studying, the active ingredient may be learner-owned prioritisation rather than tutor-built scheduling.

If the tutor cannot name what is supposed to make the route work, fidelity becomes a checklist of surface features rather than protection of mechanism.

5. Fidelity is not perfect sameness

Good implementation is not rigid cloning.

A Mathematics tutor can change the numbers in an example. An English tutor can choose a passage that interests the learner. A Science tutor can shorten a task because fatigue is changing performance. A study plan can move from Tuesday to Wednesday because school workload changed.

The crucial question is whether the adaptation preserves the active ingredient.

If the aim is independent method selection, changing the numbers preserves the mechanism. Naming the method does not. If the aim is delayed retrieval, changing the day by one day may preserve the mechanism. Replacing the delayed attempt with same-session rereading does not. If the aim is learner-owned planning, changing the planner format may be harmless. Having the tutor choose all priorities is not.

6. Dose matters

An intervention can be conceptually sound and practically too small to have a fair chance.

A learner receives one two-minute retrieval attempt and then returns to rereading for the rest of the month. The tutor later concludes retrieval practice did not help. A learner uses one mixed set after weeks of blocked exercises and the tutor concludes interleaving confused them. A tutor attempts one scaffold fade, restores the full prompt immediately after hesitation and concludes the learner is not ready for independence.

These conclusions may be correct, but the dose is too small for confidence.

At programme level, the National Student Support Accelerator’s Tutoring Quality Standards treat dosage, setting, instructional materials, formative assessment, tutor coaching and other implementation characteristics as separate quality elements. Its public Design Badge explicitly warns that alignment of programme design does not show how well implementation occurs because real sites differ in tutor reliability, time constraints, student absence and other conditions. This is precisely the distinction a tutor needs at lesson scale: intended design and lived dose are not interchangeable.

7. Sequence matters

The same ingredients in the wrong order can produce a different learning condition.

If a model answer is shown before the learner’s first attempt, it changes the evidence available from the attempt. If feedback arrives before the learner has exposed their reasoning, the tutor loses diagnostic information. If time pressure is introduced before method selection is stable, the learner may practise haste rather than performance. If a challenging extension arrives before a prerequisite is secure, difficulty can hide the original weak link.

Fidelity therefore includes the order in which teaching, attempt, feedback, reattempt, delay and transfer occur when that order is part of the mechanism.

8. Support boundaries matter

A route designed to test independent method selection cannot be considered faithfully implemented if the tutor repeatedly supplies method cues. A route designed to fade a planning scaffold is not faithfully implemented if the parent quietly rebuilds the same scaffold at home. A route designed to preserve legitimate access support is also not faithfully implemented if the tutor removes an accessibility accommodation that was never meant to be faded.

Volume 0073, The Access-Support Boundary, defines which assistance belongs relative to the target. Volume 0078, The Support Provenance Check, asks what help actually contributed to returned work. Fidelity uses those distinctions to decide whether the route ran under its intended support conditions.

9. Learner participation is an active ingredient when the route depends on it

An intervention can be delivered by the tutor and still not be enacted by the learner.

The tutor asks for self-explanation, but the learner repeats the worked example. The tutor assigns error analysis, but the learner copies corrections from the answer key. The tutor asks the learner to choose study priorities, but a parent has already decided the schedule. The tutor creates a peer-comparison task, but one learner does all the explaining while the others agree.

If the mechanism requires learner production, fidelity cannot be assessed only from whether the tutor delivered the instruction. The target learner operation must actually occur.

10. Materials can preserve or destroy fidelity

A tutor plans mixed method selection and then uses a worksheet whose headings announce the method for every question. The lesson plan says mixed; the material says blocked.

A tutor plans independent inference and uses a passage where every relevant sentence is highlighted. The questions may be fresh, but the material performs part of the evidence selection.

A tutor plans scaffold fading but prints an old worksheet containing the full planning frame. The stale resource quietly restores retired support.

Volume 0066, The Configuration Snapshot, helps expose what the current route is supposed to be. Fidelity asks whether the materials actually instantiate that route.

11. Time pressure can create implementation drift

Many fidelity failures are not conceptual. They are logistical.

The lesson starts late. Homework questions consume twenty minutes. A parent needs a discussion. One learner arrives without material. The tutor shortens independent practice, drops the delayed check and gives more direct prompts to finish the planned content.

These adaptations may be reasonable in the moment. The mistake comes later if the tutor evaluates the route as though the intended exposure occurred.

Implementation notes can therefore be simple and honest: “Planned mixed set reduced from six items to two because school homework required urgent repair; no adoption decision this week.” That sentence protects the learner from a false conclusion.

12. Attendance and spacing matter

A route may depend on repeated encounters. If the learner misses one of two weekly sessions, the intended sequence changes. If a planned three-day retrieval return becomes a three-week gap, the task may become much harder. If home practice never happens, the tutor’s planned dose is not the learner’s received dose.

Do not blame the learner automatically. Absence, competing school demands, family conditions and scheduling constraints can all affect delivery. Fidelity is descriptive before it is evaluative. The point is to know what actually happened.

13. Adaptations should be labelled as adaptations

A tutor sees that the learner is overloaded and reduces the number of mixed items. Good. A tutor notices the learner lacks a prerequisite and inserts a short explanation. Good. A tutor realises the original feedback target was wrong and changes it. Good.

Professional adaptation is not a fidelity failure by default. The tutor should record whether the adaptation preserved the active ingredient, weakened it, strengthened it or replaced it.

This leads to a useful four-way distinction:

• Faithful adaptation: surface or logistical features change while the active ingredient remains intact.
• Protective adaptation: the route is temporarily modified because safety, access, fatigue or another legitimate condition requires it; interpretation is adjusted accordingly.
• Mechanism-changing adaptation: a defining element changes, so the intervention is no longer the original intervention.
• Implementation failure: the intended route simply does not occur often enough or clearly enough to evaluate.

14. Fidelity should protect active ingredients, not cosmetic details

Two tutors can deliver the same active learning mechanism with different voices, examples and personalities.

One may use a diagram. Another may use a physical analogy. One may ask the learner to explain orally. Another may ask for a written justification. If the target operation remains the same and the evidence condition remains comparable, variation in style may be compatible with fidelity.

This is especially important in tutor handoffs. Volume 0063, The First Change, separates tutor style from learning route. Fidelity should not demand imitation of the previous tutor’s personality. It should protect the mechanism that current evidence still supports.

15. Worked case: Alicia and the mixed-practice route that was never mixed

This is a fictional teaching case.

Alicia’s tutor decides to improve algebra method selection by mixing three familiar question families. After two weeks, Alicia still chooses methods poorly on a school paper. The tutor is ready to declare mixed practice ineffective.

The fidelity check reconstructs what actually happened. During tuition, the first “mixed” worksheet was divided into three sections, each with a heading that named the method. On the second worksheet, the tutor verbally reminded Alicia which method family each question resembled. The third planned mixed set was replaced by urgent school homework.

Alicia received varied execution practice. She received very little independent method-selection practice.

The conclusion changes. The evidence does not yet show that mixed selection training failed. It shows that the defining selection demand was not consistently present. The tutor redesigns one short set with no topic labels and protects the first choice from cues. Only then does the route receive a fair trial.

16. Worked case: Beatrice and feedback without reattempt

This is a fictional teaching case.

Beatrice receives detailed feedback on comprehension explanations for three weeks. Her answers remain weak. The tutor concludes that feedback “doesn’t work for her.”

The fidelity check reveals that most feedback was written after the task and discussed briefly at the end of lessons. Beatrice rarely produced a fresh answer using the feedback. Corrections were sometimes copied into the original response.

The active ingredient was supposed to be feedback followed by changed learner action. Volume 0072 calls the later changed attempt the Feedback Receipt. The route mostly delivered information, not uptake opportunities.

The tutor changes implementation before changing the feedback theory: one narrow feedback point, followed by an immediate fresh sentence or question, then a delayed changed-context return.

17. Worked case: Ciara and scaffold fading that never faded

This is a fictional teaching case.

Ciara is meant to move from model-supported Science explanation to independent causal reconstruction. The tutor removes the printed model answer but continues asking leading questions whenever Ciara pauses.

Ciara’s answers stay strong. The tutor believes scaffold fading succeeded.

The fidelity check asks what support actually disappeared. The visible model disappeared; the tutor’s verbal cueing replaced it. The mechanism-changing decision support remained.

The learner has therefore not yet demonstrated the intended unsupported causal reconstruction. The next step is not harsher removal of all help. It is a cleaner support boundary: preserve legitimate task access, reduce leading questions, allow a bounded think period, and observe what Ciara can construct before the next prompt.

18. Worked case: Denise and timing introduced at the wrong dose

This is a fictional teaching case.

Denise is ready to add moderate time pressure to stable Additional Mathematics work. The planned route is progressive: short representative sections first, then longer sections if accuracy holds.

Instead, a busy week leads the tutor to skip the short sections and assign a full timed paper. Accuracy collapses. The tutor says Denise is not ready for timing.

The result does show that full-paper pressure currently exceeds stable performance. It does not test the planned progressive timing route. The tutor should not confuse failure at a higher dose with evidence that the lower-dose intervention would fail.

19. Worked case: Emily and the self-directed study plan built by the tutor

This is a fictional teaching case.

Emily’s route aims to transfer weekly study prioritisation to her. The tutor plans to review Emily’s first draft, ask questions and let her revise it.

Each week the lesson runs short. To save time, the tutor creates the schedule and asks Emily whether it looks reasonable. Emily follows it inconsistently. After a month, the tutor concludes she cannot self-plan.

The fidelity check reveals that learner-owned prioritisation never occurred. Emily was asked to comply with a tutor plan, not construct and revise her own. The intervention failed to be implemented at the exact point where independence was meant to grow.

20. Worked case: Faith and an independence routine implemented only on easy tasks

This is a fictional teaching case.

Faith is learning to start tasks independently before seeking help. The tutor applies the routine on familiar worksheets but reverts to immediate guidance whenever a task looks unusual.

Faith becomes independent on familiar tasks and remains dependent on novel ones. The tutor describes this as a failure of the independence routine.

The fidelity question is whether the route ever included appropriately unfamiliar but accessible tasks. If the target claim included independent start under mild novelty, the implementation sample excluded the exact condition the route was supposed to train.

21. The fidelity check should be brief enough to use

A private tutor does not need a research-monitoring department. A useful check can fit on one card.

• Target change: What learner capability is the route meant to change?
• Active ingredient: What part of the route is believed to cause that change?
• Dose: How much exposure was intended and how much actually occurred?
• Sequence: Did attempt, support, feedback, reattempt and return occur in the intended order?
• Support condition: Was help at the intended level?
• Practice condition: Did the materials actually require the target operation?
• Learner participation: Did the learner personally perform the operation?
• Adaptations: What changed, and did the active ingredient survive?
• External interference: Did school, home, absence or workload materially alter delivery?
• Receipt: What predicted evidence appeared or failed to appear?
• Judgement: Mechanism failure, implementation failure, ambiguous result or successful route?

22. Use four possible verdicts instead of “worked” or “failed”

A binary verdict is often too crude.

Implemented and effective: the active ingredients were present and the predicted receipts appeared.

Implemented and ineffective: the route was delivered with sufficient fidelity but the predicted receipts did not appear or meaningful costs outweighed gains.

Not implemented sufficiently: the intended route did not occur often, clearly or consistently enough to judge the mechanism.

Mechanism changed through adaptation: the route became a different intervention. Evaluate the intervention that actually ran rather than pretending it was the original.

These categories are practical descriptors, not validated measurement scales. Their purpose is to make the next decision more honest.

23. Fidelity and personalisation are not enemies

The word fidelity can sound like standardisation. Tutoring often promises personalisation. The two can coexist if the tutor distinguishes mechanism from surface.

The active ingredient can remain stable while examples, pacing, language, interests and task context adapt to the learner. A learner who needs more processing time can receive it while still performing the same target reasoning. A learner with stronger prior knowledge can receive harder examples while still encountering the same method-selection demand. A learner using accessibility support can retain it while answer-giving cues fade.

Personalisation asks what form the route should take for this learner. Fidelity asks which elements must remain intact for it still to be the route we think we are evaluating.

24. Fidelity should not protect a bad plan

A dangerous interpretation of fidelity is: “Follow the plan even when the learner shows it is wrong.”

No.

If the learner becomes distressed, the task is inaccessible, the diagnosis is contradicted, a legitimate accommodation is missing, the school context changes materially or the intervention is creating a clear cost, adapt or stop. Protect the learner before the purity of the intervention.

Then label the change honestly. “Trial stopped because the prerequisite was missing” is better than forcing completion so the tutor can claim fidelity. “Support increased because the task was inaccessible” is better than treating avoidable failure as data.

25. Fidelity can explain why two tutors get different results

Two tutors may say they use the same method and produce different outcomes.

One gives the learner five seconds before prompting; the other gives thirty. One uses mixed questions without labels; the other groups them by method. One asks for a fresh reattempt after feedback; the other discusses the correction verbally. One schedules a delayed return; the other moves on after immediate success.

The names of the intervention are identical. The active ingredients are not.

This is why professional coaching should observe live teaching rather than rely only on tutors describing what they believe they do. Volume 0075, The Coaching Receipt, asks whether professional feedback changed later practice. Fidelity adds a useful question for coaching: did the intended teaching move actually occur in the form that mattered?

26. Fidelity in three-student tutorials

Small groups create a special implementation risk: an individual route can be diluted by group convenience.

Alicia is supposed to work without a method cue, but the tutor’s explanation to Beatrice reveals the method aloud. Ciara is meant to attempt independently, but a peer answers before she begins. Denise is supposed to experience uninterrupted timing, but the tutor pauses the whole group repeatedly. Faith is meant to practise help-seeking after one self-check, but the class routine encourages immediate questions.

Volume 0077, The Three-Learner Orchestration, owns the moment-to-moment distribution of attention and activity. Fidelity asks whether the individual intervention survived the shared room.

27. Parent actions can change implementation

A tutor plans learner-owned study scheduling. At home, the parent reorganises the timetable every evening. A tutor plans independent homework attempts before help. At home, a parent sits beside the learner from the first question. A tutor plans model answers only after a first draft. At home, the model answer is opened before writing begins.

The parent may be acting with excellent intentions. The problem is not blame. The problem is that the home condition has changed the intervention.

Parent communication should therefore explain the active ingredient, not merely the homework instruction. “Please let her struggle” is vague and potentially unhelpful. “Please let her choose the first method before offering a hint; if she is still blocked after she records the relationship she sees, help is welcome” is an implementable boundary.

28. AI and digital tools can change implementation silently

A route designed around learner-generated explanation can become AI-assisted explanation between lessons. A route designed around retrieval can become search-supported recall. A route designed around learner planning can become app-generated scheduling.

The Support Provenance Check asks what the tool contributed. Fidelity asks whether the intended learner operation still occurred often enough for the intervention to retain its mechanism.

The answer is not necessarily to remove the tool. The route may be adapted so that the learner performs the target operation before using the tool, or uses the tool in a role that preserves the active ingredient. But the adaptation should be explicit enough that the tutor knows which intervention is now being evaluated.

29. The Configuration Snapshot can serve as the intended-state reference

Fidelity requires a comparison between intended and actual delivery. The Configuration Snapshot can provide the intended state: target, support, practice condition, evidence rule, active priority and learner ownership.

After a week, the tutor can ask whether that configuration actually ran. If not, the snapshot should not remain a fictional description of practice. Either delivery must change or the configuration must be updated to the route that is genuinely sustainable.

30. The Change Bandwidth can explain implementation overload

Sometimes fidelity fails because the system is carrying too many changes.

A new timing routine, support fade, study planner, feedback code and checking process all launch in the same fortnight. The tutor forgets parts. The learner confuses rules. Parents implement old instructions. Materials lag behind.

Volume 0067, The Change Bandwidth, owns the rate and amount of active change. Fidelity failure can be a receipt that the route exceeded the system’s implementation capacity.

31. Do not punish the learner for adult implementation failure

A learner is told to use a new checking routine. Different adults explain it differently. The worksheet uses the old format. The tutor forgets to review it for two weeks. Then the learner is criticised for “not following the system.”

That is not a fair conclusion.

Before attributing poor adherence to motivation or discipline, check whether the route was clear, stable, practicable and consistently supported by the adults and materials around the learner.

32. Parent communication: “We are checking whether the plan really happened”

A useful parent update can sound like this:

The first result is weaker than I hoped, but I do not yet want to conclude that the strategy failed. Two of the planned practice returns were displaced by school work and the final attempt still had more prompting than intended. I am tightening the delivery for one more bounded cycle. If the route is then implemented properly and the same problem remains, I will change the intervention rather than simply repeat it.

This communicates accountability without pretending the tutor already knows the cause.

33. Learner communication: separate “the plan failed” from “we did not really run the plan”

Learners often internalise failed routines as personal failure.

A tutor can say:

We planned to practise this three times with less prompting, but we only managed one proper attempt. That means I cannot fairly say you failed the method. We need a cleaner run before we decide whether to keep it.

This language protects learner identity while preserving responsibility for completing the next agreed route.

34. Research foundation: implementation fidelity is part of educational evaluation

The Education Endowment Foundation’s implementation guidance explicitly treats fidelity as part of judging implementation and advises educators to protect the active ingredients that make an intervention work while allowing intelligent adaptation around them. Its guidance notes that fidelity can include structural components such as dosage and dynamic components such as whether key pedagogical strategies are actually present.

That principle is supported more broadly by educational implementation and process evaluations. Current EEF trial protocols routinely examine implementation quality, reach, adaptation, differentiation and setting fidelity alongside impact rather than assuming programme design automatically equals programme delivery. These are programme-level research practices; this volume does not claim that a private tutor can reproduce a formal implementation study. It borrows the evidence discipline: before attributing an outcome to an intervention, check whether the intervention substantially occurred.

35. Research foundation: design alignment is not implementation quality

The National Student Support Accelerator’s current Tutoring Program Design Badge makes an unusually clear distinction: its design badge indicates alignment to research-informed quality standards but does not show how well implementation occurs because local conditions vary. That distinction is highly relevant to tutors. A route can be well designed and poorly delivered. A tutor’s reputation, qualifications or beautiful materials cannot substitute for evidence that the active teaching practices actually reached the learner.

NSSA’s broader Tutoring Quality Standards also separate elements such as training, coaching, formative assessment, progress measurement, materials, dosage and programme improvement. The practical lesson is that tutoring quality is not one feature. Implementation is a system of conditions.

36. Research foundation: monitoring is for adjustment, not paperwork

AERO’s Monitor Progress guidance links checks for understanding to additional instruction, guidance and feedback. Fidelity monitoring should serve the same practical purpose. Do not create elaborate compliance records that consume more time than the learning route. Record the few implementation features that would change interpretation or the next decision.

The tutor is not auditing for the sake of proving that procedures were followed. The tutor is trying to answer a useful question: should we keep the mechanism, improve delivery, adapt the route, or abandon the intervention?

37. Common failure: blame the method when the dose was almost zero

One attempt is made, the result is weak, and the tutor declares the approach ineffective.

A low dose can still reveal immediate harm or obvious mismatch. But when the mechanism requires repeated learning opportunities, one partial exposure may not be a fair evaluation. Distinguish “insufficient dose to judge” from “sufficient evidence of failure.”

38. Common failure: protect the method by blaming implementation forever

The opposite error is equally dangerous.

Every weak result is explained away: the learner was tired, the homework was interrupted, the dose was not perfect, the parent helped too much, school changed the topic, one session was missed. The method becomes impossible to falsify because implementation is never considered good enough.

Set a reasonable fidelity threshold before the trial. Decide which active ingredients truly matter and what minimum exposure is sufficient for a judgement. Once the route has received a fair implementation and predicted receipts remain absent, change the hypothesis.

39. Common failure: measure surface compliance instead of mechanism

The learner filled in the planner. The learner completed the worksheet. The tutor delivered the three sessions. The checklist was used.

These observations may say little if the active ingredient was learner prioritisation, method discrimination, independent retrieval or feedback-driven revision.

Fidelity is not paperwork completion. It is preservation of the mechanism that justified the intervention.

40. Common failure: confuse learner resistance with implementation failure

A route can be implemented faithfully and the learner can dislike it. A learner can also participate minimally, which may be evidence that the route is poorly fitted, poorly explained, excessively demanding or simply requires more support.

Do not label the learner “resistant” before checking whether the task is accessible and meaningful. Do not label the tutor’s delivery faithful merely because the instructions were spoken. Learner participation is part of implementation when the mechanism depends on active production.

41. Common failure: change several active ingredients and keep the old name

The tutor says they are still using “retrieval practice,” but students now look at notes during every attempt. The tutor says they are using “mixed practice,” but worksheet headings identify each method. The tutor says support is fading, but leading questions have replaced the visible scaffold. The tutor says the learner owns planning, but priorities are pre-selected.

When active ingredients change, rename the intervention in your own records or update the configuration. Do not preserve a convenient label after the mechanism has disappeared.

42. A practical tutor fidelity audit

• What exact learner change was this route designed to produce?
• What active ingredient was supposed to cause that change?
• What minimum dose was actually planned?
• How much dose was actually received?
• Did the key steps occur in the intended sequence?
• Was the support level faithful to the target?
• Did the material require the intended operation?
• Did the learner personally perform the operation?
• What adaptations were made?
• Which adaptations preserved the mechanism?
• Which adaptations changed it?
• Did outside conditions materially alter implementation?
• What receipt was predicted?
• Did the receipt appear?
• Is the correct verdict effective, ineffective, not sufficiently implemented or transformed into a different intervention?
• What must change next: mechanism, delivery, dose, support, materials or diagnosis?

43. A parent fidelity audit

• What was the tutor actually trying to change?
• What part of the plan was essential rather than optional?
• Did the planned practice really happen?
• Did home support accidentally replace a learner decision?
• Were missed sessions or school demands large enough to change the dose?
• Is the tutor saying the method failed, or that implementation was too inconsistent to judge?
• How many clean attempts will be enough before the route is reconsidered?
• What would make the tutor stop defending the method and choose another one?

44. A learner fidelity audit

• What am I supposed to be practising?
• What part am I meant to do before getting help?
• Did I actually get enough chances to practise that part?
• Did the task keep changing?
• Did someone else do the difficult decision for me?
• What happened when I tried it independently?
• What will we keep the same next time so we can tell whether the plan works?

45. What this volume does not authorise

The Implementation Fidelity Check does not demand rigid scripts, prohibit teacher judgement, blame learners for imperfect participation or require formal research measurement in private tuition. It does not validate a route merely because it was followed faithfully. A faithfully delivered intervention can still be wrong, ineffective, inequitable, unnecessarily burdensome or harmful.

It also does not permit the tutor to defend a preferred method indefinitely by declaring every poor result an implementation problem. Fidelity is a gate to interpretation, not an escape from falsification.

46. The ethical standard

Learners should not become the test site for adult inconsistency.

If the tutor changes the rules, misses the planned practice, restores old support, uses materials that contradict the route and then blames the learner when improvement does not appear, the learner is carrying the cost of an implementation failure they did not create.

Professional tutoring should own its part of the system. That means distinguishing a learner who did not yet improve from an intervention that was not delivered, and distinguishing both from a well-implemented route that genuinely failed.

Before asking whether the learner failed the plan, ask whether the plan ever reached the learner in the form that was supposed to make a difference.

Evidence and connected reading

• The Tutor Handbook Vol No.0078 | The Support Provenance Check
• The Tutor Handbook Vol No.0077 | The Three-Learner Orchestration
• The Tutor Handbook Vol No.0076 | The Evidence Sample
• The Tutor Handbook Vol No.0054 | The Drift Check
• The Tutor Handbook Vol No.0052 | The Trial Run
• Education Endowment Foundation | A School’s Guide to Implementation
• National Student Support Accelerator | Tutoring Quality Standards
• National Student Support Accelerator | Tutoring Program Design Badge
• Australian Education Research Organisation | Monitor Progress

Final principle

A result is never only a result.

It is the result of a learner meeting an intervention under actual conditions.

When the route underperforms, first check what truly ran. Protect the active ingredient. Distinguish cosmetic adaptation from mechanism change. Check dose, sequence, support, participation and materials. If the intended intervention received a fair implementation and still failed, change it. If it never really ran, do not pretend you have already learned what it would have done.

The Implementation Fidelity Check protects one of the most important distinctions in professional tutoring: a bad idea, a badly delivered idea and an idea that was quietly replaced by something else are not the same failure.