🧭 Tutor's handbook · this page is for you, Nat, not for Kelcy
10-week home program

Building word-problem thinking

A 10-week at-home track that runs alongside your sessions. Your sessions follow what she is doing in class. This builds the thing class rarely teaches directly: how to think your way into an unfamiliar problem and not freeze.

What this is, and what it is not

One problem a week, done at home, where Kelcy records herself thinking out loud while she works. The goal is not a tidy correct answer. The goal is to make her thinking visible, so she can hear her own detective at work and you can see exactly where she gets stuck.

It runs in parallel with your live sessions, which stay reactive to class. This track is deliberate and slow-burning: the same Detective ritual every week, on complex unfamiliar problems, with your help fading out over the ten weeks until she runs the whole thing herself.

The one-line version

Same routine every week + problems that don't look like the last one + your prompts slowly disappearing = a kid who can start a problem she has never seen before.
Why it is built this way

Four pieces of research shape the whole program. You do not need to teach any of this to Kelcy, it just explains the choices.

IMPROVE · Mevarech & Kramarski
Gives us → the Detective skills

Students who are taught to ask themselves four kinds of question (do I understand it? have I seen one like it? what strategy fits? does my answer make sense?) outperform those who just practise. Those four are exactly Open the Case, Check the Case Files, Pick Your Gadget, and the Sniff Test.

Schoenfeld · monitoring
Gives us → the Three Knocks

Novices pick one approach and ride it off a cliff. Experts constantly stop and ask "what am I doing and is it working?" Teaching that one mid-problem check is the highest-leverage move, so it gets its own ritual.

EEF · metacognition guidance
Gives us → the non-negotiables

Metacognition has to be taught inside the maths, not as generic "thinking tips", the teacher has to model it out loud, and prompts must be faded or the child becomes prompt-dependent. That is the spine of the ten weeks.

Chinofunga & Chigeza · 2024
Gives us → the questions

What makes a question "complex unfamiliar": a real-world story, no clear path, needing you to combine topics, often work backwards, multi-step, and justify that the answer is reasonable. That is the recipe for the harder weeks.

Anchored in IMPROVE (Mevarech & Kramarski), Schoenfeld's monitoring work, the EEF metacognition and self-regulated learning guidance, and Chinofunga & Chigeza (2024, JCU / MERGA) on complex unfamiliar problem design. QCAA chilli bands: 1 simple familiar, 2 complex familiar, 3 complex unfamiliar.

The non-negotiables

If you keep only these, the program still works. Drop these and it quietly stops working.

What one week looks like

About 20 to 30 minutes at home. Same shape every week so it becomes automatic.

  1. ~5 minFluency warm-up. Two or three quick familiar sums (times tables, a percentage, a simple solve). Wake up the basics.
  2. ~3 minOpen the case, no solving. Read it twice, underline the question, circle the clues. Pen down. She is only allowed to decode, not answer, yet.
  3. ~12 minSolve while recording. She talks out loud the whole time (voice memo on). Working goes in the journal. Stuck = Three Knocks, then a smaller case if needed.
  4. ~3 minSniff test. Read the answer back into the story. Right size? Right units? Could it happen?
  5. ~4 minCase notes. What kind of case was it, what got her unstuck, a note to Future Kelcy. This is the archive that makes next time easier.

🎙️ The recording is the point

She records a voice memo talking through her thinking as she works. You are not marking it. You are listening for where the thinking stalls or skips a step. A silent stuck moment is invisible on paper but obvious on tape. Keep the memos, they are your data across the ten weeks.
The 10-week arc

Three checkpoints (red, orange, green) bracket the program so you can measure movement, not just feel it.

WeekFocusCase typesChilliYour prompting
1 ● DiagnosticThe Cold Case: see where she breaks downMixed (one per blocker)1 to 3None. No help, just record.
2You model the full ritual out loudTank2You think aloud; she echoes
3Decode drill, name the case typeTank + Backwards2Full scaffold, all 6 skills visible
4Build the toolkitContainer + Crossover2 to 3Full scaffold, she leads
5 ● Check-inThe Case Review: measure movement, retuneMixed (parallel set)1 to 3None. No help, just record.
6Step into the unfamiliarMixed3Prompts fade: she self-asks first
7Complex unfamiliarMixed3Minimal: only "Three Knocks?"
8Complex unfamiliarMixed3Minimal
9Near-soloMixed3You observe only
10 ● FinalCase Closed: same problems as Week 1Mixed (Week 1 set)1 to 3Solo, then compare to Week 1
How your help fades

This is the whole game. Your prompting starts high and is gone by the end. Picture it as a slope, not a switch.

Wk 1
none*
Wk 2
you model
Wk 3
full
Wk 4
full
Wk 5
none*
Wk 6
she asks
Wk 7
minimal
Wk 8
minimal
Wk 9
observe
Wk 10
solo

* Weeks 1, 5 and 10 are the checkpoints. No prompting there on purpose, because you are measuring what she can do unaided. The fade happens across the teaching weeks in between.

The three checkpoints

Same shape, spread across the program, so you have three honest snapshots instead of a vibe.

● Week 1 · Diagnostic

Five carefully chosen problems, no help, recorded. Each one probes a different blocker. Gives you her starting profile: is it decoding, connecting, strategy, monitoring, fundamentals, or checking?

Open the diagnostic →

● Week 5 · Check-in

Five fresh problems that probe the same six blockers as Week 1. Re-record, compare profiles, and retune which blocker the back half leans into.

Open the check-in →

● Week 10 · Final

Re-run the exact Week 1 problems. Nine weeks on, she meets the same cases that beat her and cracks them solo. Powerful evidence, and a real confidence moment.

Uses the Week 1 set →
Listening to the voice memos

You are not grading. You are locating the stall. As you listen, you are sorting what you hear into six layers (the diagnostic page has the full map and a sheet to fill in):

Decode (does she work out what is being asked?) · Connect (does she link it to a type she knows?) · Strategy (does she pick a sensible tool?) · Monitor (does she notice when she is off track?) · Fundamentals (are fractions, negatives, percentages, or algebra the real wall?) · Evaluate (does she check the answer makes sense?)

Whichever layer shows up most is where the teaching weeks should lean. A right answer with no checking is a finding. A wrong answer with great process is a win. Listen for the thinking, not the result.

When a case goes wrong, mid-week

The first move is never to fix it. It is to find where it broke. "Talk me through it, where do you reckon it went sideways?" Getting her to locate her own slip is self-explanation, the highest-value thing you can do, and it is "Dead Ends Are Clues" in action. Only once you know why it is wrong do you pick the response, and the response depends on the blocker:

If it broke at...Your move (don't just re-explain)
🔍 DecodeRe-read together, hunt the question, strip the distractor. "What is it actually asking?"
🗂️ Connect"Cracked one shaped like this before? Flip your journal." Show a solved parallel case.
🧰 StrategyGive a worked example to study, or a half-done one to finish. Then "pick your gadget, and say why."
🚪 MonitorReplay with a knock at each step. "Did you answer the whole question, or stop early?"
🧮 FundamentalsStep out of the word problem, drill just that skill, then come back. Drop the chilli until it is solid.
👃 EvaluateOnly the sniff test, nothing else. "Is that reasonable? Right size, right units?"

Two tools that work for any blocker

The cousin question (enabling prompt). Hand her the same kind of case with friendlier numbers, never the method. She cracks the cousin, then the original looks easy. This is your go-to "another practice question", done so it builds the pattern instead of just repeating the wall.

The funny worked example (find-the-mistake). Show the goblin solving it and making the common slip, and have her catch it. Studying a wrong example builds the same radar that prevents it, and the humour takes the sting out of being wrong (which matters most for the anxious moments).

🚪 When you are not in the room

Between checkpoints she will hit wrong answers on her own (at home, on a light day). The kid-facing Dead Ends Are Clues page carries this same flow self-serve: find the slip, pick what kind of wrong it was, try a cousin, plus Spot the Goblin's Mistake puzzles. Point her there so a wrong answer never strands her.
Start here

🔴 Run the Week 1 diagnostic

Your running notes, the blocker map, the observation sheet, and the verified answer key.

Open →

📄 Kelcy's diagnostic sheet

The clean question sheet to print or send to GoodNotes. No hints, lots of space.

Open →

🟠 The Week 5 check-in

Parallel problems and a compare-to-Week-1 panel, ready for when you get there.

Open →

📓 The weekly journal page

The write-on GoodNotes page she uses for every case in between checkpoints.

Open →