These checks tell the teacher what to model next; they are not a Year 3 achievement-standard decision. Keep exact response or model | access/prompt | mathematical idea shown | next move for a rotating sample. The whole class attempts each exit; observe a planned few closely. Arrange a quiet individual check when independent reasoning is needed and write not yet observed otherwise.
| Day | Keyed check | If the response suggests a gap |
|---|---|---|
| 1 | 1 ten-thousand + 3 thousands + 0 hundreds + 2 tens + 5 ones = 13 025 |
Rebuild 13 025 beside 1 325; point to the absent hundreds. |
| 2 | 30 506 has 0 thousands and 0 tens | Say every place while a learner positions each tile. |
| 3 | 12 405 > 12 045 because hundreds differ first | Align places; compare from left and stop at first difference. |
| 4 | 18 counters is about 20, more than 10; exact count after estimate | Provide a visible ten benchmark, then count in groups. |
| 5A | 14 305 < 14 350 because tens 0 < 5 | Keep zero tens visible and compare both on a mat. |
| 5B | 38-counter tray: estimate first, then group 3 tens/8 ones | Do not grade closeness alone; practise interval and revision. |
| 6 | 184 + 137 = 321, with an exchange from 4+7 ones | Use ten single pieces exchanged for one ten, then recompute. |
| 7 | 302 − 157 = 145, with 145+157=302 as check |
Count up 157→200→300→302 if exchange across zero is hard. |
| 8 | □ + 137 = 400 → 263; 400 is whole |
Mark whole/parts before choosing subtraction. |
| 9 | Book problem 513 books; tile problem 325 tiles | Re-enact which group arrives/leaves, then check unit. |
| 10 | 14 370 > 14 307; 346+178=524 tickets | Diagnose place comparison and exchange independently. |
Day 5 check A: place and estimate
New place prompt: Build 14 305 and 14 350, say each numeral, compare, then explain which digit settles it. Key: both have 1 ten-thousand, 4 thousands, 3 hundreds; 0 tens < 5 tens, so 14 350 is larger. A child who says 14 305 is larger “because 5 is greater than 0” is comparing ones before tens; ask them to point to each place, not rehearse a symbol. A child who reads 14 305 as 1 435 has dropped a place; rebuild in all five labelled columns.
Estimate prompt: show a new tray containing 38 large paper counters for a brief look. Retain the estimate before opening. Group count 3 tens and 8 singles. Strong evidence includes a plausible interval (e.g. between 30 and 50), a reason and willingness to revise. Do not mark an estimate of 35 “incorrect” just because the exact count is 38. If the initial estimate is 300, compare with a visible ten; if the counting result is 37, check one-to-one organisation without treating estimation itself as weak.
Recheck later: use a new numeral, e.g. 16 402 versus 16 420, and a different-size collection with a ten benchmark. Do not repeat only the answer the child has now seen.
Day 10 check B: place plus additive model
Place: 14 307 has 1 ten-thousand, 4 thousands, 3 hundreds, 0 tens, 7 ones. 14 370 is greater: hundreds tie; 7 tens exceed 0 tens. Accept a well-explained equivalent expanded representation.
Problem: “There are 346 tickets in one box and 178 in another. How many tickets altogether?” 524 tickets. A reasonable estimate is near 350 + 180 = 530. A place-value path: 300+100=400, 40+70=110, 6+8=14, total 524. Check 524−178=346 or another correct inverse.
Example feedback decisions:
- Child gets 524 with a model but omits “tickets”: ask what is being counted, keep the correct calculation.
- Child writes 414 after correctly summing hundreds/ones but dropping a ten exchange: exchange 10 tens for a hundred with cards, then try 257+164.
- Child adds 346−178 even though boxes combine: ask them to mark the two parts and draw a whole; use 24+13 before returning to three digits.
- Child produces 524 by copying a neighbour but cannot explain any part: mark independent strategy not yet observed and use a private new-number check.
Reading support, translated oral explanation from an appropriate speaker, AAC or a scribe must be recorded, without confusing them with a mathematical prompt. Respond to the child's actual model; do not assign an ability label from a single fortnight.
Rights: Original checks and examples © NeuroForgeIO Pty Ltd 2026, SubjectNest, CC BY 4.0. Credit, link and note changes.