Adaptive
DefaultThe main mode. A balanced mix of statutory words, any word lists the teacher or parent has assigned to that child, and this week's spellings if the class has them.
Statutory Assigned lists Weekly words
Under the hood
How Spello actually works
A plain-English tour of how we pick words, find each child's level, and adapt as they play. Written with teachers in mind, but anyone's welcome.
The short version
Two-minute read- 1 A short assessment, on first use, places each child at a starting working-at year — not necessarily their school year.
- 2 Every word shown is picked from a balanced mix of words they've seen before (for review) and new words — tuned to the right difficulty for that child, right now.
- 3 As they play, their working-at year drifts up or down based on real performance — nudging harder words in when they're ready, easing off when they're not.
That's the short version. The unreasonably detailed one keeps scrolling — we got a bit carried away. Tea recommended.
1 · Practice modes
Four practice modes
The practice mode determines which words will be used during the game. This is used in combination with the five game types.
Tricky words
Statutory common-exception words only. Still filtered to the child's working-at level so a Year 1 child doesn't get Year 6 "conscience".
Statutory only Level-aware
Weekly spellings
Just this week's words from the teacher, nothing else. Shown to the child only if their class has a list for the current week.
Teacher-set Class only
Specific list
The child picks one topic or spelling-rule list — "Ancient Egypt", "Magic e", "Silent letters" — and plays only words from that list.
One list Child chooses
2 · The assessment
Finding the right starting point
On first login, most children sit a short placement assessment. It opens a few years below their age with a comfortable warm-up, climbs toward their age year if things feel easy, and probes more deeply once it reaches a tier that matters. There are no stars, no right/wrong feedback, no scores shown — just a handful of words to spell until we've quietly found the right year to place them in.
Year 1 children skip the assessment
There's nothing meaningful to measure below Year 1, so new Year 1 children go straight to Year 1 play — no placement session, no waiting, just start spelling. Parents and teachers can still request an assessment manually at any time if they'd like one.
1
How the session unfolds
Three phases — warm-up, probe, and an optional stretch probe at the top.
3words Warm-up
Feel out the floor. Starts three years below their age, three words at a time. 2 of 3 or better unlocks the next rung up toward their age year; below that flips the session into a gentle descent, one rung at a time, looking for comfort.
10words Probe
Zero in. Ten words at the year the warm-up lands on. 8 or more opens a stretch probe one year up; 5–7 settles here; under 5 drops a year and probes again. Year 1 is the floor.
10words Stretch
Only after an ace. Ten words at the year above. 7 or more places them at the higher tier; otherwise they settle at the year where they aced.
★
Four example journeys
What actually happens for different children on their first session.
Ellie
Year 4 · reads well, spells confidently up to a point.
- Y1warm-up3/3
- Y2warm-up3/3
- Y3warm-up2/3
- Y4probe6/10
Climbs cleanly through warm-up, then passes the probe at her age year. Placed at Year 4.
Noah
Year 5 · confident speller, wide vocabulary.
- Y2warm-up3/3
- Y3warm-up3/3
- Y4warm-up3/3
- Y5probe9/10
- Y6stretch7/10
Aces warm-up through Y4, aces the Y5 probe, and clears the Y6 stretch too. Placed at Year 6 — above his age.
Mia
Year 3 · aced the easier years but Year 3 words didn't click.
- Y1warm-up3/3
- Y2probe9/10
- Y3probe4/10
Y3 probe fails, so the engine drops back and extends Y2 from warm-up into a full probe. She aces it. Placed at Year 2.
Finley
Year 5 · finds spelling tough, hasn't caught up with peers yet.
- Y2warm-up1/3
- Y1probe6/10
Y2 warm-up falls short, so the session flips direction and extends Y1 from warm-up into a full probe. Placed at Year 1. The adaptation layer eases him in from there.
2
What the child actually sees
Neutral by design. We're measuring, not teaching — and definitely not testing.
No right / wrong feedback
Answers look the same whether correct or not — no ticks, no crosses, no colour change, no celebration. We don't want children tensing up on a hard word or second-guessing ones they actually know.
Doesn't feed the spacing engine
Assessment attempts are for placement only — they don't affect mastery tracking or spaced repetition. Real play starts with a clean slate at the decided year.
Short and hard-capped
Most children settle in 20–30 words — a warm-up climb of a few short rungs and one ten-word probe. The absolute cap is 60, enough to cover a full warm-up + probe + stretch and a descent-and-reprobe for the trickiest placements.
3 · Word selection
Choosing the best word
Each mode decides the pool of words available; then the engine picks the next one from that pool and chooses one of the five game types to present it with. Here's how each of the four modes scopes its pool, starting with the main one.
Adaptive — the main mode
Every word choice is two steps. First, Spello decides which words are eligible for this session. Then for each question it picks the single best word from that pool. Here's what happens inside each step.
1
Build the session pool
Decide the budget by source, then the year split inside statutory.
1a · The source budget
A 20-question session is budgeted across up to three sources. The split depends on what's actually been put in front of this child.
No extras
All 20 come from the statutory curriculum.
Assigned lists only
Half and half. Every assigned word is eligible; 10 are used this session.
Weekly words only
Half and half. If the class has fewer than 10 weekly words, the shortfall moves back into statutory.
Both assigned and weekly
40 / 35 / 25. Weekly gets the smaller share — it's usually a shorter list the class is drilling this week; assigned sits in the middle as the ongoing one.
Not enough words for a source? If the class only has three weekly words ready this week, the weekly budget caps at three and the shortfall is redistributed to the other sources in proportion to their own share — so the session still runs the full 20.
1b · The year split inside statutory
Whatever size the statutory slice ended up — 20, 10, or 8 — it's then split across three year tiers around the child's working-at year. Only statutory words are year-filtered; weekly and assigned stay as the adults picked them.
- Current year (W)
- 60% — the bulk of practice happens where the child actually is.
- One below (W−1)
- 25% — keeps earlier words fresh and fills gaps.
- One above (W+1)
- 15% — a small stretch. How the system notices they're ready to move up.
At Year 1 there's no "below", so that 25% rolls into current-year practice. At Year 6, the "above" does the same.
2
Pick the next word
Four lanes, tried in order. The first with a word to offer wins.
The lanes don't care which source a word came from — a weekly word and a statutory word compete on the same terms once they're in the pool.
- 1
Overdue review
A word they've seen before that's due for a spaced-repetition check-in. Gets first dibs — spacing only works if reviews happen on time.
- 2
Still learning
A word they've got wrong recently and the system isn't yet sure they know. Keeps them on it.
- 3
New word
Brand new to them, ordered so easier words appear first — a confidence foothold before harder ones arrive.
- 4
Fallback
Rare. A random in-scope word if the first three lanes are all empty.
Adapting to struggle
The engine watches rolling accuracy over the last five meaningful answers and runs the session in one of three gears. The 60/25/15 distribution above is only the first of them.
Baseline
60 / 25 / 15
The default mix. Current year does the bulk, a quarter stays below to keep earlier words fresh, a sliver above for stretch.
Struggle
30 / 70 / 0
Triggers when rolling accuracy falls below 40%. Above-year stretch pauses entirely; prior-year reinforcement takes over.
Deep struggle
10 / 90 / 0
Triggers below 20%. Almost all prior-year reinforcement until something clicks again.
Hysteresis, on purpose. The exit threshold sits deliberately above the entry one — to leave struggle, rolling accuracy has to clear 60%, not just 40%. That stops a single right answer after a rough patch from yanking the child back to stretch words before the wobble is really over. Deep struggle drops to struggle at 40% or skips straight back to baseline at 60%; baseline can also drop directly into deep struggle if accuracy collapses past 20%.
Carried across sessions. The rolling window is seeded from the tail of the child's last session. A child who ended yesterday in struggle mode starts today in the same mode — it relaxes naturally as fresh correct answers replace old ones, so there's no jarring reset overnight and no false confidence from a blank slate.
Weekly and assigned words keep flowing regardless — the teacher and parent decided those were important.
★
A worked example
Sam, Year 3 — class has weekly words, parent has assigned "Silent letters".
Sam hits the "both" scenario — 8 statutory · 7 assigned · 5 weekly. Inside the eight statutory slots, 60/25/15 lands on five Year 3, two Year 2, one Year 4.
Y3 wk as Y3 wk Y2 as wk Y3 as Y4 wk Y3 as Y2 wk Y3 as as as
- Statutory — current year ×5
- Statutory — below ×2
- Statutory — above ×1
- Assigned lists ×7
- Weekly words ×5
With a different setup, the tiles shift: assigned-only → 10 statutory · 10 assigned. Nothing set → 20 statutory, with 60/25/15 applied to all of them.
The other three modes
Each still uses the same four lanes for which word to pick next — the difference is the pool they're choosing from.
Tricky words
Statutory common-exception words only — but still filtered around the working-at year, using the same 60/25/15 split.
Weekly words and assigned lists aren't included. If a Year 3 child picks Tricky words, they won't see "conscience" — that's Year 6.
Weekly spellings
Just this week's words from the teacher — nothing else. No year-group filtering, no statutory mix.
If the list has fewer than 20 ready words, the session is shorter. Hidden from the child entirely if their class hasn't set any words this week.
Specific list
Exactly the list the child picked — "Ancient Egypt", "Magic e", "Silent letters" — and nothing else.
Useful if a child (or their teacher) wants a focused drill on one thing. No adaptive tier mixing, no year filter.
4 · Spacing
Spacing it out
Every word has its own little status. The system remembers when it last saw a word, how the child did, and when it should come back. This is where spelling actually sticks.
A little theory
Why spacing beats cramming
In 1885, the German psychologist Hermann Ebbinghaus ran a now-classic experiment on himself, memorising nonsense syllables and testing how quickly he forgot them. He found memory drops off sharply at first — most of what you learn slips away within a day — then tails off more gently after that. It's called the forgetting curve.
The clever part: each time you review something just before you'd otherwise forget it, the curve resets higher and decays more slowly. A handful of well-timed reviews is more effective than a long cramming session — and it's why revising everything the night before a test barely sticks.
Spello uses this deliberately. Every word has its own forgetting curve, and we schedule it to come back just as the child is on the verge of forgetting — so each review counts.
Each yellow spike is a well-timed review. Each one resets retention higher and the curve flattens.
Testing effect
Recalling a word — not just re-reading it — is what strengthens the memory. Every question in Spello is an active recall.
Spacing effect
The same amount of practice is much more effective when spread across days than packed into one sitting.
Desirable difficulty
A word that's just hard enough to recall is the sweet spot. Too easy and it's wasted; too hard and it's discouraging.
Learning
Still getting wrong sometimes. Comes back soon.
Reviewing
Got it right a few times. Comes back at wider intervals.
Mastered
Consistently correct. Only checks in occasionally to stay fresh.
Get it right
The interval before we show it again grows — first a day, then three, then a week, then longer.
Get it wrong
The interval resets to zero. It'll come back very soon. The system also notes it's harder than it thought.
▸ The technical bit: it's a modified SM-2 algorithm
SM-2 is a well-studied spaced-repetition algorithm used by flashcard apps. Each word has an "ease factor" — a number that grows when answers are easy and quick, and shrinks when they're slow or wrong. Spello tweaks the standard SM-2 in a few ways: confident answers on words well below the child's level promote faster, hesitation (backspaces before submitting) caps the reward, and we never let intervals grow unboundedly for words they might genuinely not know.
5 · Levelling
Moving up and down
The initial assessment gives a starting point — real play then refines it. The working-at year drifts up or down based on the last three sessions of evidence.
Moves up when all three are true
- Getting at least 85% right at the current year.
- Has had at least 20 tries at this year across their last three sessions.
- At least half of the current-year words they've seen are now "reviewing" or "mastered".
Drops down when both are true
- Getting under 40% right at the current year across the last three sessions.
- At least 15 tries at the current year in that same window — enough evidence to trust the dip isn't just a bad day.
Parents can also request a fresh assessment from the dashboard at any time — useful after a long break or if something's clearly wrong.
Example over a term
Y5 Y4 Y3 Y2 Y1
- Initial assessment: Year 3
- Strong run — promoted to Year 4
- Rough patch — dipped back to Year 3 briefly
- Recovered and kept climbing
6 · You in the loop
Teacher-set content
Spello's default scope is the statutory curriculum, but you can add your own words into the mix.
This week's words
Type or paste a list into your class's weekly words screen. If a word's new to Spello, we automatically generate example sentences and audio for it in the background — usually within a minute or two.
- Appears in the child's adaptive mode as part of the mix.
- Also available as its own "Weekly spellings" mode, if they want to drill just those.
- Wipe and replace each week — previous weeks' words keep their progress.
Assigned wordlists
Pick specific wordlists — topics or spelling-rule sets like "Silent letters" or "i before e" — for an individual child. Useful for targeting a weak area without disrupting everything else.
- Assigned wordlists feed into the adaptive mode's mix.
- The child can also pick any of their assigned lists directly.
- Parents with a family account can assign lists the same way.
7 · Deliberately not
What we deliberately don't do
A few choices worth flagging so no-one's surprised.
Tell them right/wrong in the assessment
The first-use assessment is placement, not teaching — feedback during it would turn it into a test.
Punish a wrong answer
A wrong answer doesn't cost coins and doesn't drop their level. It simply comes back again soon — the system just noted it needs more practice.
Time them out or rush them
There's no timer on the answer. Speed earns a bonus but isn't required. Slow thinkers aren't penalised.
Give them a grade
Children never see a percentage, a pass/fail, or a mark out of twenty. They see coins, streaks, and progress toward the next collectable — never a grade.
8 · Under the hood
For the curious
The actual numbers, if you want them. These are tunable — if real usage shows we need to adjust, we will.
- 20 questions per normal session — roughly ten minutes. The first-use assessment can run up to 60 questions so the algorithm can fully probe every year group for very confident children; most settle long before the cap.
- Starts three years below the child's age year (floored at Year 1). Warm-up: three words per rung; 2/3 unlocks the next rung up toward their age year, under 2/3 flips direction and descends looking for comfort. Probe: ten words at the landed year. ≥8/10 opens a stretch probe one year up; 5–7/10 settles here; under 5 drops a year and re-probes. Stretch probe: ten words one year up; ≥7/10 places them at the higher tier. Floors at Year 1. Children whose age places them at Year 1 skip the assessment entirely; parents and teachers can still request one manually.
- Baseline 60/25/15 (current / below / above) per session, redistributed when the child is at Year 1 or Year 6. Shifts to 30/70/0 if rolling accuracy drops below 40%, and 10/90/0 below 20% — above-year words pause in both struggle modes. Returns to baseline above 60%.
- All of: ≥85% accuracy at current year across the last three sessions; ≥20 current-year attempts in that window; ≥50% of seen current-year words at "reviewing" or "mastered".
- <40% accuracy at current year across ≥15 attempts in the last three sessions.
- 5 base coins per correct answer, plus a +1 accuracy bonus when the word is typed cleanly with no corrections, plus a +2 speed bonus when a listen-and-type or flash-card answer is submitted in under 4 seconds. Wrong answers and giving up earn nothing — and never deduct. The initial assessment pays a flat 10 coins on completion — the same small reward for every child, so the placement never feels like a score to chase.
Session length
Assessment rules
Distribution
Moving up
Moving down
Coin rules
Questions we haven't answered here? — we're happy to explain.