Concepts for students, parents and tutors
Start from what a person needs to do next
The current app organises around features: assignments, quizzes, reminders, marketplace. These sketches organise around moments: the lesson tomorrow, the note a parent reads, the tutor a parent is choosing. Open each on your phone and on a laptop.
A. Before the lesson
For Ava, Year 7
Home is the next lesson and what to do before it. Late homework starts with "tell Maya", not a warning. The AI helper asks questions, is labelled as a program, and a person is the fallback when it is off.
B. The weekly letter
For Jordan (parent) and Maya (tutor)
A parent reads one short note: what we did, what went well, one thing to try. The tutor writes it beside their private notes. AI wording is optional, labelled and never sees the private side.
C. Meet before you book
For a parent choosing a tutor
Start from the child's goal. Compare tutors by their own answer to a real teaching question. Times are shown in both time zones. First step is an intro chat, not a payment.
D. Home for every role
Student, parent, tutor and administrator. Apple-influenced
Tiles that show a live fact and offer one simple action, in three sizes, with a plain list of every existing tool underneath. The tutor home keeps the app's Guided workflows and adds "describe it in a sentence". Includes loading, empty, stale and larger-text states.
E. Create a lesson
For Maya, creating teaching content. Canva-influenced
Start from a purpose or a sentence, edit blocks on a canvas, see the 45 minutes as a timeline, then preview as the student before sharing. Every block says who sees it, so the answer key stays with the tutor. AI suggestions appear as a change to accept, edit or reject.
F. One home for Grades 1 to 12
Students. Fun but useful, icon-first, animated
One structure for every age: a hero for what is next, a trail of four steps, and big picture tiles with one or two words each. A switch tunes it for Grades 1 to 3, 4 to 8 and 9 to 12. Motion has an off switch.
G. Primary school version
Noah, about 7. Kid-friendly
Huge targets, read-aloud, a friendly bookmark character, tap-to-colour puzzles, a sticker book, and help as three choices instead of free typing.
H. Secondary school version
Ava, older student. Serious
No mascot. Deadlines with honest wording, a weekly checklist, a worked-practice area that checks reasoning, a week planner, a pausable focus timer, and an AI helper the student controls.
I. Vibrant palettes for children
Colour only. Playground, Candy and Arcade beside the current soft palette
The same student home repainted in three bolder palettes with the same five colour roles, a black and white check, and an honest note on what the evidence does and does not say about children and colour.
J. Touch-first for iPad
Any role on a tablet. Bigger targets, adaptive layout
The student home sized for fingers: 56 px controls, a button for every swipe, a sidebar in landscape and a tab bar in portrait or Split View. Every page also has a "Touch size" switch in the top bar that makes its own controls bigger, on by default on touch screens.
How they fit together
These are ten surfaces for different people, one direction rather than ten rivals. Spacing now follows the audience: parents get the most air for clarity, tutors get more on one page without cramping, and students get expressive layouts. Each says plainly what is not built.
Students from Grade 1 to 12: one structure, three tunings
Sources: Nielsen Norman Group's children's UX study (125 children aged 3 to 12) and physical development guidance, and Apple's motion guidance. I could not read NN/g's teen article, so H rests on thinner evidence.
| Finding | What I did with it |
|---|---|
| You cannot design for "children aged 3 to 12" as one group. At minimum, separate young, mid-range and older children. Children react badly to designs for a younger age. | F is a shared structure with grade bands, and G and H are separate tunings. H drops the mascot and the colour accents, so older students are not given something that feels "for babies". |
| Young children need big targets (about 2 cm, four times the adult recommendation) and simple taps. Dragging is hard under 9, and designs that allow tap or drag work best. | G uses tap-to-colour only, with very large buttons. A real build should also allow drag where it helps. |
| Pre-readers rely on pictures and real-life metaphors; young children skip long instructions. Children like animation and sound; adults usually do not. | Icons first with one or two words (F, G), a read-aloud button, and animation that is short and tied to an action. H keeps motion subtle. |
| Apple: add motion purposefully, make it optional, keep feedback brief, and do not make people wait for animations. | Every page has a visible Motion switch. If the device asks for reduced motion, nothing animates until the student turns it on. I tested this: zero animated elements with reduced motion set. |
| Children are wary of giving out personal information; they cannot tell advertising from content. | No adverts, no free-text AI chat in G, and a "for grown-ups" link kept small at the foot, as NN/g suggests for service content. |
Trade-off I chose and you should decide on: stars only ever go up and nothing is lost on a missed day. Streaks and countdowns would feel more exciting but can punish absences, which is a risk for anxious students. I have not tested this with children.
What the running app already does well, and where it shows up here
From reviewing the local app as a student, parent and tutor, its stylesheet and its own usability notes (docs/features/usability-improvements.md). These are things to keep, not replace.
| In the app today | Where it shows up in these concepts |
|---|---|
| A recognisable identity. A tilted bookmark mark, the "my learning" wordmark, a purple (#5650b4) with a deep plum, and a caption that names the role: Tutor workspace, Student portal, Parent portal. | All five concepts use the same palette and top bar. In D and B the caption and account name change with the role you pick. |
| One launcher for every role. The sidebar was removed on purpose; an "All apps" button returns home and "More tools" sits below. | D keeps the "All tools" list and the All apps button, so nothing is harder to reach. Tiles are added above it, not instead of it. |
| Guided workflows. A lilac strip with five named workflows and the label "CBSE · Class 8 · Mathematics". | D's tutor home reuses the exact five names and the label. E's first step is the same flow. |
| A three-step wizard. Set up, Review & edit, Save, with Preview and Edit tabs. | E's steps are named after them, with one addition: a lesson-flow step for the 45-minute timeline. |
| Honest provenance. A "Your starting point" panel: "No AI generation is needed", "not an official syllabus", check your textbook. | E keeps this panel, and the sample draft works with AI off. |
| Care already built in. 44 px targets, a skip link, drafts that warn before being discarded, saved versions you can restore, AI availability shown beside the action that needs it. | Kept: 44 px controls, skip links, a visible "AI is off" state with a human fallback, undo and version history in E. |
| Clear role boundaries. Private tutor notes never reach portals; administrators moderate public content only. | B keeps private and parent-visible text separate. D's admin home shows no student data. |
Where these concepts differ, and what it would cost
| Difference | Why | Cost to build |
|---|---|---|
| Home tiles show a live fact, not an icon and a description. | Apple's widget guidance prefers changing information to a copy of the app icon. | Each tile needs a count or status from the API (late homework, next lesson). Today's launcher uses no live data. Tiles with nothing to say should stay as list rows. |
| Content is split into blocks, each marked "Only you" or "Students see this". | Today the answer key sits in the same text as the lesson, with a note to remove it. | A library item is a single text field (up to 8,000 characters). Blocks would change that model. A cheaper first step is a marked private section inside the existing text, shown to students without it. |
| A student preview inside the editor. | The quiz studio already has one; lessons and assignments do not. | Reuse the quiz preview pattern. Low. |
| Home is organised by the next action, with the tool list below. | Matches the evidence on linking tutoring to lessons. | Mostly front-end once the counts exist. The tool list stays as the fallback. |
The running app is also stronger than these sketches in ways I cannot match: it is real, and its authors ran a 31-item usability pass, browser tests in five profiles and accessibility checks. My sketches have not been tested at all.
What the evidence says, and what we did with it
I read these pages directly. They are mostly UK school studies, not Australian online tutoring, so treat them as design guidance, not proof.
| Finding | Where it shows up | Source and strength |
|---|---|---|
| One-to-one tuition is more likely to help when it is additional to and explicitly linked with normal lessons. | A and B organise around the next lesson and the last lesson's note, not standalone tasks. | EEF, one to one tuition. About +5 months on average, moderate evidence. Lower in maths (+2 months), and commercial evaluations tend to show larger effects. |
| Short, regular sessions appear to work best: about 30 minutes, three to five times a week, up to ten weeks. | C's profile suggests a first-month pattern and lets the parent ask for one. | Same EEF page. A pattern seen across studies, not a rule for every child. |
| Effective feedback is specific, on the task and on how the learner manages their learning, and it matters when work is correct too. | A shows what Ava did and the next step, with no scores. B names something specific the child did. | EEF, feedback. About +6 months, high evidence. Feedback can also backfire. |
| Feedback from technology still helps but is slightly smaller (about +4 months) than from teachers. | The AI helper in A is a fallback and a "show Maya" route, not the main teacher. | Same EEF feedback page. |
| Parental engagement is linked with better outcomes (about +3 months), but evidence on what works for schools to do is mixed, and work directly with children has stronger evidence. | B asks parents for one small thing to try, not to supervise. This is a design bet to test, not a finding. | EEF, working with parents. Guidance report. |
| Products position their AI tutor as not giving the answer and parents as having a window into learning with messages. | A's helper asks a question first. B is a message-based window. | Khanmigo and Seesaw. These are the vendors' own marketing claims, not independent evidence. |
Not covered: I could not read Preply, Superprof or ClassDojo (blocked or redirected), so no claims are made about them. No interviews or usability testing were done. Everything about what real parents and students want is still an assumption.
Alternatives I considered for the student home
- A skill map showing every topic. Clear for tutors, but it asks a 12-year-old to read a curriculum.
- A chat-first home where the AI plans the day. Easy to build, but it hides the tutor and invites over-reliance.
- The next lesson as the organising idea (chosen for A) fits the evidence on linking tutoring to lessons, and it can be tested quickly.
Self-review against generic AI-design patterns
- Avoided: gradients, glass, nested cards, KPI strips, streak and score gamification, emoji, bento grids. A to C use ruled sections and plain type; D uses tiles, which must carry live content to justify themselves.
- Distinct choice: the app's own purple and bookmark mark, margin notes for teacher voice, and real fraction bars drawn from the maths, not decoration.
- Colour is never the only signal. Status always has a word ("Not handed in"). I calculated every text and background pair in the app's palette: the lowest is 4.67:1 (WCAG AA needs 4.5:1).
- Remaining risk: it may feel too plain or too much like a document for some learners. That is a testable question, not a settled one.
- Reuse test: swap "fractions" for "deals" and the student view would not make sense. The parent letter and fraction bars are specific to teaching.
Home tiles and creation: what I took from Apple and Canva
Sources: Apple's widgets, typography and accessibility guidance, and Canva's education and docs pages. Canva's pages are marketing, so they show what it claims to do, not how well it works. Apple's are design guidance.
| Principle | How it appears |
|---|---|
| Apple: replicating an app icon "offers little additional value"; prefer dynamic information. | D's tiles carry a live fact ("2 not handed in", "Tomorrow, 10:30 am") instead of an icon and a label. |
| Apple: small sizes show one thing; larger sizes add detail. Keep interactions simple and deep-link to the right place. | Three tile sizes. Inline actions are limited to a button, a link or a checkbox; each tile's title opens its detail. |
| Apple: use the system font, avoid light weights, meet 44 pt targets, don't rely on colour alone, support larger text. | System font stack, weights 400 and up, 44 px controls, status always in words, and a larger-text tile state. |
| Apple: support light and dark, with clear placeholders while data loads. | D follows the system theme and shows placeholder, empty and could-not-refresh states. |
| Canva: start from a template by purpose or a prompt, then edit with comments and suggestions in one shared space. | E starts from five purposes or a sentence. AI changes show as a suggestion to accept, edit or reject. |
| This app today: five guided workflows produce one block of text with the answer key inside it, "to be removed from the learner copy". | E splits the document into blocks and marks each "Only you" or "Students see this", with a student preview before sharing. |
What to ask real people
- Show a Year 7 student A with two late items. What do they do first, and how do they feel?
- Give a parent B. Ask them, without scrolling back, what they would do this week.
- Give a parent C. Which tutor would they request, and why? Do they trust "None yet"?
- Watch Maya write a note in B. Does it take less time than today?
- Show D to each role with no instructions. Ask what they would do first, and time it.
- Ask a tutor to build a lesson in E and in today's Content creator. Compare time, and ask whether they would trust the student preview.
- Ask a 6 to 8 year old to get from G's home to a puzzle with no instructions. Watch where they hesitate, and whether the read-aloud is used.
- Show F's Grades 9 to 12 setting and H to teenagers. Ask which feels like "theirs" and why. Ask whether the checklist feels helpful or like surveillance.
- The app's sample content covers Years 5 to 10 Mathematics. Grades 1 to 4 and 11 to 12 would need new content and curriculum. Children under 13 also need a privacy and consent review before any of this is built; I have not done one.
- The mascot is the app's bookmark mark with a face, drawn by me. It is an idea to test, not an approved character.
- Reading aloud uses the browser's own voice, so quality varies by device. Animations were tested for running, stopping and reduced motion, not for performance on low-end devices.
- Student colour pairs (text on tiles and the warm primary theme) were calculated: the lowest text pair is 5.29:1 and the icons on tinted discs are at least 5.09:1 (AA needs 4.5:1 for text, 3:1 for icons).
- Student colour roles are fixed: indigo is the brand, plum is the deeper tone for Grades 9 to 12, sun yellow means your next move or a reward, mint means done, sky means people or information, lilac means the AI helper and coral means gentle attention (children never see red). Each screen has one saturated tile, the thing to do now. Sun yellow does not appear in the running app or its logo, so it is a proposal. Mint and coral become near-grey and khaki for red-green colour-blind viewers, so every state also has an icon or a word. The sticker colours in G are collectibles and are the one exception to the roles. Nothing was tested with children.
- Static HTML with synthetic data. Tabs work; saving, sending and AI do not. In E, selecting a block updates the inspector.
- D and E now use the running app's colours, font, top bar and button shape. What they add is Apple-style tile sizing and a Canva-style editor layout. The system font only looks like Apple's on Apple devices, and no Apple or Canva assets are used.
- The app has no dark theme today. D's dark theme is a proposal using the same hues; it is contrast checked but not an existing feature.
- The first round of concepts used an invented paper-and-serif look and the wrong tutor name for Ava (she is taught by Maya Chen, not Alex). Both are fixed. This was a miss on my part: I had not yet reviewed the running app's tutor side.
- The earlier self-review said "no icon tiles". D uses tiles because you asked for them, so they must earn their space with live content. If a tile has nothing live to say, it should not be a tile.
- Colour pairs in D were calculated in light and dark; the lowest is 4.67:1 (AA needs 4.5:1). Larger-text behaviour is shown for one tile only, not tested at all system sizes.
- Time zone conversions and dates are hand-written examples (the sample date is Tuesday 20 October 2026).
- Several features (intro chat, reply-time promise, cancellation terms, "show Maya this question") are proposals that would need product and policy decisions first.
- Accessibility was designed in (skip link, tab keys, labels, focus rings, 44 px targets) but not audited with assistive technology.