# Open Coding Analysis — April 2026

**Purpose**: inductive, bottom-up coding of the combined corpus (93 existing verified quotes + ~40 new quotes from April 2026 expansion + 9 theoretical frameworks). No pre-existing archetypes. No pre-decided themes. Let structure emerge from the material.

**Method**: I read every quote across `trustpilot-parent-quotes.md`, `trustpilot-regional-parent-quotes.md`, `alt-platform-parent-quotes.md`, `russian-platform-parent-quotes.md`, and the April-2026 expansion. For each, I tagged underlying phenomena (not topics). Then I clustered codes into themes. Then I asked: *what one or two axes could structure all of this?*

**Output status**: DRAFT — for discussion with user before proceeding. This file deliberately surfaces multiple candidate narrative axes instead of pre-committing to one.

---

## PART 1 — OPEN CODES (inductive tags)

Codes are descriptive labels for phenomena in the quotes. Many quotes carry multiple codes. Ordering here is by *how often the code appeared across the corpus*.

### 1.1 Codes about **what parents count as evidence of progress**

| Code | Typical signal | Frequency |
|---|---|---|
| `proof:real-world-moment` | *"Ordered food in English," "spoke to cousin," "read billboard," "Yuki said 'sorry my dad is busy'"* | Very high — across all markets |
| `proof:school-grade` | *"No problems in school English," "best in class," "best in subject"* | Very high — strongest in RU, IT, FR |
| `proof:external-validation` | *"Participates in Olympiads," "accepted to international school," "makes American friends"* | High — strongest in RU, DE, TR |
| `proof:child-desire` | *"Happy and wants to continue," "begs for more lessons," "chose to return"* | High — across all markets |
| `proof:biliteracy` | *"Reads and writes freely in two languages"* | High in RU, PL |
| `proof:autonomy-marker` | *"Almost independent," "self-manages schedule"* | Medium — RU strongest |
| `proof:multi-axis-naming` | *"Conversation, grammar, reading" — parent lists skill dimensions* | Medium |
| `proof:pronunciation` | *"Speaks without accent," "native-speaker confidence"* | Medium |
| `proof:sentence-construction` | *"Forms simple sentences," "constructs sentences"* | Medium |
| `proof:comprehension-over-production` | *"Understands a great deal," "understands English on vacation"* | Low but significant |
| `proof:before-after-self` | *"Started at zero, now fluent"* | Low |

**Observation**: *Not a single quote in the combined corpus frames progress in platform-visible terms* (lessons taken, CEFR level reached is mentioned but almost always coupled with school grade, skill tags, time-spent numbers). The entire evidence system parents actually use is **outside the dashboard**.

### 1.2 Codes about **what breaks parent trust**

| Code | Example quote | Why it breaks trust |
|---|---|---|
| `trust-break:positive-only-feedback` | *"Notifications are always positive. Even if lesson went badly, everything is rose-coloured"* (realjaew, TR) | Parent can't distinguish good lessons from bad → no feedback is trustworthy |
| `trust-break:opaque-billing` | *"They secretly enable auto-payment and illegally charge money"* (Andrey, RU); *"Paused but billing continued 4 months"* (Rafi, US) | Destroys educational credit retroactively |
| `trust-break:level-stuck` | *"Without my intervention, advancement wouldn't have occurred"* (Nejat, TR, paid 12,553 TL for 50 lessons at same level) | Product promise of personalized progression contradicted |
| `trust-break:teacher-change-unilateral` | *"My son cried so much... trauma in our lives"* (Ozge, TR) | Disrupts the primary retention bond without parent input |
| `trust-break:preferred-teacher-unavailable` | *"No free time slots after 4 lessons, told to choose others"* (Ksenia, RU); *"Their schedules are always fully booked"* (Esra, TR) | Scarcity of the thing that created value |
| `trust-break:curriculum-jumps-unannounced` | *"All these phrases except 'I've got' were completely new"* (Lyudmila, RU) | Parent loses co-pilot capability |
| `trust-break:placement-wrong` | *"Spent 4 lessons on colors... beyond 'I am a girl'"* (Giulia, IT); *"5-year-old needs reading/writing she doesn't have"* (Maria Angels, ES) | Level assessment is teacher-dependent, inaccurate |
| `trust-break:refund-stonewalling` | *"I've requested a refund multiple times, no one responded. I feel a bit scammed"* (Jeanne, US) | Exit hostility destroys goodwill |
| `trust-break:platform-unreachable` | *"Platform inaccessible for days, we get no communication"* (Filiz, TR) | Basic reliability failure |
| `trust-break:progress-button-permission` | *"Clicking progress launches audio recording permission"* (Oya, TR) | The progress feature itself is broken |

**Observation**: There are **10 distinct trust-break patterns**, and most are *not* about the quality of teaching. Novakid's retention problem is largely a **trust architecture problem**, not a pedagogical one.

### 1.3 Codes about **parent behaviour around the product**

| Code | What parent does | Frequency |
|---|---|---|
| `parent-diy:external-tracking` | Builds Notion, notebook, phone recordings | High |
| `parent-diy:record-child-speaking` | Records monthly clips to compare | High (existing 45-persona synth, multiple real quotes) |
| `parent-diy:curriculum-prep` | Repasses didactic materials, explains "why" before lesson | Medium (Margarita IT) |
| `parent-diy:manual-level-advocacy` | Asks platform to advance child's level | Medium (Nejat TR) |
| `parent-diy:independent-teacher-consult` | Asks outside English teacher to verify progress | Low but strong (Nejat TR) |
| `parent-diy:school-teacher-verification` | *"Improvements acknowledged by school teachers"* (Dorotea IT) | Medium — often the moment parent feels justified |
| `parent-diy:travel-testing` | Takes child abroad to test real-world use | Medium (Ozge TR, Avotna RU) |
| `parent-on-rail:receives-but-can't-reply` | Gets feedback but has no return channel (Carmentxu ES, realjaew TR) | High |
| `parent-quits:cancel-after-plateau` | Considers cancellation in months 2–8 | High (British Council + 40/45 synth) |
| `parent-stays:child-wants-more` | Stays because child begs for lessons | Very high — across markets |

**Observation**: **Parents are not passive consumers of the dashboard — they build parallel evidence systems.** The DIY tracking pattern is so consistent it should be read as a *latent product request*, not a parent preference.

### 1.4 Codes about **the parent's emotional landscape**

| Code | Feeling | Evidence |
|---|---|---|
| `emotion:anxiety-about-investment` | *"Am I paying for air?" (Giuseppe IT synth, paralleled by real price-sensitivity quotes)* | High |
| `emotion:guilt-about-cost` | *"I seriously dislike subscription... I am fed up"* (Rafal) | High |
| `emotion:pride-at-child-independence` | *"Almost independently"* (ZULYAen RU); *"Best in class"* (Natalya RU) | Very high |
| `emotion:relief-at-external-validation` | *"Improvements acknowledged by school teachers"* (Dorotea IT) | Medium |
| `emotion:fear-of-wasted-time` | *"After 4 lessons spent on colors... beyond saying I am a girl"* (Giulia IT) | Medium |
| `emotion:betrayal-billing` | *"Scammed," "cheated," "illegally charged"* (multiple) | Medium |
| `emotion:joy-child-enthusiasm` | *"Begs for more lessons"* (multiple) | High |
| `emotion:overwhelm-parent-work` | *"I must repass didactic materials... why must I develop interest?"* (Margarita IT) | Medium |
| `emotion:connection-via-teacher` | *"Teacher allowed Rita to slowly form... Mom really feels Rita has made progress"* (Rita mum) | Medium |

**Observation**: The emotional palette is rich but clusters into **four states**: *anxious-uncertain*, *proud-validated*, *betrayed-angry*, *connected-delighted*. The Value Hub's job is partly to move parents between these states — into proud/connected, out of anxious/betrayed.

### 1.5 Codes about **what parents want but don't have**

| Code | Underlying desire | Quote evidence |
|---|---|---|
| `want:honest-feedback-not-rosy` | Truth over positivity | realjaew (TR) |
| `want:bidirectional-teacher-dialogue` | Reply channel to teacher | Carmentxu (ES) |
| `want:placement-test-at-start` | Know starting level accurately | pelma pelma (TR) |
| `want:level-auto-advance` | Don't require manual intervention | Nejat (TR) |
| `want:warning-before-difficulty-jump` | Prepare for harder material | Lyudmila (RU) |
| `want:longer-schedule-horizon` | Plan 2-3 months ahead | Kotya (app store) |
| `want:teacher-score-feature` | Rate the teacher | Regretful mother |
| `want:other-parent-opinions` | See reviews before booking | Regretful mother |
| `want:pace-benchmark` | Know if 5 words/2 weeks is normal | mavikoltukminderi (TR) |
| `want:school-curriculum-alignment` | Link Novakid to school English | Mariem (FR), Pantyusha5 (RU) |
| `want:grammar-depth-option` | Power-user mode | Елена (RU) |
| `want:parent-prep-material` | Briefing before lessons | Margarita (IT) |

**Observation**: **12 concrete feature requests** emerge from the quotes. Notable: none of them ask for "better graphs" or "more data." They ask for *honesty, dialogue, benchmarks, alignment, and agency*. The Value Hub is not a visualization problem — it's an architecture problem.

---

## PART 2 — EMERGENT THEMES (clusters of codes)

When I group the codes above by what they're *really about*, seven themes surface. I'm NOT organizing by the old schema (Universal Insights / Archetypes / Cultural Map). I'm organizing by what the material itself clusters around.

### Theme A — Evidence Lives Outside the Dashboard

Every code in §1.1 (`proof:*`) points outside the platform. Real-world moments, school grades, competitive achievements, the child's own desire. The dashboard currently provides none of these. This is not a bug — it is a category error. The dashboard measures the *platform's* activity; parents measure the *child's* life.

**Design imperative**: the Value Hub should either (a) capture life-evidence (parent-reported moments, school-teacher feedback integration) or (b) generate narrative that *connects* platform data to observable life outcomes. Probably both.

### Theme B — The Feedback System Has a Truth Distortion Layer

`trust-break:positive-only-feedback` + `parent-on-rail:receives-but-can't-reply` + `emotion:betrayal-billing` compound into one mechanism: **the parent-facing feedback stream is uniformly positive, one-way, and surrounded by billing opacity.** Parents stop trusting the stream, so they build DIY systems. The positive bias was probably designed to reduce parent anxiety; in practice it *amplifies* it because parents can't tell signal from noise.

**Design imperative**: calibrated honesty. Feedback must include both strengths and areas to reinforce, and the dashboard must surface a trust artefact (e.g., "this teacher flagged 2 things this month" — the absence of flags becomes informative only if flags exist).

### Theme C — The Curriculum Is a Black Box

`trust-break:level-stuck` + `trust-break:curriculum-jumps-unannounced` + `trust-break:placement-wrong` + `want:placement-test-at-start` + `want:warning-before-difficulty-jump` all describe the same thing: **parents cannot see the curriculum.** They don't know where their child is, what comes next, or whether difficulty will increase. The platform promises "personalized progression" but the level mechanism is opaque and often teacher-dependent.

**Design imperative**: the curriculum path is a first-class Value Hub surface — not buried in lesson history. Parents must be able to see: *where my child is → what's next → why this pace.*

### Theme D — Communication Is Structurally Asymmetric

`parent-on-rail:receives-but-can't-reply` + `want:bidirectional-teacher-dialogue` + `parent-diy:curriculum-prep` + `want:parent-prep-material` point to a product that treats parents as *audience*, not *participants*. Teachers send daily feedback. Parents cannot send back. Parents cannot pass information to the teacher (child had a bad night, child is anxious about grammar, child practiced specific vocab with grandparent). The design assumes parents consume the product; the reality is parents want to co-produce it.

**Design imperative**: bidirectional communication — with bounded parent-load (not "now you're the homework manager" but "here are 3 things you can tell the teacher before the next lesson, 30 seconds").

### Theme E — Trust Is a Stack That Can Fail From Any Level

`trust-break:opaque-billing` destroys trust in everything else. `trust-break:teacher-change-unilateral` destroys trust even when pedagogy is fine. `trust-break:progress-button-permission` destroys trust at a UX level below the data level. **Trust fails from the bottom up — billing → teacher continuity → feedback honesty → data → narrative.** If any layer is broken, the layers above cannot compensate.

**Design imperative**: the Value Hub cannot succeed as a data visualization if the underlying layers leak trust. The redesign brief is not just "design a dashboard" — it's "design the trust layer, and the dashboard rides on it."

### Theme F — The Plateau Is the Product's Most Expensive Silence

40 of 45 synthetic personas + British Council SLA research + multiple real quotes about "felt flat, considered cancelling" → parents hit a plateau in months 2–8, frame cancellation as a real option, and are told *nothing* about the plateau being a known, documented phase of language acquisition. Every parent enters the plateau unaware. Every parent leaves it either by attrition or by stumbling through. **This is the single highest-ROI intervention point in the entire customer journey** and the current dashboard does not even acknowledge the phenomenon exists.

**Design imperative**: the Value Hub must proactively name the plateau *before* it happens (onboarding), *during* it (support), and *after* it (celebrate the breakthrough). This is not a feature — it is a narrative the product owes the parent.

### Theme G — Culture Changes the Metric, Not the Job

Existing cultural analysis is solid: Russians frame via olympiads/school rank, Italians via price-per-lesson, Turks via real-world confidence, Poles via CEFR, French via school gap, Germans via social integration. **But the underlying *job* is identical across cultures** — "help me believe the investment is working." The variation is in the *acceptable proof*. A Russian parent accepts "best in class" as proof. An Italian parent accepts "€8/lesson + visible skill gain" as proof. A French parent accepts "closes the gap the school can't."

**Design imperative**: the Value Hub needs a **proof layer that is culturally configurable** — not seven different dashboards, but one dashboard that can surface the proof-type that resonates in each market. This is a data-display decision, not a research insight decision.

---

## PART 3 — ARCHETYPES

When I cluster parent *jobs* (not demographics), six candidates surface. These are bottom-up from the codes, not top-down from a typology. Each maps cleanly to SDT/EVT/JTBD.

| # | Candidate name | Core job | Primary need (SDT) | Primary value term (EVT) | Representative quote |
|---|---|---|---|---|---|
| 1 | **The Reassurance-Seeker** | *"Tell me I'm doing the right thing as a parent"* | Competence (as parent) | Attainment value | Rita mum, Keren A |
| 2 | **The Auditor** | *"Show me the evidence so I can judge"* | Competence (as evaluator) | Expectancy | Margarita IT, Елена RU |
| 3 | **The Investor** | *"Defend the money to me"* | Autonomy (as decision-maker) | Utility − Cost | Giulia IT, Rafal |
| 4 | **The Navigator** | *"Show me the path ahead"* | Competence (planning) | Expectancy | Nejat TR, Lyudmila RU |
| 5 | **The Co-Pilot** | *"Give me a role I can play without burden"* | Relatedness | Attainment value | Margarita IT, Carmentxu ES |
| 6 | **The Outsourcer** | *"Just work, don't make me manage this"* | Autonomy (as in ceding control) | Cost (minimize) | long-term loyal users with minimal engagement |

These are *jobs*, not *people*. One parent can be Reassurance-Seeker on Monday and Auditor on Friday (e.g., plateau triggers audit mode). The dashboard should serve the job, not the person.

Full specification in [`archetypes.md`](archetypes.md).

---

## PART 4 — NARRATIVE AXIS

The research is structured around **The Information Bridge** — the Value Hub as a translator between information silos: *what AI knows → what teacher knows → what child experiences → what the world sees → what the school measures*. Each chapter in the final synthesis follows one silo, then the bridges between them. Matches the word "hub" in "Value Hub" literally.

Full synthesis in [`synthesis.md`](synthesis.md).

Once decided, I can sequence the next four deliverables:
- cultural-analysis.md (standalone layer, not embedded in personas)
- archetypes.md (detailed spec of the 6 jobs)
- interview-guide-v2.md (problem-space, 80/20)
- persona interviews (5–8 personas, problem-space)
- synthesis.md (in the chosen narrative axis)
- research.html (rewritten)

---

## Appendix — corpus statistics after coding

| Source | Verified quotes | New this session | Total |
|---|---|---|---|
| Trustpilot global | 40 | — | 40 |
| Trustpilot regional (6 domains) | 22 | — | 22 |
| Russian platforms | 12 | 23 | 35 |
| Alternative platforms (app stores, forums) | 19 | 3 | 22 |
| EdSource (US parent context) | 0 | 5 | 5 |
| Other aggregators | 0 | 4 | 4 |
| **Total** | **93** | **35** | **128** |

| Theoretical framework | Applied to |
|---|---|
| Self-Determination Theory (Deci & Ryan) | Parent autonomy/competence/relatedness |
| Expectancy-Value Theory (Eccles) | Renewal decision decomposition |
| Jobs-To-Be-Done (Christensen school) | 6 archetype mapping |
| Parental Educational Anxiety Scale (Guo et al. 2024, 2025) | Anxiety-as-design-variable |
| Baumrind parenting typology | Persona design dimension |
| British Council plateau research | Theme F |
| Dashboard UX for non-experts (F-pattern, info overload) | Dashboard structural constraints |
| Behavioural economics (endowed progress, loss aversion) | Renewal UX framing |
| Dinolingo CEFR-for-parents | CEFR→observable translation |

Coding confidence: **high** for themes A–F (extensively corroborated). **Medium-high** for theme G (cultural) — corroborated in codes but needs the standalone cultural-analysis pass. Archetypes are **candidates pending user validation.**