See what happens when a student starts to drift.
What I built: four fictional learner journeys, adjustable risk signals, a simulated catch-up experience, report previews and an editable financial model. Pick a learner and change the signals to see the concept respond. What this does not do: run trained AI, call PW APIs, use real learner data, contact parents or prove growth uplift. The rules and responses are illustrative.
Part of Anchit Tandon's Senior Product Manager - Growth interview case study for PW. Step-by-step review: anchit-tandon.com/pw-study#walkthrough
On this screen: Drop-off Risk Score (engine output; the phone shows it as an on-track score), Catch-up Plan (the plan in the phone) and Next-Step Recommendation (bottom of the phone).
1 · Pick a student
Namaste Riya
This week's on-track score
Needs a push this week
6 lectures behind batch pace
AI Mentor plan · added to your Tracker 360 schedule
Ask your AI Mentor Simulated
Your next step
2 · Drag the signals
The phone and the score reactDrop-off Risk Score: engine output
Drop-off risk
/ 100
This is a transparent rule-based illustration, not a trained ML model. A production system would first validate internal data access and quality, legal eligibility, and the performance of simple rules. Predictive ML would be added only if temporal calibration, precision and intervention experiments justify it.
Weekly Progress Report on WhatsApp
The PW Parent App already shows attendance, lecture, DPP and test stats, but a parent has to open it. The Weekly Progress Report pushes a plain-language summary on WhatsApp and links into the app. Who receives it depends on the learner's age. It is built from the same learner and signals as the Student app tab: drag the signals there and this message changes.
| Learner | Examples | Weekly report goes to |
|---|---|---|
| Under 18 | K-12, Class 11-12, boards | To the parent, with consent taken at purchase (DPDP Act 2023 needs it for under-18s) |
| 18-21 | JEE/NEET droppers, early college | To the learner. The parent gets a copy only if the learner opts in |
| Adult aspirant | UPSC, govt exams, CA, NEET PG, skills | To the learner, plus a study-circle check-in. No parent channel |
Guardrails built in
- The learner always sees exactly what is shared, and with whom.
- One report a week, max one alert in between. No fear or rank-shaming language.
- High-risk learners still behind after the 7-day plan get a human mentor call, never only a bot.
- Offers appear only at natural moments (results, batch end), never inside a risk alert.
- Under 18: offers and prices go only to the parent; the child sees an academic next step (DPDP Act 2023, s.9(3)).
To Riya's parent · business account (concept)
Renewals Dashboard
The screen PW's retention team would open every Monday: who is drifting, what action each group gets, and what that is worth. Uses a made-up cohort of 5,000 paid learners so it runs without PW data.
1Who is drifting
For this demo, synthetic learners receive an illustrative risk score. Production scoring would be limited to legally eligible learners, validated events and permitted educational purposes.
2What we do about it
Learners are grouped by the action they get. Click any row to see who is in it and how the ₹ is worked out.
3What it is worth for PW
A what-if calculator on PW's real FY26 numbers. Change any assumption and the total updates.
1 · Who is drifting this week
Share of the cohort in each risk tier. High = 60+ risk score, Medium = 35-59, Low = under 35.
2 · What we do about it
The calculator's total, allocated to groups by each lever's rule; sorted by expected extra fees. Click a column header to re-sort, or a row for detail.
How I would validate this-after joining PW
First test a simple, auditable missed-practice rule. Backtest any future ML model for calibration, precision at capacity and subgroup errors; child data use requires legal approval.
Randomise a legally eligible cohort into current support versus a guided catch-up plan; preserve existing help for both groups. Recalculate sample size from the real baseline.
Verified practice recovery in seven days; check repeat practice and unassisted mastery. Read incremental next-cycle enrolment later when the cohort matures.
Mastery must not decline. Check unsafe advice, data accuracy, consent, opt-outs, complaints, cost and subgroup fairness.
3 · What it is worth for PW
Assumptions: edit meScales the idea to all of PW. Base figures are PW's FY26 disclosures (Q4 FY26 shareholder letter); uplift rates are my assumptions to be tested.
Running costs a year
Annual incremental collections
Net of running costs
PW brand context schema
Extracted from computed styles on pw.live (homepage, NEET Class 12 listing, a batch detail page) on 25 Oct 2026 with a headless browser. This prototype uses only these tokens. The PW logo is deliberately not reproduced.
Colour tokens
Type
- Reddit Sans only, weights 400/500/600/700 (Noto Sans Devanagari added here for Hindi).
- h1 40/50 700 · h2 32/48 700 · card title 20 600 · body 16/24 400 · label 12 500.
- Buttons 16px: 400 in header, 600 for primary actions.
Shape and depth
- Buttons 6px radius, 48-56px tall. Tags 4px. Panels 8-12px. Batch cards 16px with 1px #D9DCE1 border, no shadow. Pills 9999px.
- Shadows: light 0 1px 8px #00000014; detail cards 0 4px 16px rgba(27,33,36,.06).
- Hero gradient: 270deg #EEF9FF → #F9EEFF → #F8F8FF → #F1F5FF.
Batch card anatomy
- Banner with faculty art and plan strip ("Infinity, Pro").
- Orange category label (e.g. "Dropper") + language pill ("HINGLISH").
- Title 20/600, audience line, start and end dates.
- Price 18/700, struck MRP, green "% OFF"; dark "Buy Now" + outline chevron.
Voice
- Warm, aspirational, numbers-led ("Join 15 Million students", "24*7 Doubt Support").
- English UI; Hindi carries the brand in names: Arjuna, Lakshya, Yakeen, Vidyapeeth, Pathshala.
- Teachers are "Sir"; batches are labelled HINGLISH; emoji rare.
- This concept keeps plain English feature names and UI, with Hinglish only inside student chat.