Repository navigation
bear case eoy2026
A pre-mortem. Research-driven, numbers first, written 17 Jul 2026 — the morning of the first WTP attempt. Every claim is sourced or derived; every derivation shows its arithmetic. The job of this document is to be right if the project fails, so the win-plan that follows it has something real to beat.
You sell a digital register to India's affordable private schools (APS). The research consensus on that segment:
- 200,000–400,000 APS schools serving 60–90 million children — 70–85% of private enrollment in the most populated states (FSG, Edufinance).
- 70–85% of children in private unaided elementary schools pay less than ₹500/month in fees (IHD working paper). The whole APS market is estimated at ~$5.2B — that is total school revenue, not software spend.
- School management software in India prices at ₹20–₹100 per student per month, i.e. ₹1.2–6 lakh/year for a 500-student school (LMSCloud cost guide) — and the segment that actually pays those prices is the ₹2,000+/month-fee school, not the APS.
The uncomfortable arithmetic of your price point. Your offer is ₹20/student/month — the absolute floor of the market range. Take a typical APS: 300 students × ₹400/month average fee = ₹1.2 lakh/month total school revenue, most of which is teacher salaries and rent. Your ask is ₹6,000/month — 5% of gross revenue for attendance software, at a school whose alternative is a ₹60 paper register and a free WhatsApp group. The research phrase for this segment is "highly fragmented, extremely price-sensitive." You are not underpriced. You are asking a thin-margin business for a visible new line item.
- Teachmint: VC-backed, freemium, aimed exactly at budget schools (Decentro ERP survey).
- Fedena: free open-source community edition, 15 years old (Fedena pricing).
- WhatsApp itself: the incumbent for parent communication is free, already installed, and the school already trusts it. Your own strategy memo concluded absence alerts are commoditized — that is why the automated pipe is paused.
- Government rails (UDISE+ reporting) already force schools to maintain digital records once a year, and they do it under duress, not enthusiasm.
A solo founder's differentiation cannot be features (Teachmint has more), price (free exists), or brand (nobody has heard of you). The plan's honest answer is service intensity — you personally onboard every teacher. That is also the thing that cannot scale, which is Section 5.
- 2024 was the worst year for edtech VC in a decade, and 2025 was worse (EdWeek Marketbrief). There is no "raise a seed on a pilot" path in this category in 2026.
- Byju's went from $22B to zero (CNBC, TFN) — and took the sector's trust with it. Every school owner in India has now been burned by, or read about, edtech sales tactics. Your cold call this morning lands on that soil.
- The VC-era market-size claims were fiction: against the "$90B+" narratives, actual direct-to-consumer K-12 online spend realized at $2–3B (~11% of the claimed supplemental spend) (RAYSolute 2026 analysis). Assume the B2B-school equivalents are similarly inflated.
- On the AI hope specifically: Goldman Sachs' research (Gen AI: too much spend, too little benefit?) collects the sober view — Acemoglu's estimate is that AI will cost-effectively automate <5% of tasks within a decade, and most firms report unclear ROI on GenAI spend. Translation for Class-mode: AI test generation is a cost center you can't yet pay for (your Gemini key is broken today because billing isn't linked), not a willingness-to-pay driver. Your own probe data agrees: two AI-generation outages produced zero user complaints.
Sector research: 85% of edtech tools are poorly implemented or a poor contextual fit; 65% of administrators have discontinued a tool they previously adopted (SchoolAI); tools fail when they add steps to an existing workflow — and marking attendance on a phone is more steps than a tick in a paper register that doubles as a legal record the school already trusts (Cambridge).
Your production database, W29 baseline: 14 accounts, 0 active teachers, 0 attendance rows, 0 fee entries, 0 organic usage events. The product has existed for months. Nobody uses it when you are not in the room. The sector base rate and your local evidence point the same direction.
Assume everything goes right: the school signs today.
| Variable | Value | Basis |
|---|---|---|
| Revenue per school (300 students × ₹20) | ₹6,000/mo | your offer |
| Founder capacity: onboard + support | ~2 schools/month max | in-person training per teacher (your own offer's promise), solo, while also building |
| Sales conversion (APS cold outreach) | generously 1 in 5 | no brand, no channel, post-Byju's trust deficit |
| Schools live by 31 Dec 2026 (Aug–Dec, 5 months) | 4–8 best case | capacity-bound, not demand-bound |
| MRR at 31 Dec 2026, best case | ₹24,000–₹48,000 (~$290–$580) | 4–8 × ₹6,000 |
| Churn exposure | annual; decisions cluster at term boundaries | 65% admin discontinuation base rate |
₹48,000/month gross, before Azure (your bill is real; the $1K Founders Hub credit you haven't applied for yet is the mitigation), before travel, before the AI costs you'd incur if anyone used those features. That is not a company; it is a hard job with negative margin. And it's the best case: it assumes the WTP answer today is yes, teacher adoption defies an 85% base rate, and nothing churns.
The plan's own tripwires agree with this bear case, which is to its credit: no WTP by Aug 14 pauses build; the streak must start Sep 8. But note what even hitting them yields by December: one to a handful of schools and a four-figure-rupee MRR. "Worthy" was never on the 2026 scoreboard — survival to a fundable/expandable 2027 was.
- The school says a polite yes and never pays (verbal WTP ≠ collections; APS cash cycles are fee-season shaped). Mitigation exists in the plan: written answer + named signer.
- Teachers don't mark attendance after week 2 — the 0-teachers baseline replays with an audience. 85%/65% base rates above.
- The founder becomes the product — every new school adds permanent support load; build velocity → 0; the Lab/learning-map differentiators (gated on payment anyway) never ship.
- Price collapses under negotiation — ₹20 → "₹10 and free till April" → the revenue math halves before it starts.
- A funded freemium (Teachmint) or a free tool is "good enough" the moment your service intensity can't cover the next school.
- Language/localization drag — TD1 (school language unnamed) is still open the morning of the visit; the vernacular pass is scoped M and unstarted.
This document loses if, by the dates below, the following are true — all measurable in your own tables:
- Aug 14 (day-30 line): a signed offer with money moved (not promised) — beats failure mode #1.
- Sep 30: ≥60% of the pilot school's teachers marking attendance ≥4 days/week without founder presence — beats #2 and the 85% base rate.
- Oct 31: second school signed at the same price from a referral you didn't engineer — evidence of pull, beats #5.
- Dec 31: ≥5 paying schools with <1 school-day/week of founder support each — beats #3, and turns the Section 5 table from ceiling into floor.
Anything less than this by year-end and the correct 2027 move per your own escalation rule is not "try harder" — it is the mapped real buyer, a different wedge, or a different market.
Sources: Goldman Sachs — Gen AI: too much spend, too little benefit? · FSG APS landscape · IHD low-cost private schools · Edufinance · LMSCloud pricing guide · EdWeek Marketbrief on edtech VC · CNBC on Byju's · RAYSolute Indian edtech 2026 · SchoolAI adoption research · Cambridge on teacher training · Decentro ERP survey · Fedena pricing · plus Class-mode production metrics, W29 baseline (docs/dashboard/metrics/2026-W29.json).