DhoopProof — performance proof desk for solar installers

1. One-liner

Sends every rooftop customer a trusted monthly “your solar is working” report so small installers stop drowning in “not generating” calls.

2. Trend signal — why now?

India put 32 lakh households on rooftop solar through PM Surya Ghar as of March 2026, and the market is 20.84 GW growing at 18.4% CAGR. That flood of new residential systems just met the ugliest number in the industry: “98% of solar rooftop owners have no idea their system is underperforming.” The system is a black box. The bill creeps up, the homeowner blames the installer, and there’s no data to settle it.

That fight is now hitting courts. On July 6, 2026 — two days before I wrote this — a District Consumer Disputes Commission in Vizianagaram ordered a renewable-energy firm to pay homeowner Ijjada Vivekananda ₹25,265 because his rooftop system “produced far less electricity than promised” and he had nothing to prove otherwise except his own bill. Multiply that risk across a long tail of small installers who each did 20–200 jobs on thin margins, have no monitoring discipline, and can’t afford a legal fight per unhappy customer.

Meanwhile the enabling piece quietly landed: Growatt shipped a token-based open API in 2025, and the other volume brands in India (Solis, Deye, Luminous, SolarEdge, Sungrow) expose per-plant generation over cloud APIs, with mature community libraries proving the data pull is cheap and reliable. Plain kWh + an expected-yield model + a cheap LLM = a homeowner-readable report. The data exists; nobody’s packaging it for the small installer as a trust artifact.

Provenance:

• Signal 1 (demand): “98% of rooftop owners have no idea their system is underperforming”; fresh consumer-court judgment fining installer ₹25,265 for underperformance — https://www.businesstoday.in/latest/trends/story/rooftop-solar-failed-to-deliver-promised-savings-court-orders-renewable-energy-firm-to-pay-rs25265-541205-2026-07-06 — 2026-07-06
• Signal 2 (feasibility): Growatt token-based open API (2025) + community libraries for multi-brand inverter data — https://pypi.org/project/growattServer/ — 2026-07 (accessed)
• Signal 3 (economic): India rooftop solar 20.84 GW at 18.4% CAGR; SolarSquare raising ~$60M at ~$500M; AMC market real at ₹1.5–8K/kW/yr with “generation performance report” a standard line item — https://techcrunch.com/2026/05/23/solarsquare-in-talks-to-raise-up-to-60m-as-indias-rooftop-solar-market-draws-major-vc-interest/ — 2026-05-23 Category: Tech-unlock

3. The opportunity

Two adjacent tools already exist and both miss this cut:

• Full installer OS platforms (Solar Ladder, seed $1.3M; SurgePV) — CRM + proposal + design + O&M dashboards sold per-seat to the installer’s back office. Great for a 3D quote, useless as a thing the homeowner trusts. The homeowner never sees them.
• Inverter-brand apps (Growatt ShinePhone, SolarEdge Monitoring) — hardware-locked, English, engineer-facing graphs. A retired schoolteacher in Coimbatore does not read a per-string I-V curve. And if the installer put Growatt on 30 homes and Solis on 20, there’s no single view.

The gap in the middle: a homeowner-facing proof-of-performance layer, brand-agnostic, in the homeowner’s language, delivered over WhatsApp, priced per-plant for the installer. The insight is that the buyer (installer) and the reader (homeowner) are different people, and nobody has built the artifact for the reader. It converts the installer’s biggest liability — the “not generating” phone call and the consumer-court exposure — into a monthly touchpoint that also justifies an AMC renewal. That’s a 10× better outcome than a dashboard the customer never opens.

4. Target market

• Primary customer: Owner-operators of small/mid Indian residential rooftop solar installers — 20 to ~500 completed installs, lean teams of 3–15, no in-house software. Tier-1/2/3 cities. Not the SolarSquare/Tata tier (they build their own).
• Why they buy, in their words: “Vendor ghosting post-installation” and “contractors becoming unresponsive after advance payment” are the top complaints against them — but the honest ones are drowning: every dusty month a batch of customers calls saying “bijli kam ban rahi hai” and the installer sends a guy on a bike to check a system that’s actually fine. They want to (a) kill nuisance calls with proof, (b) catch the real faults before a Google 1-star or a consumer-court notice, (c) finally sell AMC — which they can’t, because they can’t produce the “generation performance report” the AMC is supposed to include.
• Rough TAM reasoning: India has thousands of MNRE-empanelled / DISCOM-registered residential installers. If even 5,000 small installers each carry 200 live plants, that’s ~1M monitorable homes — and PM Surya Ghar is adding more monthly toward a 1-crore-household target. I don’t need the whole market; a few thousand installers is a full business.
• Why now for them: Their install volume just 3–5בd off subsidy, their post-sale service burden scaled with it, and the July 2026 court ruling turned “annoyed customer” into “financial liability.” The pain moved from irritating to expensive this quarter.

5. Product sketch (MVP)

• Installer onboards a plant in 2 minutes: pick inverter brand, paste the plant’s API credentials (or homeowner OAuth), enter system kW and pin-code.
• DhoopProof pulls daily generation, blends it with pin-code irradiance/weather, and computes expected-vs-actual for that specific roof — no generic benchmarks.
• Monthly WhatsApp report to the homeowner, in their language: “Your 5 kW system made 612 units this month — 96% of what it should in July. ₹4,300 saved. All healthy. — from [Installer Name].” Installer-branded.
• Early-fault alerts to the installer, not the homeowner: string dropout, inverter offline 3+ days, sustained soiling loss >15%, so the fix happens before the complaint.
• One dashboard across all brands — Growatt, Solis, Deye, Luminous, SolarEdge in a single plant list with a red/amber/green health column.
• Consumer-dispute pack, on demand: a timestamped generation history + expected-yield methodology PDF the installer hands to a lawyer or DISCOM if a “not generating” claim escalates.
• AMC renewal nudge: flags plants approaching AMC expiry and drafts the renewal message with the year’s performance summary attached.

6. AI angle — what’s load-bearing

Two places AI does real work, not decoration:

1. The report writer. Raw kWh + weather + expected-yield delta → a paragraph a 60-year-old non-technical homeowner in Marathi or Tamil actually trusts and doesn’t panic over. Tone matters: “healthy,” “slightly low this month due to monsoon cloud — normal,” “we’ve flagged this for a technician.” Templating this across brands, languages, seasons, and fault types by hand is exactly the brittle work an LLM collapses.
2. The anomaly triage. Separating real underperformance (fault) from expected dips (monsoon, winter, genuine soiling that a cleaning fixes) so the installer isn’t buried in false alarms — the thing that kills every naive threshold-alert system. LLM + a light expected-yield model reads the pattern and classifies “dispatch a tech” vs. “reassure the customer” vs. “sell a cleaning.”

Remove the AI and you have a raw kWh CSV per brand that the homeowner can’t read and the installer won’t triage. The product is the translation and the triage.

7. Localization angle

This is India-first by construction, not by decoration:

• Language/script: the homeowner report must land in Hindi, Marathi, Tamil, Telugu, Gujarati, Kannada — the reader is not English-comfortable. This is the whole moat vs. the brand apps.
• Channel: WhatsApp, not email, not a portal login. The homeowner will never install an app; they’ll read a WhatsApp forward from “their solar guy.”
• Pricing: ₹30–60/plant/month works in India where a $10/plant Western SaaS can’t. The installer bundles it into an AMC they charge ₹3,000–12,000/year for.
• Brand mix: India’s residential fleet skews Growatt/Solis/Deye/Luminous — different from the SolarEdge/Enphase mix the Western fleet tools (Sunvoy) index on. Getting the Indian brand coverage right is a distribution edge.

8. Business model — path to $1M–$5M ARR

• Pricing: ₹40/plant/month, billed to the installer, minimum ₹1,500/mo (covers ~40 plants). Installer marks it up inside a ₹3–12K/yr AMC or eats it as a retention cost.
• ACV: an installer with 200 live plants = ₹8,000/mo = ₹96,000/yr ($1,150). Small installers with 40–60 plants ~₹24,000/yr.
• Rough math to $1M ARR (~₹8.3 crore): ~700 installers averaging 250 plants at ₹40 = ₹70L/mo ≈ ₹8.4 crore/yr. Or fewer, larger installers.
• Rough math to $5M ARR: ~3,000 installers, OR move up-market to C&I/MSME rooftop (bigger systems, ₹150–300/plant, real O&M money) once residential is proven, OR add a per-report consumer-dispute-pack and cleaning-lead-referral revenue line.
• Expansion path: plants-per-installer grows automatically as they install more (usage-priced, net-negative churn baked in). Then upsell: dispute packs, DISCOM/net-meter follow-up nudges, cleaning-vendor marketplace take-rate, and a homeowner-paid “premium alerts” tier.

9. Go-to-market wedge — first 100 customers

• Scrape the MNRE / DISCOM empanelled-installer lists (public, state-wise, thousands of names with phone numbers). These are exactly the small installers. Cold-WhatsApp them a 40-second Loom showing their own brand logo on a sample homeowner report. Expect small % reply — but the list is long and free.
• Pilot free on 25 plants. Onboarding is API-credential paste; I can light up a real installer’s fleet in an afternoon and let their next month’s reports sell it. “Show, don’t tell” is trivial here because the data already exists.
• Ride the consumer-court fear. The July 2026 ruling is a marketing gift — one WhatsApp forward (“here’s the ₹25,265 case; here’s how you never get one”) to installer groups. Every state has active WhatsApp/Telegram installer communities; seed the dispute-pack angle there.
• Inverter distributor channel. Growatt/Solis/Luminous regional distributors sell to hundreds of these installers and want stickier customers — co-market as a value-add. One distributor relationship = a batch of installers.
• Partner with AMC-hungry installers first. Target the ones already trying to sell AMC and failing on the “report” deliverable — they convert fastest because DhoopProof unlocks revenue for them, not just cost.

10. Build complexity — justification

Medium. Off-the-shelf: inverter cloud APIs (Growatt token API, community libs for Solis/Deye/Luminous), WhatsApp Business API, LLM for reports/triage, standard web stack. Custom work is the annoying-but-bounded part: normalizing 5–6 brand APIs into one schema (each is a different, sometimes flaky, undocumented-in-parts cloud), a defensible per-roof expected-yield model (irradiance × derate — not research, but must be credibly accurate or the reports lose trust), and multilingual report quality. A pair can ship a 2-brand, 2-language v1 in ~10–14 weeks; full brand/language coverage is the following quarter.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Reading a plant’s own generation data with installer/owner consent. No regulated activity.
Ethical — no harm / dark patterns	✅	The opposite — it gives homeowners honest visibility they currently lack.
Market exists (evidence above)	✅	32 lakh new systems, “98% don’t know,” live court case, AMC market with the exact deliverable gap.
1–5 person team can build this	✅	Medium build, off-the-shelf APIs, no ops fleet.
Launchable with <$50K / ₹40L	✅	Software + API costs. Distribution is cold-WhatsApp + free pilots.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	15/20	Real and now court-priced pain for installers, but it’s a fear/reputation pain, not a daily cash-bleed — some installers will ignore it until burned.
Demand evidence	15	12/15	”98% unaware” stat, fresh judgment, AMC “report” gap, ghosting complaints. Strong signals; the specific installer buying behavior for this exact SKU is still inferred.
Build feasibility	15	11/15	Off-the-shelf APIs, but multi-brand normalization + a trustworthy expected-yield model + multilingual quality is real engineering discipline.
Distribution clarity	15	12/15	Public empanelled-installer lists + free-pilot + distributor channel is concrete. WhatsApp cold-outreach reply rates uncertain.
Revenue mechanics	15	12/15	Usage-priced, net-negative churn, clear AMC attach. ACV is modest — needs volume or up-market move for $5M.
Time to first revenue	10	7/10	Free pilot → paid after the first month’s reports land. ~6–10 weeks realistic; installer buying cycles are slow-ish.
Defensibility	10	5/10	Execution moat + brand-coverage + installer relationships. Solar Ladder/SurgePV could bolt on a homeowner report; the language + WhatsApp + per-plant-pricing focus is the edge, not a wall.
Total	100	74/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy (multi-brand API integration is the core lift) · domain-expertise-required (must understand PV yield, soiling, net-metering, and how Indian installers actually sell AMC).

Key assumptions to validate (3–5)

1. Assumption: Small installers will pay ₹40/plant/mo to reduce service calls and enable AMC. How to test: 20 installer interviews + a paid pilot; kill if <5 of 20 will put a card down after seeing their own report.
2. Assumption: The volume brands’ APIs (esp. Solis/Deye/Luminous) are reliably accessible at the plant level without per-plant friction. How to test: integrate the top 3 brands and pull data from 50 real plants before writing a line of GTM.
3. Assumption: A per-roof expected-yield model can be accurate enough (±~8%) that reports build trust instead of triggering disputes. How to test: backtest against 100 plants with 6+ months of history; measure predicted-vs-actual error.
4. Assumption: Homeowners open and trust a WhatsApp report enough that it actually deflects calls. How to test: run 25 plants for 2 months; measure inbound “not generating” call rate vs. a control cohort.

Risk flags

1. Platform dependency: Built entirely on third-party inverter cloud APIs. Growatt/Solis can rate-limit, break, or close access. Mitigate by breadth (many brands) and, later, optional read-only dataloggers.
2. Incumbent bolt-on: Solar Ladder or SurgePV adds a homeowner WhatsApp report and undercuts the whole wedge. The defense is speed, Indian-brand + language depth, and owning installer relationships before they notice.
3. Weak-pain-until-burned: Reputation/legal fear is a slower buy than a cash-bleed. Risk that installers nod and don’t pay until their court notice arrives — lengthening the sales cycle.
4. Data-accuracy liability: If a report says “healthy” and the system was actually failing, DhoopProof inherits blame. Needs conservative alerting and clear methodology disclaimers.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder comfortable with flaky third-party APIs, paired with (or advised by) someone who has sold residential solar/AMC in India
Time to revenue:        6–10 weeks (free pilot → paid after first monthly report)
Capital to launch:      ₹5–8 lakh ($6–10K)
Top 3 assumptions to validate first:
  1. Installers will pay ₹40/plant/mo — 20 interviews + paid pilot, kill if <5/20 convert
  2. Top-3 Indian inverter-brand APIs are reliably pullable — integrate + pull 50 real plants first
  3. Expected-yield model hits ±8% — backtest 100 plants with history before GTM
Kill criteria:
  - Abandon if <25% of 40 cold-outreach installers agree to a free pilot
  - Abandon if the expected-yield model error exceeds ~12% and reports trigger more disputes than they deflect
  - Abandon if two of the top-3 volume brands lock or price out API access with no datalogger fallback

15. Next step — 1-week validation sprint

• Day 1–2: Integrate the Growatt token API + one more brand. Pull real generation from 30–50 live plants (borrow an installer friend’s fleet or use public/demo plants). Build the crude expected-yield model and eyeball predicted-vs-actual.
• Day 3–4: Generate 10 sample homeowner WhatsApp reports in Hindi + one regional language. Cold-WhatsApp 40 empanelled small installers with a Loom of a report carrying their brand. Book 8–10 calls.
• Day 5: On the calls, show the report and ask for a ₹1,500/mo pilot commitment (soft — “start free, pay after month one”). Go if ≥5 of 10 say yes and the model error is under ~10% on the test fleet. No-go if installers shrug at the report or the API pulls are too flaky to trust.

The falsifiable outcome: ≥5/10 installers commit to a pilot AND expected-vs-actual error <10% on a 50-plant test fleet. Miss either and it’s not ready.