SeisanCheck

1. One-liner

Recomputes the proxy-control deduction on your Japanese solar payout and shows which hours you were wrongly billed for.

2. Trend signal — why now?

Japan’s grid operators curtail solar more every year, and 2026 is the record. The Agency for Natural Resources and Energy’s short-term outlook puts FY2026 curtailment at 2.53 billion kWh — the highest ever, roughly 1.25× FY2025’s ~2.0 billion kWh — concentrated in Kyushu (1.22bn kWh), Tohoku (0.75bn) and Hokkaido (0.11bn). In March 2026 TEPCO Power Grid curtailed solar for the first time in its history (1 March 2026, 11:00–16:00 JST), meaning Tokyo — the last holdout area — is now in the game.

The part nobody has tooled is the settlement, not the curtailment. Since April 2022, Japan runs オンライン代理制御 (online proxy control / “economic curtailment”). Old-rule offline plants between 10kW and 500kW — which have no online-control obligation — don’t physically curtail at all. Instead an online plant curtails on their behalf, and at settlement the law deems that the offline operator curtailed and the online operator generated. The offline operator then pays for the deemed curtailment via a negative 精算比率 (settlement ratio) applied to their purchase price.

Three things make that deduction near-impossible for the owner to check:

1. The deemed hours are fictional and unfavourable. Shikoku Electric Power Transmission’s own FAQ states that online plants curtail only the minimum hours actually needed that day (its example: “demand rose mid-day so control was only 1 hour”), while offline plants are settled as if they had curtailed a fixed block every time — its example is “fixed at 8 hours a day”, regardless of what the online plant actually did.
2. The money lands 2–3 months later. Settlement is applied to the purchase payment two months after the curtailment month, as a minus line — long after the owner has forgotten the weather that day.
3. The utility issues no proof and won’t answer. Same FAQ, Q4-3: asked for a document certifying curtailment (for an insurance claim), the answer is “we do not issue a certificate of curtailment.” Q7-2: an O&M company asking whether its client’s plant was curtailed is told “we cannot respond to individual curtailment record enquiries” — only the ID-holding owner can log into the mypage.

The utility’s own FAQ is effectively a complaint log: Q4-6 (“am I being curtailed more than other plants?”), Q4-7 (“I can’t accept this unless you tell me how many plants are in the rotation”), and Q6-3 — “the proxy-control settlement ratio is high and a substantial amount is being deducted” — which the utility answers by recommending you spend money on going online.

Provenance:

• Signal 1 (demand): Shikoku Electric Power Transmission official curtailment FAQ — offline plants settled on assumed fixed hours (e.g. 8h/day) vs online plants’ actual minimum hours; no curtailment certificate issued (Q4-3); refuses O&M-company record enquiries (Q7-2); Q6-3 documents owners complaining that “a substantial amount is being deducted” — https://www.yonden.co.jp/nw/assets/renewable_energy/output_control/faq.pdf — updated 2026-03-05
• Signal 2 (feasibility): Every TSO now publishes machine-readable monthly 精算比率 (settlement ratios) per voltage class and rule type, plus curtailment instruction records — e.g. TEPCO PG FY2026 ratios (2026-06-18), Tohoku EPCO FY2026 ratios (2026-06-30), Shikoku ratios — making independent recomputation possible without utility cooperation — https://www.tepco.co.jp/pg/consignment/access/pdf/260618_seisanhiritsu.pdf — 2026-06-18
• Signal 3 (economic): ANRE short-term outlook — FY2026 curtailment hits a record 2.53bn kWh (~1.25× FY2025), concentrated in Kyushu/Tohoku/Hokkaido; TEPCO curtailed for the first time ever on 2026-03-01, extending exposure to the Tokyo area — https://www.tainavi-pp.com/investment/solar/27/ and https://www.pv-magazine.com/2026/03/24/solar-curtailment-reaches-tokyo-japans-last-holdout-grid-area/ — 2026-03-24 Category: Underserved niche

3. The opportunity

The grid operator is simultaneously the party that curtails, the party that classifies the curtailment, the party that computes the settlement ratio, and the party that deducts the money — and it issues no certificate and won’t answer third parties. The generator’s only record is a login-gated mypage that shows instructions, not money.

This is the classic self-graded-metric gap. Nobody sells the counter-calculation.

The incumbents are remote-monitoring vendors — エコめがね (NTT Smile Energy, the sub-50kW market leader), and the 7-odd 遠隔監視 systems compared in the trade press. They are all built around the same job: is the plant generating, and did a PCS fault. They alert on sudden generation stops and produce monthly generation reports. None of them reconciles the deemed-curtailment deduction on the purchase statement, because that requires modelling a legal fiction (deemed hours), not a physical fact (kWh produced). Their own blogs write explainers about proxy control — they treat it as content, not as a product surface.

The specific 10× : an owner today either (a) doesn’t notice the deduction, (b) notices revenue cratered and assumes the plant is broken — 九州新エネルギーメンテナンス reports fielding exactly these enquiries, owners asking whether something is damaged when the real cause was the settlement ratio — or (c) has to hand-join a PDF of monthly settlement ratios, the mypage instruction log, and their own generation data. We do (c) automatically and tell them the number.

4. Target market

• Primary customer: Owner-operators of old-rule (pre-2015 FIT), 10kW–500kW offline solar plants in the Kyushu, Tohoku and Hokkaido grid areas — the ones subject to online proxy control. Typically an individual investor, a small property/asset company, or a regional firm holding 1–20 low-voltage plants as a yield asset. Secondary: O&M and asset-management firms who are explicitly refused record access by the utility (Q7-2) and currently have to ask their client to log in.
• Why they buy: Their sell-electricity revenue drops without explanation two months after the fact, and the utility’s answer to “why is so much being deducted?” is “consider paying to go online.” They want to know whether the deduction is right, and how much the plant is actually losing per year, before deciding whether to spend ¥100k–500k on onlining or to sell the asset.
• Rough TAM reasoning: JPEA’s own survey of member companies holding old-rule solar in Kyushu (14 companies, 241MW, Dec 2021) found 97% of sub-500kW plants had not gone online and had no plan to (only 2% done); outside Kyushu 81% were non-responsive. That non-onlined majority is precisely the proxy-control population. Japan’s low-voltage FIT fleet runs to hundreds of thousands of plants; even a low-tens-of-thousands serviceable slice of old-rule offline plants in the three affected areas is far more than needed for a $1–5M business.
• Why now for them: FY2026 is the record curtailment year and TEPCO joined in March 2026. A deduction that was a rounding error in 2022 is now large enough to change whether the asset clears its loan.

5. Product sketch (MVP)

• Connect one or more plants by plant ID, area, rule class (old-rule/new-rule, online/offline) and contract price
• Automatic ingest of each TSO’s published monthly settlement ratios and curtailment instruction records — no utility cooperation required
• Deduction recomputation: for each settlement month, an independent calculation of what the proxy-control deduction should be, against what the purchase statement actually took
• Discrepancy flag when the statement’s minus line doesn’t match the recomputed figure, with the underlying ratio and kWh shown
• “Why your payout dropped” explainer for the month — separates curtailment deduction from an actual generation fault, so owners stop calling O&M about a plant that isn’t broken
• Curtailment evidence pack — a dated, exportable PDF of curtailment instructions and deemed hours, since the utility refuses to issue one (useful for insurance and for sale diligence)
• Online-conversion payback calculator — your actual annual proxy deduction vs the quoted ¥100k–500k onlining cost, so the “should I convert?” decision has a real number
• Multi-plant portfolio view for O&M firms and owners holding several assets

6. AI angle — what’s load-bearing

Honest answer: AI is a supporting actor here, not the engine. The core is a deterministic recomputation of a published formula against published ratios — that’s arithmetic and parsing, and it should be arithmetic, because a customer disputing money needs a number they can defend, not a model’s opinion.

Where AI genuinely earns its place:

• Document extraction. Every one of the ten TSOs publishes settlement ratios in its own PDF/Excel layout, changing shape year to year, in Japanese. Vision-model extraction into a normalised table is the difference between a solo builder covering three areas and covering all of them. This is the load-bearing use.
• Statement parsing. Owners receive purchase statements in inconsistent formats (paper scans, PDFs, utility portal exports). Extracting the minus line from a photographed statement is a real vision task.
• Explanation generation. Turning “ratio −0.0412 applied to 4,180kWh in the June settlement month” into plain Japanese an individual investor understands.

If you removed the AI you’d still have a product — a slower one that supports fewer areas and makes the owner type numbers in by hand. I’d rather state that plainly than pretend a deduction calculator is an AI play. The moat here is domain modelling, not model choice.

7. Localization angle

This is Japan-only by construction, and that’s the point. The 代理制御 settlement fiction exists nowhere else — it’s a specific 2022 Japanese regulatory invention to bring un-onlined old-rule plants into the curtailment burden without forcing them to install hardware. The product is Japanese-language by default, priced in yen, and sold to an owner class (individual/small-corporate FIT investors) that is unusually large in Japan because the 2012 FIT boom pulled in retail investors at scale.

Practical implications: monthly per-plant pricing in the ¥1,000–3,000 range fits an asset earning a few hundred thousand yen a year; Japanese-language support is mandatory, not optional; and distribution runs through the 販売施工店 / O&M channel and the very active Japanese solar-investment blog and 発電所売買 (plant resale) ecosystem rather than through English SaaS channels.

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

• Pricing: ¥1,500/month per plant (~$10) for owners, with volume tiers; ¥30,000–80,000/month portfolio plans for O&M and asset-management firms managing 30–200 plants. One-off ¥15,000 evidence pack for plant sale/insurance diligence.
• ACV: ¥18,000/plant/year ($120) for a single-plant owner; realistic blended ACV ¥120,000 ($800) because most buyers hold several plants, and O&M accounts pull the average up hard.
• Rough math to $1M ARR: ~1,250 accounts at ~$800 blended ACV. Concretely: ~1,000 small owner accounts (avg 3 plants) plus ~150 O&M/asset-manager accounts. Against a non-onlined old-rule population in the tens of thousands, that’s low-single-digit penetration.
• Rough math to $5M ARR: needs two things true — coverage of all ten TSO areas (so the product sells nationally as curtailment spreads past Kyushu/Tohoku/Hokkaido into TEPCO and beyond), and the O&M channel becoming the primary motion, with a few hundred firms on portfolio plans reselling it as part of their monitoring package. FY2027+ curtailment growth is the tailwind.
• Expansion path: per-plant seats grow as owners add assets; upsell from reconciliation into the onlining decision (payback analysis, then referral fees to 施工店 doing the conversion); then plant-sale diligence — curtailment history is a material valuation input in Japan’s active secondary market for FIT plants, and we hold it in structured form.

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

• Free public “your area’s deduction” calculator, seeded on the three areas that hurt. A page where an owner enters area + rule class + contract price + capacity and gets their estimated FY2026 proxy deduction. This is the single highest-intent query in the market right now — owners searching 出力制御 精算比率 / 代理制御 マイナス after their payout drops. Convert to paid for the per-month actuals.
• Target the O&M firms the utility explicitly refuses. There is a finite, listable set of Japanese solar O&M companies (the trade comparison sites enumerate them). Every one of them has been told by the grid operator “we cannot respond to individual curtailment record enquiries — the owner must log in.” Cold email with a worked example on a real plant in their area: you get portfolio-level curtailment answers without chasing client logins. These are multi-plant accounts on day one.
• The “my plant isn’t broken” wedge. Regional maintenance firms publicly report owners contacting them convinced of equipment damage when the cause was the settlement ratio. Partner with 施工店/O&M to bundle SeisanCheck as the first triage step — it saves them truck rolls, which is a cost argument, not a favour.
• Japanese solar-investment media and 発電所売買 brokers. タイナビ発電所 and similar publish heavily-read curtailment explainers and run plant resale marketplaces. A co-published “FY2026 deduction by area” data piece, using our own recomputation, puts us in front of exactly the owner segment weighing sell-vs-convert.
• Guest calculations in the blog/YouTube layer. Japan’s FIT-investor content scene is dense and specific. Offer creators a free portfolio audit and let them show the recomputed number on their own plants.

10. Build complexity — justification

Medium. No hardware, no utility integration, no approvals — everything needed is published. The real work is (a) ingesting and normalising ten TSOs’ differently-shaped Japanese ratio publications and keeping up as they change annually, (b) implementing the deemed-hours/settlement-ratio calculation correctly per rule class and voltage class, which is genuinely fiddly domain modelling, and (c) statement parsing from messy owner-supplied documents. A pair could ship a credible three-area (Kyushu/Tohoku/Hokkaido) v1 in 10–14 weeks; national coverage is a further couple of months of grind. The domain research is the bottleneck, not the engineering — and a Japanese-reading founder or advisor is effectively mandatory.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Recomputing published ratios against the owner’s own statement. No scraping of gated systems; owner-supplied data only.
Ethical — no harm / dark patterns	✅	Gives the paying party visibility into a deduction the counterparty computes. Strictly pro-transparency.
Market exists (evidence above)	✅	Record FY2026 curtailment, utility FAQ documenting owner complaints about the size of deductions, JPEA data on the non-onlined majority.
1–5 person team can build this	✅	Two people, ~3 months to a three-area v1.
Launchable with <$50K / ₹40L	✅	Data is public. Main cost is Japanese-language domain time.

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Real recurring money, and the utility’s own FAQ records owners complaining the deduction is substantial. Held back from higher because it’s a quarterly-ish sting on a passive asset, not a daily operational fire — many owners have tolerated it for three years.
Demand evidence	15	12/15	Strong indirect evidence: primary-source utility FAQ answering these exact complaints, JPEA survey on the non-onlined population, record curtailment forecast, maintenance firms reporting owner enquiries. Docked because I found no direct verbatim owner quote demanding this product — the pain is documented by intermediaries, not by owners in their own words.
Build feasibility	15	11/15	All inputs public, but ten TSOs × annually-changing Japanese PDFs plus per-rule-class settlement logic is real grind.
Distribution clarity	15	12/15	The O&M list is finite, enumerable, and has been explicitly refused service by the utility — an unusually clean cold-outreach story. High-intent search wedge is real. Docked because Japanese SMB cold email converts slowly and trust-building is slower than the West.
Revenue mechanics	15	11/15	Per-plant pricing is small; the model leans on multi-plant and O&M accounts to make ACV work. Path to $1M is credible, $5M needs the O&M channel to carry it.
Time to first revenue	10	8/10	Sellable as soon as one area is accurate; owners can be charged immediately since the value is a number they can check against a statement they already hold.
Defensibility	10	6/10	Soft moat: accumulated normalisation of ten TSOs’ formats, per-plant curtailment history that compounds and becomes valuation data, and hard-won rule-class logic. Copyable by a determined Japanese competitor in ~6 months; the incumbents’ disinterest is the real protection, and that’s not permanent.
Total	100	76/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

Japanese-language fluency and willingness to read METI/TSO documents is non-negotiable. This is not a market you can serve through translation.

Key assumptions to validate (3–5)

1. Assumption: The recomputed deduction differs from the utility’s applied figure often enough, or is opaque enough, that owners will pay just to see it verified. How to test: Recompute 20 real plants’ FY2026 settlements from owner-supplied statements across Kyushu/Tohoku; measure how many show a discrepancy or a figure the owner couldn’t previously explain.
2. Assumption: O&M firms will pay for portfolio-level curtailment visibility rather than continuing to ask clients to log in. How to test: Cold-call 30 Japanese solar O&M companies; offer a free 10-plant audit; count how many hand over plant lists.
3. Assumption: Per-plant deduction is large enough to justify ¥1,500/month. How to test: Compute actual FY2026 annual deduction for a representative 50kW old-rule offline plant in Kyushu. If the annual deduction isn’t comfortably into six figures of yen, pricing must drop or shift entirely to portfolio plans.
4. Assumption: Owners can obtain and share their purchase statements without friction. How to test: Ask 15 owners for a statement; measure how many produce one within a week.

Risk flags

1. Regulatory obsolescence: The whole product exists because of the proxy-control fiction. If METI reforms deemed-hours settlement, or mandates/subsidises onlining for the sub-500kW class, the core reconciliation shrinks. Mitigation: the onlining-payback and sale-diligence surfaces survive that change; watch the 次世代電力系統WG agenda.
2. Incumbent absorption: エコめがね already writes explainer content about proxy control and owns the sub-50kW monitoring relationship. If NTT Smile Energy decides to add a settlement line to its monthly report, distribution beats us instantly. This is the single biggest risk.
3. Population decay: Old-rule plants are a fixed, non-growing cohort — FIT contracts expire and owners convert or sell. The addressable base slowly shrinks, which caps the terminal size and argues for expanding into new-rule/FIP curtailment economics early.
4. Evidence without remedy: There is no published formal dispute channel for contesting a settlement ratio. If the answer to a discovered discrepancy is “the utility won’t reopen it,” the product’s value narrows to explanation and decision support rather than recovery. This is why the pitch is deliberately “know your number and decide,” not “we get your money back.”

14. Structured verdict

Score:                  76/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Japanese-reading technical founder, ideally with energy/FIT asset background or a solar O&M advisor
Time to revenue:        10–14 weeks
Capital to launch:      ¥1.5–3M ($10–20K)
Top 3 assumptions to validate first:
  1. Recompute 20 real FY2026 settlements — do discrepancies or unexplainable figures show up at a meaningful rate?
  2. Cold-call 30 O&M firms — will the party the utility refuses pay for portfolio curtailment visibility?
  3. Size the actual annual deduction on a representative 50kW Kyushu old-rule offline plant — does ¥1,500/mo clear the value bar?
Kill criteria:
  - Abandon if fewer than 3 of 20 recomputed plants show a discrepancy or a figure the owner couldn't previously explain
  - Abandon if fewer than 3 of 30 O&M firms will share a plant list for a free audit
  - Abandon if NTT Smile Energy ships settlement reconciliation inside エコめがね before v1
  - Abandon if METI publishes a timetable to abolish deemed-hours proxy settlement for sub-500kW plants

15. Next step — 1-week validation sprint

• Day 1–2: Pull FY2026 published settlement ratios for Kyushu, Tohoku and Shikoku. Implement the deemed-hours calculation for old-rule offline sub-500kW by hand in a spreadsheet. Compute the annual deduction for three representative plant sizes (10kW, 50kW, 250kW). This alone answers whether the money is big enough to sell against.
• Day 3–4: Recruit 10 owners through Japanese solar-investment forums and plant-resale communities. Offer a free recomputation in exchange for their purchase statement. Count how many can produce a statement, and how many of the resulting numbers they say they could not previously explain.
• Day 5: Call 15 O&M firms with a worked example from their own grid area. Ask directly for a plant list for a free audit.

Falsifiable outcome: Go only if (a) the modelled annual deduction on a 50kW Kyushu plant exceeds ¥100,000, (b) ≥5 of 10 owners say the recomputed figure told them something their statement did not, and (c) ≥3 of 15 O&M firms hand over a plant list. Anything less and the pain is real but not purchasable, and this stays a spreadsheet.