KyogiMinutes

1. One-liner

Turns every supplier price talk into a dated record that shows you negotiated, not dictated.

2. Trend signal — why now?

Two dated changes land on the same buyer eight weeks apart, and they pull in opposite directions.

1 January 2026 — the law changed what counts as a violation. Japan’s 下請法 (Subcontract Act) was renamed and rewritten as the 中小受託取引適正化法 (取適法). The headline addition: unilaterally deciding a price without responding to a supplier’s request to negotiate is now itself prohibited. The JFTC’s own leaflet is explicit that this covers not just outright refusal but “協議の求めを無視したり、協議を繰り返し先延ばしにするなど、協議を困難にさせた場合” — ignoring the request, or repeatedly stalling it, counts (JFTC, 取適法 leaflet; Cabinet Office, gov-online, Nov 2025). Practitioner commentary reads the shift the same way: “問いの重心が、いくらで買ったかから、どう決めたかへ移っている” — the question moved from what you paid to how you decided (Nexaflow). Note the law also added 運送委託 (transport commissioning) as a fifth covered transaction type, dragging logistics buyers into scope who were previously outside it.

1 October 2026 — the tax gave every buyer a reason to reopen prices. Under the invoice system’s transitional relief, a buyer purchasing from a non-registered (免税) supplier could deduct 80% of the notional input tax. From 1 October 2026 that drops to 70%, then 50% in 2028 and 30% in 2030 (小谷野税理士法人; 道濟会計事務所). The 2割特例 also terminates for corporations with the tax period containing 30 September 2026 (FOLIO). Real cost lands on the buyer’s P&L on a fixed date.

The squeeze. The buyer now has a documented cost increase and a legal obligation to consult before acting on it. The JFTC has already published what the wrong move looks like: unilaterally notifying a tax-exempt supplier that the consumption-tax equivalent will be cut from their price. It issued 36 注意 (cautions) on invoice-related conduct by end-September 2023 (三井住友海上 MSコンパス). The JFTC Q&A sets the standard for a lawful reduction — agreement reached after substantive discussion of the supplier’s own input tax burden, with good-faith consultation evidenced “書面、電子メール等”; the key phrase is substantive discussion, not mere notification (JFTC Q&A).

Enforcement has teeth and names. FY2025: 39 勧告 (recommendations, published with the company name), 8,261 指導, and ¥2.557 billion returned by 177 commissioning businesses to 5,165 SME suppliers (JFTC, 10 June 2026). Named precedents for 買いたたき include BigMotor (Mar 2024) and KADOKAWA (Nov 2024) (Cloudsign). METI’s own follow-up survey found only 31.5% of price negotiations were initiated by the buyer side, down ~3 points (中小企業庁, 26 June 2026) — buyers are structurally bad at starting these conversations, which is exactly the behaviour the new law punishes.

Provenance:

• Signal 1 (demand): 取適法 effective 1 Jan 2026 makes unilateral price-setting without consultation a violation, including ignoring or stalling a request; FY2025 enforcement = 39 勧告, 8,261 指導, ¥2.557bn returned by 177 buyers — https://www.jftc.go.jp/file/toriteki_leaflet.pdf + https://www.jftc.go.jp/houdou/pressrelease/2026/jun/260610.html — 2026-06-10
• Signal 2 (feasibility): JFTC accepts 書面・電子メール等 as consultation evidence, and defines the bar as reconstructable who/when/on-what-basis timeline — a records problem, not a legal-judgement problem, now tractable with Japanese-language ASR + summarisation — https://www.jftc.go.jp/dk/guideline/unyoukijun/invoice_qanda.html — 2026-08-29
• Signal 3 (economic): input-tax deduction on non-registered suppliers drops 80%→70% on 1 Oct 2026 and 2割特例 ends for corporations, putting a dated, quantified cost increase on every buyer with 免税 suppliers — https://koyano-cpa.gr.jp/nobiyo-kaikei/column/8872/ — 2026-08-29 Category: Regulatory arbitrage

3. The opportunity

The accounting vendors already won the half of this problem that is arithmetic. freee ships an automatic update so that purchases dated on or after 1 October 2026 compute at 70% (freee), and it checks registration numbers against the NTA Web-API automatically (freee press release). Money Forward and Yayoi do the equivalent. Nobody should build a deduction calculator.

The unbuilt half is the record of how the price was decided. That artefact lives nowhere: it’s in a sales rep’s inbox, a purchasing manager’s memory, a meeting that produced no minutes, and a spreadsheet that shows the new price but not the reasoning or the sequence. When the JFTC asks — and in FY2025 it asked 8,261 times at 指導 level alone — the company reconstructs from recollection, and recollection loses.

The current market answer is free Excel. bizocean distributes a 価格交渉記録簿 template; METI publishes a 価格交渉ハンドブック (last revised January 2026) and runs 価格転嫁サポート窓口 at all 47 prefectural よろず支援拠点 (中小企業庁 support tools). That’s the incumbent: a downloadable form and a government help desk. It fails for a specific structural reason — a template records the conclusion after the fact, but the violation the new law defines is about the process and its timing. An Excel row filled in the week the auditor calls proves nothing. What proves something is a contemporaneous, timestamped trail showing the supplier’s request arrived, the buyer responded within a sane interval, materials were exchanged, and the number moved because of a stated reason.

This is the scoring-vs-proving gap in its cleanest form: every vendor sells the number, nobody sells the evidence that the number was arrived at lawfully.

4. Target market

Primary customer: The 購買 (procurement) or 管理部 manager at a Japanese company with 50–500 employees in manufacturing, construction, or logistics/transport, who buys from between 20 and 300 suppliers, of whom a meaningful slice are 免税事業者 — typically 一人親方 (independent tradespeople), owner-drivers, and small workshops. Annual revenue roughly ¥1bn–¥30bn. This is the band that is unambiguously a 委託事業者 under 取適法 but has no in-house legal team and no 法務部 headcount to run a consultation programme.

Why they buy: Two fears, one date. On 1 October their input-tax deduction shrinks and someone in finance says “renegotiate with the non-registered suppliers.” The procurement manager knows — because their industry association has told them all year — that doing this the obvious way (send a letter, cut the tax equivalent) is the exact fact pattern the JFTC publishes cautions about. They need to have the conversation and be able to show they had it. Construction is the acute case: the sector’s multi-tier subcontracting means a huge count of 免税 一人親方, and commentary notes that in construction “対応の巧拙がコストや受注機会に直結” — handling this well or badly directly hits cost and order flow (ケンテク).

Rough TAM reasoning: METI’s FY2025 price pass-through special survey alone covered over 120,000 businesses, and the logistics slice surveyed 30,000 shippers plus 40,000 carriers (JFTC, June 2026). The 50–500 employee band across manufacturing, construction and transport in Japan runs to the low hundreds of thousands of firms. I don’t need a big share: 1,000 customers is a rounding error on that base and is a ¥1.4bn business at my pricing.

Why now for them: The 1 October date forces the conversation. The January law change means the conversation is now itself regulated. Before 2026, sloppy price-setting was a commercial matter; now it is a named-and-published enforcement matter.

5. Product sketch (MVP)

• Supplier registry with 免税 flagging — imports the supplier master from freee / Money Forward / a CSV, checks each registration number against the NTA API, and flags which counterparties are non-registered and therefore in the 1 October exposure set, with the yen impact of the 80%→70% step per supplier.
• Consultation clock — the moment a supplier requests a price discussion (by email, form, or logged phone call), a timer starts and the responsible buyer is assigned. Stalling is the violation, so the product’s core UI is a list of open requests aged in days.
• Meeting-to-minutes — record the negotiation call or upload the recording; the product produces a Japanese-language 協議記録 with attendees, date, the materials cited, the positions taken, and the conclusion. This is the piece the template can’t do.
• Cost-basis attachment — attach the supplier’s stated input costs, labour cost changes, and the buyer’s own tax-impact figure to the record, so the file shows the price reflects the supplier’s actual burden rather than a flat pass-down.
• Auto-drafted response letters — generates the written explanation of a price change in JFTC-safe language, flagging phrasings that match published caution patterns (e.g. conditioning continued trade on registration).
• Reconstruction export — one click produces a per-supplier, per-negotiation chronological PDF an auditor or the JFTC can read: who, when, on what basis, what was concluded.
• Risk dashboard — which suppliers have had a price cut without a matching consultation record. That list is the company’s actual exposure, and no existing system can produce it.

6. AI angle — what’s load-bearing

Remove the AI and this collapses back into the free Excel template that already exists and already fails. The load-bearing work is turning unstructured Japanese negotiation talk into a structured, defensible record — a recorded call or a messy email thread in, a minutes document with attendees, cited materials, stated positions and conclusion out. Japanese ASR plus summarisation is the only thing that makes contemporaneous recording cheap enough that a procurement manager will actually do it for 200 suppliers instead of writing up three and forgetting the rest. Second load-bearing job: classifying draft language against the JFTC’s published caution patterns, so the tool catches “unless you register we’ll cut your price” before it’s sent rather than after it’s been reported.

7. Localization angle

This is Japan-only by construction and that’s the point. The statute (取適法), the tax schedule (the 80/70/50/30 taper), the enforcement body (公正取引委員会), the document conventions (稟議, 議事録), and the language are all Japanese. A global procurement suite cannot serve this without rebuilding around a Japanese legal artefact, and the segment — 50–500 employees, no legal department — is below the altitude at which SAP Ariba or Coupa sell. Pricing must be yen-native and land in the ¥30,000–150,000/month band where Japanese SMB SaaS actually transacts, not translated US pricing. Distribution is likewise local: 商工会議所, industry associations, and the 税理士 who already sits inside these companies’ books.

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

• Pricing: ¥48,000/month ($320) for up to 50 suppliers; ¥98,000/month ($650) up to 200; ¥180,000/month (~$1,200) unlimited plus multi-site. Annual prepay standard in Japan — expect most contracts booked yearly.
• ACV: ¥1,050,000 ($7,000) blended.
• Rough math to $1M ARR: 145 customers at blended ACV. At the ¥98,000 mid-tier that’s ~128 customers — a single vertical (construction subcontracting in Kanto) can carry that.
• Rough math to $5M ARR: ~700 customers, which requires winning through channel rather than direct — 税理士法人 and industry-association reseller arrangements — plus expansion into the adjacent 労務費転嫁 (labour cost pass-through) records that the same law governs.
• Expansion path: supplier count tiers first, then seats as 購買 + 経理 + 法務 all get logins, then the 2028 step-down (70%→50%) which re-triggers the entire renegotiation cycle on a fresh date. The taper running to 2031 gives this product a five-year cadence of forced re-engagement, which is unusually good retention structure for a compliance tool.

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

• The 税理士 channel is the whole game. Japanese SMBs of this size have a 顧問税理士 in the building monthly, and that accountant is already fielding “what happens on 1 October” questions. Recruit 30 税理士法人 with a revenue-share; each has 50–200 corporate clients and can identify which have 免税 suppliers from the books they already keep. Ten productive firms converting five clients each = 50 customers.
• Ride the METI calendar. 価格交渉促進月間 runs in March and September with government publicity, follow-up surveys naming good and bad buyers, and prefectural 窓口 activity. Publish a free “1 October exposure calculator” — upload your supplier list, get the yen impact and the count of suppliers you must consult — timed to the September month. That’s a lead magnet with a genuine number attached, and it self-qualifies: anyone with a non-zero result is a prospect.
• Industry associations, construction first. 建設業協会 chapters and 専門工事業 associations run member seminars on exactly this and are actively looking for speakers who can explain the 一人親方 problem. Sponsor/speak at 15 chapter events; these rooms are 50–150 owner-operators of precisely the target size.
• Named-risk outbound. The JFTC publishes 勧告 with company names, and METI’s follow-up surveys publish buyer-side negotiation records by industry. Firms in poorly-rated industries are the warmest possible list — approach with “your sector scored X on buyer-initiated negotiation; here’s what the auditor will ask for.”
• The 労務費指針 hook. The Cabinet’s labour-cost pass-through guidelines expect buyers to proactively open negotiations. Same buyer, same record, adjacent trigger — one piece of content covers both and doubles the search surface.

10. Build complexity — justification

Medium. The heavy lifting is Japanese-language ASR and summarisation into a fixed document shape — off-the-shelf models handle this well now, and the output is a structured form rather than free prose, which keeps quality controllable. The NTA registration-number API is public and simple. The genuine work is the accounting-system integrations (freee and Money Forward both have documented APIs, but supplier-master mapping is fiddly) and getting the 協議記録 document format right enough that a 税理士 will vouch for it — that requires a real advisor, not guesswork. Call it 12–16 weeks to a v1 a design partner can run, with a Japanese-speaking founder or co-founder non-negotiable.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Record-keeping aid; makes no legal determination. Must not be marketed as legal advice — 弁護士法 boundary is real but easily respected by staying on documentation.
Ethical — no harm / dark patterns	✅	The product’s incentive is aligned with the law’s intent: it pushes buyers to actually consult and to price on the supplier’s real costs. It gets worse, not better, if used to fake consultations — the trail is contemporaneous and timestamped.
Market exists (evidence above)	✅	39 勧告 / 8,261 指導 / ¥2.557bn returned in FY2025; 120,000+ businesses surveyed; free templates in circulation proving felt need.
1–5 person team can build this	✅	12–16 weeks, 2 people plus a 税理士 advisor.
Launchable with <$50K / ₹40L	✅	Model API costs, two integrations, seminar sponsorships. Well under.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Dated, quantified, and enforced with published company names. Docked because the pain is periodic (negotiation cycles) rather than daily, and many buyers will rationally gamble on not being audited.
Demand evidence	15	12/15	Strong institutional evidence — enforcement stats, government surveys, free templates in circulation, association seminars. Docked: I found no verbatim buyer-side complaints, and the existence of free Excel means demand is real but not yet priced.
Build feasibility	15	11/15	Japanese ASR + structured summarisation is off-the-shelf; freee/MF integrations and document-format credibility are the real work. 12–16 weeks.
Distribution clarity	15	12/15	税理士 channel is specific and warm, METI’s March/September calendar gives timed hooks, construction associations are addressable. Docked because the 税理士 channel is slow to activate and Japanese SMB sales cycles are not fast.
Revenue mechanics	15	11/15	¥48–180K/month is credible for Japanese SMB compliance SaaS and 145 customers reaches $1M. Docked: free government templates cap willingness-to-pay, and this is a budget line that didn’t exist last year.
Time to first revenue	10	7/10	The 1 October date creates genuine urgency and design partners are gettable pre-build, but Japanese procurement cycles and annual-contract norms mean 8–12 weeks to real money, not 4.
Defensibility	10	5/10	Accumulating negotiation history creates switching cost, and 税理士 relationships compound. But freee or Money Forward could bolt a records module onto software these companies already run — that’s the central risk.
Total	100	74/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

Japanese language fluency is a hard gate, not a preference. A 税理士 or ex-procurement advisor on the cap table is close to mandatory for document credibility.

Key assumptions to validate

1. Assumption: Procurement managers at 50–500 employee firms perceive consultation-record risk as their problem, not the 税理士’s or the lawyer’s. How to test: 20 interviews via 商工会議所 introductions; ask who owns the record today and watch whether they name a person or shrug.
2. Assumption: A company will pay ¥48,000+/month for a record that free Excel nominally produces. How to test: show the reconstruction export next to a filled-in bizocean template to 15 prospects and ask which one they’d hand an auditor — then ask for a signed LOI at price.
3. Assumption: 税理士法人 will actively recommend rather than passively tolerate. How to test: pitch 10 firms; a real yes is them naming three specific clients on the call.
4. Assumption: The 1 October step-down actually triggers renegotiation waves rather than buyers quietly absorbing 10 points. How to test: ask 20 firms directly whether a renegotiation programme is scheduled and who signed off on it.

Risk flags

1. Incumbent absorption: freee and Money Forward sit inside these companies already and own the supplier master. If either ships a consultation-record module, the standalone case weakens fast. Mitigation is depth in the negotiation artefact and the 税理士 relationship, not feature count.
2. Free-substitute ceiling: METI and bizocean give the artefact away in template form. The pitch must be about the contemporaneous trail, not the document — if prospects hear “nicer Excel,” pricing collapses. This is the single likeliest reason the idea underperforms.
3. Enforcement is probabilistic: 8,261 指導 is large in absolute terms but small against hundreds of thousands of firms. Many buyers will correctly calculate they won’t get caught. The product sells better to the risk-averse and the recently-audited than to the median firm.
4. Language and market access: a non-Japanese-speaking founder cannot run this. That narrows the builder pool severely — which is also part of why the gap is still open.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Japanese-speaking technical founder with a 税理士 or procurement-compliance advisor
Time to revenue:        8–12 weeks from launch
Capital to launch:      ¥4–6M ($27–40K)
Top 3 assumptions to validate first:
  1. Procurement managers own the consultation-record risk personally — 20 interviews via 商工会議所, listen for a named owner
  2. ¥48,000/month clears against a free template — side-by-side demo to 15 prospects, ask for LOI at price
  3. 税理士法人 will actively refer — pitch 10 firms, count how many name specific clients on the call
Kill criteria:
  - Abandon if fewer than 3 of 15 prospects sign an LOI at ¥48,000/month after seeing the reconstruction export
  - Abandon if freee or Money Forward ships a consultation-record module before v1 launches
  - Abandon if 20 interviewed firms show no scheduled renegotiation programme for the 1 October step-down

15. Next step — 1-week validation sprint

• Day 1–2: Build the free exposure calculator only — upload a supplier list, hit the NTA API, return the count of non-registered suppliers and the yen cost of the 80%→70% step. No product, just the number. Ship it in Japanese.
• Day 3–4: Push it through three 商工会議所 chapters and two construction association mailing lists ahead of the September 価格交渉促進月間. Every submission is a qualified lead with a self-reported exposure figure attached.
• Day 5: Call the 20 highest-exposure submissions. Show the reconstruction export mock next to a filled-in bizocean template and ask for an LOI at ¥48,000/month.

Falsifiable outcome: fewer than 3 LOIs from 15 real conversations, or a majority answering “our 税理士 handles that” when asked who owns the record — either kills it. The exposure calculator also independently tests assumption 4: if firms with large computed exposure have no renegotiation programme scheduled, the 1 October trigger isn’t real and the whole thesis fails.