EstornoEscudo — MED defense desk for Brazilian high-ticket Pix sellers

1. One-liner

Auto-builds the good-faith dossier a Brazilian merchant needs to unfreeze a wrongful Pix MED block in time.

2. Trend signal — why now?

MED 2.0 (Mecanismo Especial de Devolução do Pix) became mandatory for every Brazilian financial institution on 2 February 2026, reinforced by Banco Central Resolution nº 551 in March 2026. The contestation button now sits inside every banking app. A defrauded sender files; within 72h the receiver’s balance is blocked (up to the contested value) for as long as 80 days while the bank runs an automated analysis. There is no formal in-system channel for the receiver to file a defense — analysis is internal to the banks.

The predictable side effect: legitimate merchants get frozen. Press already covering it — golpistas weaponizing MED against wholesalers on Rua 25 de Março, São Paulo. Documented merchant losses of R$650 to R$6.000, not recovered even after a police report. A phone-repair shop did a R$650 job; the customer contested claiming “service quality” and the bank froze the money — even though MED legally cannot be used for commercial disagreement or buyer’s remorse. The bank automation doesn’t make that distinction; the merchant has to prove good faith, fast, with organized evidence. Banks confirm a recebedor de boa-fé gets unblocked quickly — if they can demonstrate it inside the ~7–14 day window. Almost no merchant has that evidence assembled.

Provenance:

• Signal 1 (demand): Golpistas exploiting MED against 25 de Março lojistas; merchant losses R$650–R$6k unrecovered; phone-repair “service quality” contestation froze a legitimate R$650 sale — terra.com.br / web search 2026-05-17 — observed 2026-05-17
• Signal 2 (feasibility): MED 2.0 mandatory 2 Feb 2026 + Resolution 551 (Mar 2026); standardized nationwide dispute flow; “no formal channel for companies to present prior defense” — pagbrasil.com / bcb.gov.br MED 2.0 circuito / agenciabrasil 2026-02 — observed 2026-05-17
• Signal 3 (economic): 97% of small Brazilian firms use Pix; 170M+ users; B2B Pix R$1.5T in Dec 2025 (up from R$1T Dec 2024); Nubank alone 3M PJ accounts (76% MEI); PayPal added Pix for SMBs Apr 2026 — paymentscmi.com / pymnts.com / paypal newsroom — observed 2026-05-17 Category: Regulatory arbitrage

3. The opportunity

MED 2.0 created a structural asymmetry overnight: the sender gets a one-tap, in-app, bank-pushed contestation flow. The receiver — the merchant who shipped goods or did the work — gets a frozen balance and a vague “respond to your bank” with no standard interface, no checklist, and a clock measured in days. Banks process MED analysis with automation; they reward a clean, fast, structured boa-fé response and default to the sender otherwise (Brazil merchant fraud-chargeback win rate is ~17%).

This isn’t a payments product and it isn’t an antifraud product — those incumbents (PagBrasil, ClearSale, Konduto) sell acceptance-time risk scoring to mid/large e-commerce. Nobody owns the post-block defense workflow for the small high-ticket seller. The incumbent here is effectively “a WhatsApp thread, a shoebox of prints, and the merchant’s accountant scrambling.” An AI-first tool that captures defensible evidence at sale time and assembles the exact dossier the bank wants — on the deadline, in the bank’s expected format — does in 2 minutes what currently doesn’t get done at all.

4. Target market

• Primary customer: Brazilian SMB sellers who take high-ticket Pix into a PJ account — Rua 25 de Março / Brás wholesalers, electronics and phone-repair shops, furniture/appliance dealers, and Instagram/WhatsApp social-commerce sellers. 1–5 staff, R$30k–R$500k/month revenue, ticket sizes R$300–R$15k where a single block actually hurts.
• Why they buy (their words): “O cliente contestou, o banco bloqueou, e eu já tinha entregado a mercadoria.” “Não posso abrir um MED porque tecnicamente não fui fraudado — fiquei sem o dinheiro e sem o produto.” “Prejuízo de R$6 mil, registrei BO, não recuperei nada.”
• Rough TAM reasoning: 97% of ~20M+ Brazilian small firms use Pix; even narrowing to high-ticket PJ sellers exposed to MED, that’s well into the millions. Nubank alone has 3M PJ accounts. A wedge of 5,000 paying merchants is a rounding error on the market and a >R$5M ARR business.
• Why now for them: Before Feb 2026 MED was bank-discretionary and patchy, so the pain was anecdotal. Mandatory MED 2.0 made the contestation button ubiquitous and the freeze automatic — the pain went from “rare unlucky event” to “recurring cost of doing high-ticket Pix business,” and it’s in the news.

5. Product sketch (MVP)

• Sale capture: at point of sale (web form, WhatsApp bot, or maquininha-adjacent flow) records the transaction, links a nota fiscal / pedido, customer identity, and a timestamped good-faith trail.
• Evidence locker: auto-pulls and stores the defensible artifacts banks ask for — order, NF-e, delivery/collection proof, signature/photo, customer chat transcript, device/IP — per transaction.
• Block detector: merchant forwards the bank’s MED notification (email/print/screenshot) or flags a frozen balance; system parses it, identifies the disputed Pix, and starts the deadline countdown.
• Dossier builder: assembles a one-click contestação de boa-fé packet — structured PDF + cover letter citing that MED is barred for desacordo comercial (BC norm) — formatted for the receiving bank’s process.
• Deadline cockpit-free tracker: plain countdown + escalation reminders (bank ouvidoria, then Banco Central RDR / Procon templates) if the bank stalls.
• Pattern flags: warns when an incoming Pix looks MED-prone (new payer, high ticket, scheduled transfer) so the merchant can hold delivery past the 72h window.
• Portuguese-first throughout, MEI/PJ-aware, mobile-first.

6. AI angle — what’s load-bearing

AI does the work humans won’t: (1) parsing heterogeneous bank MED notifications and merchant document dumps (NF-e XML/PDF, WhatsApp screenshots, photos) into a normalized case; (2) drafting the bank-specific good-faith argument, correctly invoking that MED excludes commercial disagreement and remorse, tailored to whether the case is alleged fraud vs. disguised buyer’s remorse; (3) classifying incoming Pix risk to trigger a delivery-hold nudge. Remove the AI and you’re left with a folder app — the merchant still has to read the notice, figure out the argument, and write the letter, which is exactly the part they don’t do today. The drafting + parsing is the product.

7. Localization angle

This is Brazil-only by construction — it’s built on Pix, MED 2.0, Banco Central resolutions, NF-e, MEI/PJ structures, Procon and the BC’s RDR escalation path, in Portuguese. That’s the moat, not a limitation: a generic global chargeback tool cannot touch this because MED is a Pix-native, regulation-specific mechanism with no analog. Pricing must be R$-native (a R$79–R$249/mo tier works where a US$49 tool would be mispriced). Distribution rides Brazilian rails — WhatsApp, contador networks, marketplace/wholesaler associations.

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

• Pricing: tiered SaaS. Starter R$79/mo (capture + locker, low volume); Pro R$199/mo (block detection, dossier builder, deadline tracking); plus a R$149 per assisted dossier success-priced add-on for non-subscribers’ acute first incident (land-and-expand hook).
• ACV: ~R$2,000 (≈US$370) blended on Pro after expansion.
• Math to ~$1M ARR (R$5.5M): ~2,300 merchants on Pro at R$199/mo ≈ R$5.5M/yr. Reachable within the 25 de Março + electronics-repair + social-commerce wedge alone.
• Math to ~$5M ARR (R$27M): ~11,000 Pro merchants, or 6,000 Pro + a per-dossier transactional layer + a B2B2C deal with one mid-tier bank/acquirer or accounting platform (ContaAzul-style) reselling it. Plausible but assumption-heavy — depends on a channel partnership.
• Expansion path: seats for multi-store; per-dossier overage; a “Pix risk hold” premium; eventual white-label to acquirers who’d rather offer merchant defense than absorb churn.

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

• 25 de Março / Brás physical canvass: these districts are geographically dense, already in the press as MED victims, and talk to each other. Walk the street with a one-page demo and a “your last frozen Pix, recovered” hook. 30–50 closes from a focused two-week push is realistic — acute, named pain.
• Phone-repair & electronics shop lists: scrape Google Maps + assistência técnica directories in SP/RJ/MG, send a personalized WhatsApp/voice demo referencing the exact R$650 “service quality” MED scenario. High-ticket, high-frequency victims.
• Contador channel: Brazilian SMBs run on their accountant. Partner with 10–20 contadores serving PJ/MEI clients; they already field the panicked “meu Pix foi bloqueado” call and have no answer — give them a referral cut.
• MED-pain content + WhatsApp groups: a single sharp explainer (“MED não vale para desacordo comercial — e o que fazer quando o banco bloqueia mesmo assim”) seeded into wholesaler and seller WhatsApp/Telegram groups and r/empreendedor.
• Reactive capture: monitor Reclame Aqui / Procon / Twitter for fresh “bloquearam meu Pix” posts and DM the affected merchant with the per-dossier offer.

10. Build complexity — justification

Medium. Off-the-shelf: LLM parsing/drafting, NF-e XML is a documented standard, doc storage, WhatsApp Business API, PDF generation. Custom work: the bank-notification parser across many institutions’ inconsistent formats, the legally-correct dossier templates (needs a Brazilian payments/consumer-law advisor), and the deadline/escalation engine. No regulatory approval to operate — it’s a merchant-side document and workflow tool, not a payment institution. Small team, ~3–4 months to a credible v1; the domain/legal accuracy is the long pole, not the engineering.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Merchant-side evidence/document tool; cites public BC norms; not a PSP, not legal practice (templates + self-service, lawyer advisor for content)
Ethical — no harm / dark patterns	✅	Helps legitimate merchants defend against wrongful blocks; explicitly will not help contest genuine fraud — abuse-screening needed
Market exists (evidence above)	✅	Press-documented losses, mandatory MED 2.0, millions of Pix PJ sellers
1–5 person team can build this	✅	Medium complexity, off-the-shelf AI + standard stack
Launchable with <$50K / ₹40L	✅	No capex; cost is build time + legal advisor

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	Frozen balance + lost goods, recurring for high-ticket sellers, money on the line this week. Hair-on-fire for the affected segment.
Demand evidence	15	12/15	Multiple independent signals: press coverage, documented R$650–6k losses, regulatory mandate, market size. Slightly short of top band — no public competitor revenue to point at.
Build feasibility	15	11/15	Off-the-shelf AI + standard stack, but multi-bank notification parsing and legally-correct templates need real care. Pair in ~3–4 months.
Distribution clarity	15	12/15	Geographically concentrated, named victims (25 de Março), contador channel, reactive Reclame Aqui capture. Conversion math credible, not yet proven.
Revenue mechanics	15	11/15	R$-native pricing benchmarked vs ContaAzul tiers; $1M path clear; $5M needs a channel partnership assumption.
Time to first revenue	10	8/10	Acute pain + per-dossier transactional offer enables paying customers within weeks of launch; not pre-sold.
Defensibility	10	5/10	Execution + accumulating bank-format/template knowledge and merchant evidence data. Copyable, but regulatory domain depth and a 6–12mo head start plus channel lock give a real window. A bank/acquirer could in-source it (risk flag).
Total	100	76/100

13. Qualitative modifiers

Founder-fit tags

domain-expertise-required (Brazilian payments + consumer/Pix regulation) · technical-heavy (parsing + LLM drafting pipeline). A Brazilian founder or a founder with a strong local payments-law advisor is near-mandatory.

Key assumptions to validate (3–5)

1. Assumption: High-ticket Pix sellers experience wrongful/abusive MED blocks often enough (≥ a few times/year) to pay a recurring subscription, not just a one-off. How to test: 30 structured interviews across 25 de Março + phone-repair shops; quantify block frequency and R$ exposure per merchant.
2. Assumption: Banks materially favor a structured, well-argued boa-fé dossier vs. an unstructured reply (i.e., the dossier changes outcomes). How to test: run 15–20 real cases through partner merchants with vs. without the dossier; measure unblock rate and time-to-unblock.
3. Assumption: A merchant-side document/template tool stays clear of “exercício ilegal da advocacia” and PSP licensing. How to test: written opinion from a Brazilian payments/consumer-law firm before launch.
4. Assumption: Merchants will forward bank MED notifications quickly enough to act within the window. How to test: measure notification-to-upload latency in a 20-merchant pilot.

Risk flags

1. Platform/regulatory dependency: Entire product is downstream of Banco Central MED rules; BC could add a native receiver-defense channel and collapse the wedge. Mitigate by owning the evidence-capture + bank-format layer that survives a rule change.
2. Incumbent in-sourcing: A bank, acquirer, or accounting platform (ContaAzul, Cora, Asaas) could bundle this. Mitigate via speed, niche brand, and pursuing them as channel partners early.
3. Abuse / ethics: Tool must not become a way to defeat legitimate fraud contestations. Needs hard screening and a documented refusal policy, or it becomes a reputational and legal liability.
4. Market timing: Pain is real now but bank automation may mature and auto-protect boa-fé receivers, shrinking the manual-defense gap over 12–24 months.

14. Structured verdict

Score:                  76/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Brazilian technical founder (or strong BR payments-law advisor) with SMB GTM hustle
Time to revenue:        6–10 weeks (per-dossier offer can monetize before full subscription product)
Capital to launch:      R$60k–R$120k ($11–22K) — mostly build time + legal advisor
Top 3 assumptions to validate first:
  1. Block frequency & R$ exposure per high-ticket seller — 30 merchant interviews in 25 de Março + repair shops
  2. Dossier changes unblock outcome/speed — 15–20 real cases A/B via pilot merchants
  3. Legal posture (no PSP/advocacia issue) — written opinion from BR payments/consumer-law firm
Kill criteria:
  - Abandon if <30% of 30 interviewed high-ticket sellers report ≥1 wrongful/abusive MED block in the last 12 months
  - Abandon if structured dossier shows no measurable lift in unblock rate or speed across the pilot cases
  - Abandon if Banco Central ships a native receiver-defense channel before v1 launch

15. Next step — 1-week validation sprint

• Day 1–2: Build the interview script. Walk Rua 25 de Março and call 30 phone-repair/electronics shops. Quantify: how many MED blocks in the last 12 months, R$ exposure, what they did, how it ended.
• Day 3–4: Recruit 5 willing merchants. Hand-assemble a boa-fé dossier for any live or recent block (no software — concierge). Submit through their bank. Get a payments lawyer’s quick read on legal posture.
• Day 5: Decide. Go if ≥30% of interviewed sellers report a wrongful/abusive block in 12 months and ≥2 of the 5 concierge dossiers produce a faster/positive unblock vs. their prior unaided attempts. Otherwise no-go.

The result is falsifiable: a counted block-frequency rate and a binary per-case unblock outcome — not “merchants seemed interested.”