RetenSync — retention-credit reconciler for Mexican marketplace sellers

1. One-liner

Stitches Mercado Libre and Amazon retention CFDIs to payouts and SAT-ready monthly credit packets for Mexican marketplace sellers.

2. Trend signal — why now?

Three independent signals converge in Q1–Q2 2026:

1. Retention spike Jan 1 2026. SAT lifted platform ISR retention from 1% to 2.5% (a 150% jump), kept IVA at 8%, and for the first time extended the regime to personas morales (S.A., S. de R.L., SAPI). Without an RFC the rate climbs to 20% ISR + 16% IVA — non-recoverable. Effective cash-flow drag on a 2–5%-margin reseller: ~10.5% of gross sales locked up every month. Mercado Libre’s own corporate communications publicly opposed the measure, citing “millions of SMEs” whose ISR refunds come yearly and IVA refunds take months. Sources: Contadores Mismo Lenguaje, Cajanauta blog, La Política Online.

2. SAT real-time platform access Apr 1 2026 (Rule 2.9.21). Marketplaces must hand SAT a permanent online window into every transaction, broken down by seller RFC, with searchable 5-year history. Noncompliance = “kill switch” suspension of platform access. SAT now sees seller-level revenue before the seller files. Discrepancia fiscal penalty: 55–75% of tax + criminal sanctions added in the Dec 2025 reform. Sources: Bloomberg Tax, Fonoa, KPMG.

3. 8M+ CFDIs rejected in 2025 over simple errors. SAT updated CFDI catalogs 12 times since 4.0 launched; 24-business-day expedited audits with 5 business days to produce supporting docs. Forum noise: Amazon sellers on Seller Central México publicly venting they “He solicitado ayuda en todos los centros de atención de Amazon para obtener el desglose detallado para verificar a qué facturas específicas corresponden dichas retenciones” — and getting no response. Sources: Gigstack CFDI 4.0 guide, Amazon Seller Central forum thread.

Provenance:

• Signal 1: ISR retention 1%→2.5%, IVA 8%, extended to corporations, ~10.5% cash drag on SMB sellers — https://contadoresmismolenguaje.mx/inicio/blog/retencion-iva-isr-plataformas — 2026-01
• Signal 2: SAT Rule 2.9.21 real-time platform data access effective Apr 1 2026 + criminal sanctions in Dec 2025 reform — https://news.bloombergtax.com/tax-insights-and-commentary/mexicos-new-digital-access-law-creates-continuous-tax-oversight — 2026-04
• Signal 3: 8M+ CFDI rejections in 2025 + active forum complaints about retention breakdown opacity — https://sellercentral.amazon.com.mx/seller-forums/discussions/t/8d6b74fb-3c09-49f5-b692-96c04866605c — 2026-02 Category: Regulatory arbitrage

3. The opportunity

The retention regime turns every Mexican marketplace seller into a monthly reconciliation case. Each month the seller must pull:

• the platform’s retention CFDI (Mercado Libre under RFC DCM991109KR2, Amazon under its own RFC — 100+ pages of detail at scale),
• the platform’s payout report (gross sales, commission, FBA fees, returns),
• their own emitted CFDIs (Ingreso, plus Egreso for refunds, plus Global for non-RFC buyers),
• bank deposits to reconcile gross-vs-net,

and match them line-by-line so the contador can credit the retention against the monthly ISR/IVA declaration. Miss the match: seller pays the tax twice — once via retention, once on the declaration — and waits 6–18 months (sometimes never) to recover IVA balance-in-favor via a SAT devolución that itself requires a separate procedure each month.

Incumbents miss this niche cleanly:

• Heru ($249/mo, heru.app) is freelancer-issuer-centric — emit CFDIs, do the monthly declaration. Doesn’t reconcile marketplace retentions against payouts as a primary workflow.
• Gigstack (gigstack.pro/marketplaces) sells to marketplaces themselves — infrastructure for platforms emitting CFDIs to buyers, not seller-side retention-credit reconciliation.
• CONTPAQi / Aspel SAE are accountant-tier desktop suites ($1,000s setup + ongoing). Wrong wallet for a 50–500K MXN/mo seller.
• Contador único (independent CPA) charges 2,000–5,000 MXN/mo for PFAE and still misses the marketplace-specific reconciliation 30%+ of the time because the volume is too high to eyeball.
• Quiero Mi Factura, FacturoPorTi — automate CFDI emission for marketplace orders, not retention reconciliation downstream.

Wedge: a 399 MXN/mo (~$20) SaaS focused on one job — turn the platform’s retention CFDI + payout report + your CFDIs into a SAT-ready monthly credit packet, flag discrepancies before SAT does.

4. Target market

• Primary customer: Mexican-based marketplace sellers (Mercado Libre, Amazon México, also Mercado Shops + Walmart Connect later) doing 50K–1M MXN/month in gross sales. RESICO or PFAE regime (90% of the base). Run by 1–3 people, with either a shared contador or none.
• Why they buy: Cash is locked up. ~10.5% of gross retained every month, IVA-favor balance climbing fast, anxiety about getting the credit applied correctly so it doesn’t compound. Forum-verbatim: “Estrangulando al emprendedor.” “Ahora tendrás 12% por lo menos de flujo de efectivo.” “Su IVA a favor comenzará a crecer rápidamente.” “Tendrán que contratar un abogado fiscalista para recuperar el saldo a favor de IVA y les va a cobrar por lo menos el 15% – 20%.”
• Rough TAM reasoning: Mexico e-commerce 2024 = US$97B (eCDB). Mercado Libre Mexico + Amazon Mexico are the two share-gainers through 2027 (eMarketer). Conservative seller base: 300–500K active sellers on MeLi MX, 100K+ on Amazon MX (sellers maintain on multiple platforms). Target the SMB segment doing 50K–1M MXN/mo — call it 150K addressable sellers. At 399 MXN/mo, 5% penetration = 7,500 paying users = ~36M MXN ARR (~US$1.8M).
• Why now for them: Three things: (a) Jan 1 2026 retention spike turned a 1% annoyance into a 10.5% balance-sheet event. (b) Apr 1 2026 SAT gets real-time view — discrepancia fiscal detection automated. (c) Personas morales swept in for the first time — corporate sellers who had no platform-side fiscal burden now have one.

5. Product sketch (MVP)

• One-click connect to Mercado Libre Seller Central and Amazon Seller Central México (OAuth where supported, scoped session-token elsewhere) — pull monthly retention CFDI XML, payout reports, refund logs.
• Auto-fetch the seller’s emitted CFDIs (Ingreso, Egreso, Global) via the SAT Buzón Tributario (CIEL/e.firma) — the standard authenticated path.
• Monthly reconciliation view: every transaction lined up — gross sale, IVA collected, IVA retained, ISR retained, commission, refund, payout deposit — with a status flag (matched, discrepancy, missing CFDI).
• Auto-fix common errors: missing Egreso for refund, mis-coded uso de CFDI, wrong forma de pago (key cause of the 8M+ 2025 rejections).
• Monthly credit packet PDF + XML bundle to hand to the contador or upload directly to the SAT mensual declaration portal: total retention to credit, IVA-favor accumulated, ISR creditable, line-item evidence for each peso.
• IVA-favor alert: when the favor balance crosses the threshold where devolución is worth filing (~10K MXN with documented evidence), produce the aviso de devolución draft.
• Discrepancia-fiscal preflight: estimate what SAT will see via the Apr 2026 Rule 2.9.21 feed, compare against your declared income, warn before SAT auto-flags.

6. AI angle — what’s load-bearing

• CFDI XML parsing + multi-document matching. Platform retention CFDIs are 100+ pages of line items; payout reports are CSV with platform-specific schemas; refund logs sit in a third place. LLMs map fields across the three sources reliably where naive joins miss 20–40% (date drift, partial cancellations, currency conversions on USD listings). Without AI you’re hand-mapping per platform, per month, per seller.
• Error classification. The 8M 2025 rejections are dominated by ~15 root causes (wrong uso de CFDI, missing complemento de pago, RFC of receiver malformed, wrong unit code 84111506 vs ACT, etc.). LLM diagnoses and rewrites the field — deterministic rules cover the obvious 60%, LLM picks up the long tail.
• Plain-language explainer in the seller’s voice (not contador-speak): why this retention got credited, why this didn’t, what to fix.
• Remove the AI and you have a glorified CSV viewer. AI is load-bearing.

7. Localization angle (if any)

This is a Mexico-only play and the localization is the moat:

• All flows in Spanish — both castellano formal (for contador-facing reports) and conversational (for seller in-app).
• Native SAT e.firma / Buzón Tributario / PAC integration. Native Mercado Libre + Amazon México auth.
• Native Mexican payment rails: SPEI, OXXO, MercadoPago — for the subscription itself (399 MXN/mo via MercadoPago beats a Stripe USD-charge UX every time).
• WhatsApp-first delivery of monthly packet — WhatsApp is the de-facto business channel in Mexico for SMB.
• Adjacent expansion: Brazil (Mercado Livre BR, Nota Fiscal Eletrônica, similar marketplace retention regime in progress) — same architecture, different fiscal rules. After 18 months of Mexico traction.

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

• Pricing: 399 MXN/mo (~$20 USD) Solo (one platform connection), 699 MXN/mo Pro (multi-platform + IVA-devolución draft + WhatsApp delivery), 1,499 MXN/mo Studio (≤5 sellers under one contador login, white-label PDF).
• ACV: ~$240 USD/yr per Solo seller, ~$420 Pro, ~$900 Studio. Weighted ACV ~$300/yr.
• Path to $1M ARR: 3,300 paying sellers at ~$300 ACV = $1M. Mexico has 150K+ addressable sellers in the target band; 2.2% penetration.
• Path to $5M ARR: 16,500 paying sellers — 11% penetration of the addressable band, plus contador-channel Studio plan expansion. Plausible within 24 months because the regulatory trigger is forcing the decision on every seller in the same quarter.
• Expansion path: (1) contador-channel: bundle Studio plan to the 5K+ independent contadores serving SMB sellers — they bring 5–20 sellers each. (2) IVA-devolución filing as a 12% success fee on amounts recovered (fiscalistas charge 15-20% per forum — undercut and automate). (3) MeLi/Amazon → Walmart Connect / Liverpool Pro / Mercado Shops / Shopify MX expansion.

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

1. Facebook seller groups — direct DM + value-first posts. “Vendedores de Mercado Libre México 🇲🇽” and 3+ sister groups have tens of thousands of members. Post weekly: “I parsed 50 random retention CFDIs from January — here are the top 5 errors I saw and how to fix them.” Comment-DM the engaged. Conversion math: 4 posts/week × 50 DMs to engagers × 5% trial = 10 trials/week, 30% paid conversion = 3 paying customers/week from this channel alone.
2. Amazon Seller Central México forum — answer questions, drop signature with link. The retention threads (thread 1, thread 2, thread 3) have hundreds of comments by anxious sellers explicitly asking for the reconciliation help.
3. Contador partnership — Studio plan affiliate. ~5,000 independent contadores serve SMB marketplace sellers. Cold-pitch the top 200 via LinkedIn + Colegio de Contadores Públicos chapter newsletters. Studio plan at 1,499 MXN/mo replaces 5–10 hours of monthly reconciliation work; contador keeps 30% rev-share on sellers they onboard.
4. YouTube + TikTok creator partners. Mexican “vende en Mercado Libre”-niche YouTubers (Vende+, Vendedor Premium, Cajanauta) have audiences of 50–300K. Sponsor a 60-second demo. CPM at this niche ~$15–25. ~$500–1,500 per video, expect 30–80 trial signups per video.
5. Cajanauta-style fiscal-advice blogs (cajanauta.com and similar) — affiliate revshare 20% for the first year of subscription.

The wedge is time-of-pain: every seller logs into MeLi/Amazon between the 1st and 5th of the month to download last month’s retention CFDI. Be present in their feed that week.

10. Build complexity — justification

Medium. Off-the-shelf: CFDI 4.0 XML parsing libraries exist (Facturama, Edicom SDKs), SAT Buzón Tributario download is documented, MercadoLibre API gives orders + payout reports under OAuth, Amazon Seller Central exposes Tax Document Library reports. Custom work: the matching engine + the catalog-of-errors LLM classifier + the contador-grade PDF packet generator + the WhatsApp delivery layer. Realistic v1: 1 backend dev + 1 designer/PM-builder + 1 fiscal advisor (part-time) → 12–14 weeks. The fiscal advisor is non-negotiable; mis-classifying a uso de CFDI code burns the customer.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Authenticated user-side access to their own SAT + platform accounts. No scraping of third-party data.
Ethical — no harm / dark patterns	✅	Helping sellers credit retentions they’re entitled to.
Market exists (evidence above)	✅	8M+ CFDI rejections, public forum complaints, Mercado Libre’s own public objection.
1–5 person team can build this	✅	3-person team, 12–14 weeks.
Launchable with <$50K / ₹40L	✅	<$15K USD: dev, fiscal advisor retainer, MercadoPago + WhatsApp Business API onboarding.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	18/20	Hair-on-fire. 10.5% cash drag every month, criminal sanctions added Dec 2025, real-time SAT view from Apr 2026. Forum quotes verbatim about “estrangulando”.
Demand evidence	15	13/15	Multiple independent verbatim complaint threads, Mercado Libre corporate-level objection, regulatory text explicit, paid alternatives (contadores at 2-5K MXN/mo, fiscalistas at 15-20% success fee).
Build feasibility	15	11/15	Medium. CFDI parsing + SAT Buzón auth + multi-platform integration. 12–14 weeks, off-the-shelf libraries, but fiscal-correctness bar is real.
Distribution clarity	15	12/15	Named Facebook groups, named forum threads, named contador channel. Conversion math is realistic. Slight risk: paid acquisition through ads = expensive in this niche.
Revenue mechanics	15	11/15	399 MXN/mo ($20) is a believable wallet for a seller doing 50K-1M MXN/mo gross — single retention error costs more. ACV $240-900. Path to $1M ARR clear at 2.2% penetration. Churn risk on seasonal sellers.
Time to first revenue	10	8/10	Paid pilot in week 4. Real product live week 12. Revenue from day-of-launch via Facebook groups.
Defensibility	10	7/10	Workflow lock-in (12 months of reconciled history), fiscal-domain knowledge moat, contador channel partnerships. Not a deep moat — incumbents could copy in 6–9 months but only if they re-tool away from their current ICP.
Total	100	80/100

13. Qualitative modifiers

Founder-fit tags

domain-expertise-required · technical-heavy

A fiscal advisor (Mexican CPA / fiscalista) on the founding team or as a paid retained partner is non-negotiable. The technical work is standard SaaS + integrations + LLM. Without the fiscal partner you’ll mis-credit retentions and damage customers in month 1.

Key assumptions to validate (3–5)

1. Assumption: Sellers will pay 399 MXN/mo for monthly reconciliation, not bundle it with their existing contador. How to test: 20 in-person/Zoom interviews with sellers in the 50K-1M MXN/mo band; ask explicitly what they pay their contador, what’s missing, and price-anchor at 399 MXN.
2. Assumption: Amazon + MeLi seller-account integrations are stable enough for 99%+ monthly reconciliation reliability. How to test: Build a read-only prototype that pulls 30 days of retention CFDI + payout from 10 friendly sellers. Measure: % of line items auto-matched without human review.
3. Assumption: Contadores will partner via revshare rather than block as competitive threat. How to test: 20 cold contador calls; pitch Studio plan as a replacement for 5-10 hours of their monthly reconciliation grunt-work, not their tax-advice billable.
4. Assumption: The 8M-rejection error catalog converges to <20 root causes that an LLM can classify reliably. How to test: Hand-label 500 real rejection notices, build the classifier, target 90%+ accuracy on a held-out 200.

Risk flags

1. Platform-API risk: MercadoLibre or Amazon could close the seller-side data pulls or change auth. Mitigation: scope to authenticated user-token access (the seller delegates their own credentials), not platform partnership. SAT Buzón is the canonical source for retention CFDIs and won’t change.
2. Heru / Konfio / Gigstack moves down-market: Heru already has consumer-grade UX and an issuer-side product. If they pivot to retention reconciliation, they bring brand. Mitigation: ship a focused MVP in 12 weeks while they’re focused on freelancer-issuance and lending respectively.
3. SAT rule change: Reforma fiscal in late 2026 could simplify the regime (unlikely) or further tighten (likely). Both favor the product, but a one-off simplification could compress willingness-to-pay. Watch the November 2026 Resolución Miscelánea draft.
4. Contador backlash: Contadores who see the product as commoditizing their work could resist. Mitigation: the Studio plan positions the product as a productivity tool for the contador, not a contador replacement. Reframe.

14. Structured verdict

Score:                  80/100
Verdict:                STRONG GO
Confidence:             High
Best-fit builder:       Solo or pair technical founder with a retained Mexican fiscalista
Time to revenue:        4 weeks (paid pilot), 12 weeks (live product)
Capital to launch:      $12K–$18K USD (dev tooling, fiscalista retainer, WhatsApp Business + MercadoPago onboarding)
Top 3 assumptions to validate first:
  1. Will sellers pay 399 MXN/mo on top of contador — 20 interviews
  2. Read-only Amazon + MeLi integration reliability — 10-seller prototype
  3. Error-classification accuracy — hand-label 500 rejections, hit 90% on held-out set
Kill criteria:
  - Abandon if <20% of 50 surveyed sellers say they'd pay 399 MXN/mo standalone
  - Abandon if Heru or Konfio ships a head-on competing product before week 14 of build
  - Abandon if a SAT rule revision in late 2026 reverses the 2.5% ISR retention back below 1.5%

15. Next step — 1-week validation sprint

• Day 1–2: Recruit 25 Mexican marketplace sellers from “Vendedores de Mercado Libre México 🇲🇽” Facebook group via 1-paragraph DM offering free first-month reconciliation. Goal: 15 booked Zoom calls.
• Day 3–4: Run 15 Zoom calls. Show a Figma click-prototype of the monthly credit packet. Ask: “Today, how do you handle this? What does it cost? Would you pay 399 MXN/mo for this PDF every month?” Price-anchor explicitly. Capture verbatim objections.
• Day 5: Tally responses against a falsifiable cut: ≥8/15 say “yes, I’d pay 399 MXN/mo within the next 30 days” = GO. ≤3/15 = PASS. In between = build a 2-week paid-pilot landing page + Facebook ad, target 50 trial signups at 199 MXN first-month price-anchor; go/no-go on conversion to 399 MXN month-2.

Falsifiable result: payment commitments collected during the call window, not vague interest.