TindaBalik — deduction-recovery desk for Shopee & TikTok Shop sellers

1. One-liner

Audits every SEA marketplace settlement, files the claim for wrongful deductions, and takes a cut of what it recovers.

2. Trend signal — why now?

SEA marketplace sellers are getting quietly bled by the deduction stack, and the volume of leakage just exploded because of livestream commerce.

• Effective take-rate across Shopee/TikTok/Lazada has climbed to 20–25% of post-discount sales in 2025–2026, and SEA commission rates are up more than 10 percentage points since TikTok Shop entered Indonesia in 2021 (digitalinasia, shiply.ph).
• Payouts routinely don’t match sales: TikTok “frequently applies fee corrections, clawbacks, and promotional adjustments to future settlements — not the one where the original sale occurred,” producing zero-dollar deposits when it claws back from the next cycle (signalytics, settlebridge). Sellers also report TikTok applying free-shipping and discounts without authorization, with the seller eating the cost via payout deductions.
• Livestream commerce runs at 5×–20× baseline volume in 30–90 minute surge windows with elevated impulse-return rates — every order is another place a wrong shipping/weight charge or clawback can hide (locus.sh).
• The recovery model is already proven next door: Amazon FBA reimbursement is a mature market with 10+ players charging a standard 25% of recovered funds, no upfront cost (smartscout). Nobody runs this playbook on Shopee/TikTok.

The tools that exist in SEA — BigSeller (free), SmartB, SiteGiant — reconcile and flag the discrepancy. SiteGiant will even show you the “overcharged shipping fee” and a “success claimed amount” column. But they hand you a report and leave you to file the claim yourself, manually, per order, per platform. That last mile — draft the claim, submit it, chase it, book the recovered cash — is nobody’s product.

Provenance:

• Signal 1 (Demand): SEA take-rate at 20–25%, payouts don’t match sales, unauthorized free-shipping deductions, wrong-period clawbacks — signalytics / smartb — 2026-07-26
• Signal 2 (Feasibility): Marketplace settlement APIs already pulled by SiteGiant/BigSeller; SiteGiant auto-computes overcharged shipping; cheap multilingual LLM drafts per-platform claims — sitegiant — 2026-07-26
• Signal 3 (Economic): FBA reimbursement is a proven 25%-of-recovery market with 10+ funded players; SEA seller-finance tooling actively shipping — smartscout — 2026-07-26 Category: Geographic arbitrage

3. The opportunity

The Amazon world figured out years ago that platform accounting leaks money and that sellers won’t chase it themselves — so a whole category of performance-based recovery services grew up around it (GETIDA, Seller Investigators, SPS Revenue Recovery), all taking ~25% of what they claw back, all Amazon-only.

SEA has the same leak, worse (higher take-rates, more clawbacks, livestream chaos, multi-platform) — but the seller tooling stopped at “here’s a reconciliation report.” BigSeller conditioned the whole market to expect reconciliation for free, so everyone built free matching and nobody built the hard, valuable part: actually recovering the cash.

The incumbent to beat isn’t a company — it’s the seller’s own spreadsheet and their resignation to “that’s just the platform tax.” SiteGiant/SmartB/BigSeller are complements, not competitors: they detect, we recover. The 10× is going from “you can see you were overcharged RM4,200 this month” to “we already filed 63 claims and RM2,900 landed back in your wallet — here’s your cut of what we got.”

4. Target market

• Primary customer: Owner-operators of Shopee / TikTok Shop / Lazada stores in the Philippines, Malaysia, and Indonesia, doing roughly $5K–$80K monthly GMV, 1–5 staff, mixing livestream selling with regular listings. The ones big enough to bleed real money (a 1–2% recovery on $40K GMV/mo is $400–800/mo found) but too small to hire a finance person to chase it.
• Why they buy — in their words: “My payout never matches my orders.” “Where did my money go?” “TikTok deducted free shipping I never turned on.” “The commission got adjusted after the sale.” They already know they’re being overcharged; they just have no time to fight order-by-order across three platforms.
• Rough TAM reasoning: SEA has millions of active marketplace sellers; the slice doing >$5K/mo across PH/MY/ID is comfortably in the low hundreds of thousands. Capturing a few thousand of them at performance pricing clears $5M ARR. This is a wallet-share play on money that already exists, not net-new spend.
• Why now for them: Livestream pushed order counts up 5–20×, take-rates crossed 20%, and margins are compressing “disguised as growth” — the leakage is now large enough that recovery is worth the mental overhead of connecting an account, especially when it’s free-if-we-fail.

5. Product sketch (MVP)

• Connect Shopee / TikTok Shop / Lazada seller accounts (OAuth / seller-center API) in a few clicks — no spreadsheet uploads.
• Deduction audit engine: every settlement is scanned for the known wrongful-deduction patterns — overcharged/weight-discrepancy shipping, unauthorized free-shipping charges, promo subsidies the seller didn’t opt into, wrong-period clawbacks, refund double-deductions.
• Claim drafter: for each flagged item, auto-generate the platform-specific dispute in the seller’s language (Tagalog / Bahasa / English), pre-filled with the order IDs, amounts, and the exact policy the platform violated.
• File & track: submit through the platform’s claim path where an API exists; where it’s manual, produce a one-tap-ready packet and track the claim to resolution.
• Recovery ledger: running dashboard of money flagged → claimed → recovered → your cut, per store, per platform.
• Weekly “you were overcharged” digest on WhatsApp/email — the hook that keeps sellers engaged even before they’ve recovered a peso.
• Pricing shown as you keep 75–85% of everything we recover; if we recover nothing, you pay nothing.

6. AI angle — what’s load-bearing

AI does two jobs that make this a product instead of a services shop:

1. Classification at scale. Each settlement line has to be judged: is this deduction legitimate (real commission) or wrongful (weight overcharge, unauthorized free-shipping, mis-timed clawback)? The rules differ per platform and change constantly. An LLM + a maintained rule library classifies thousands of noisy, inconsistently-labeled settlement lines per seller per week. Remove it and you’re back to a human reading statements — the exact task sellers already refuse to do.
2. Claim drafting in local language, per platform. Turning “order #X was charged RM12 shipping but buyer paid RM4, policy Y says the difference is claimable” into a submitted, correctly-worded dispute in Tagalog or Bahasa — at volume — is generative work. Without it, you’re not filing 63 claims a week; you’re filing 3.

If you remove the AI, this collapses into a slow manual bookkeeping service that can’t scale past a handful of clients. The AI is the reason one small team can run recovery for thousands of sellers.

7. Localization angle

This is a SEA-native play, not a translated global tool. The whole wedge is local:

• Language: claims and digests in Tagalog, Bahasa Indonesia, Bahasa Malaysia, English — the language the seller and the platform’s support actually use.
• Platform-specific dispute rules: Shopee, TikTok Shop, and Lazada each have different fee structures, claim windows, and overcharge policies, country by country. That rulebook is the moat.
• Distribution: WhatsApp-first digests; sellers live in WhatsApp and platform seller groups, not email.
• Pricing psychology: performance-based (% of recovery) works where a $30–50/mo subscription would get ignored by a seller who’s already conditioned to free reconciliation tools.

A generic Western recovery tool (Amazon-only, English, USD) cannot serve this market. That’s the arbitrage.

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

• Pricing: performance-based — we take 15–25% of funds recovered, seller keeps the rest, no upfront and no monthly. Directly benchmarked to the FBA-recovery standard of 25%. Undercut incumbents’ psychology by leading with “you keep 80%.”
• ACV: assume a target seller does $30K/mo GMV, ~1.5% of GMV is recoverable wrongful deductions = ~$450/mo found; we recover ~60% of that = ~$270/mo recovered; our 20% cut = ~$54/mo → ~$650 ACV/year. Bigger sellers scale linearly.
• Rough math to $1M ARR: ~1,540 active sellers at $650 ACV. Very reachable within PH+MY+ID.
• Rough math to $5M ARR: ~7,700 sellers, or a mix of a smaller base with larger livestream sellers (a $100K/mo store is ~$180/mo cut → $2,160 ACV, so ~2,300 of those). Add Vietnam + Thailand for headroom.
• Expansion path: move from recovery-only to a paid “leak-prevention” tier (alerts before you get overcharged, auto-appeal of unauthorized free-shipping, margin dashboards) at a flat monthly on top of the recovery cut — turning a spiky performance revenue into a recurring base.

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

• Seller Facebook/WhatsApp groups: SEA marketplace sellers cluster in huge, active FB groups (Shopee/TikTok seller communities per country) — the exact overcharge-shipping-claim threads showing up in FB group posts. Post a “we found RM X overcharged for 5 sellers this week, free audit” offer; performance pricing removes all buyer risk.
• Free reverse-audit as the hook: connect an account, we show you exactly how much you were wrongfully deducted last 90 days for free. The number itself sells — it’s their own money on a screen. Convert to “let us file the claims” with one click.
• Livestream-seller creators: partner with 3–5 mid-tier SEA livestream sellers who already train other sellers; a “here’s how I clawed back ₱X” case study is native content their audience wants.
• Reconciliation-tool adjacency: SiteGiant/BigSeller users already know they’re overcharged (their reports say so) but don’t file. Target that exact frustrated cohort — they’re pre-educated on the pain and one step short of the fix.

The first 100 come from one country’s biggest seller FB group + the free reverse-audit hook. That’s a 2–4 week sprint, not a channel guess.

10. Build complexity — justification

Medium. Off-the-shelf: marketplace seller APIs (already pulled by SiteGiant/BigSeller, so the surface is known-feasible), LLM for classification + multilingual claim drafting, standard web stack, WhatsApp Business API. The custom work is the maintained per-platform-per-country deduction rule library and the claim-submission flows (some API, some semi-manual packet generation), plus handling settlement data that’s messy and inconsistently formatted. A pair can ship a single-platform, single-country v1 (Shopee PH, recovery-only) in ~10–14 weeks; multi-platform/multi-country is the scaling grind, not a v1 blocker.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Filing legitimate overcharge claims on the seller’s own behalf via official platform channels. No scraping of buyers, no ToS-hostile behavior.
Ethical — no harm / dark patterns	✅	Recovers money the seller is genuinely owed; performance pricing aligns incentives; nothing hidden.
Market exists (evidence above)	✅	Proven demand + proven recovery model next door (FBA).
1–5 person team can build this	✅	Pair for v1, small team to scale platforms/countries.
Launchable with <$50K / ₹40L	✅	API access + inference + WhatsApp; no capex.

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Real, recurring (every settlement), money-losing pain sellers already complain about loudly. Just short of hair-on-fire because they’ve resigned to it as “platform tax.”
Demand evidence	15	12/15	Multiple independent signals + a proven parallel market (FBA recovery at 25%). Strong. Docked slightly: SEA sellers’ willingness to actively pay for recovery (vs expect it free) is inferred from Amazon, not yet directly observed in SEA.
Build feasibility	15	11/15	Known-feasible APIs + LLM, but the per-platform rule library and messy settlement data are real engineering; not a 4-week solo build.
Distribution clarity	15	12/15	Named channels (seller FB/WhatsApp groups), zero-risk performance offer, free-audit hook. Conversion of “flagged” → “let us file” is the uncertainty.
Revenue mechanics	15	12/15	Pricing benchmarked hard to FBA standard; math to $1M is a few thousand sellers. Risk: recovery volume per seller could be lower than the 1.5%-of-GMV assumption.
Time to first revenue	10	7/10	Performance model = first recovered dollar can come weeks after launch once claims resolve, but claim resolution has platform-side lag.
Defensibility	10	4/10	Rule library + local claim flows + workflow lock-in are soft moats; a well-funded reconciliation incumbent (BigSeller/SiteGiant) could bolt recovery on. Platform dependency is a real risk.
Total	100	74/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy (API integrations + LLM classification pipeline), operations-heavy (maintaining per-platform claim rules and handling semi-manual claim flows). A domain advisor who’s run a SEA marketplace store is a strong plus.

Key assumptions to validate (3–5)

1. Assumption: Wrongful/recoverable deductions average ~1–2% of GMV per seller. How to test: run free reverse-audits on 20–30 real sellers’ last-90-days settlement data across PH/MY/ID; measure recoverable % distribution.
2. Assumption: Platforms actually honor these claims at a meaningful rate when well-drafted. How to test: file 50 real claims manually across Shopee/TikTok/Lazada; measure approval rate and payout lag before building automation.
3. Assumption: Sellers will connect their account and let us file, given performance pricing. How to test: offer the free audit to a seller FB group; measure connect rate and “yes, file it” conversion.
4. Assumption: Marketplace seller APIs expose enough settlement granularity to classify deductions. How to test: technical spike against Shopee PH + TikTok Shop APIs before committing.

Risk flags

1. Platform dependency (high): platforms can change fee/claim policies, restrict API access, or tighten claim windows — they have no incentive to make overcharge recovery easy. This is the top risk.
2. Recovery-rate uncertainty: if platforms reject most claims or recoverable % is <0.5% of GMV, the performance model doesn’t generate enough revenue per seller.
3. Incumbent bolt-on: BigSeller/SiteGiant already have the data pipe and seller base; if they add a recovery loop, the moat is thin. Speed + local claim-drafting depth is the only defense.
4. Market timing on willingness-to-pay: SEA sellers are conditioned to free tools; performance pricing mitigates this but the behavior is unproven locally.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + SEA-marketplace-seller domain advisor
Time to revenue:        8–14 weeks (build) + platform claim-resolution lag
Capital to launch:      ₹6–10 lakh ($7–12K)
Top 3 assumptions to validate first:
  1. Recoverable deductions ≈1–2% of GMV — free reverse-audit on 20–30 real sellers
  2. Platforms honor well-drafted claims at a workable rate — file 50 manual claims first
  3. Sellers connect + say "file it" under performance pricing — FB-group free-audit test
Kill criteria:
  - Abandon if <0.5% of GMV is recoverable across the 20–30 audited sellers
  - Abandon if manual claim approval rate is <30% across the 50 test claims
  - Abandon if a reconciliation incumbent ships a performance-based recovery loop before your v1

15. Next step — 1-week validation sprint

• Day 1–2: Technical spike — connect one Shopee PH and one TikTok Shop test/real account, confirm the settlement API exposes per-order fee/shipping/clawback granularity fine enough to classify deductions.
• Day 3–4: Recruit 8–10 real sellers from one country’s biggest seller FB group; run a free reverse-audit on their last 90 days; hand each a number: “you were wrongfully deducted ₱X.”
• Day 5: Manually file 10–15 of the strongest claims and measure two things: (a) how many sellers said “yes, file it” when shown their number, and (b) early platform response. Go if ≥40% of audited sellers had ≥1% recoverable GMV AND ≥50% said “file it.” No-go if the recoverable pool is thin or sellers shrug at their own number.

The falsifiable result: a distribution of recoverable-% across real sellers and a real conversion rate on “file it” — not vibes.