ClawBack — refund-dispute defender for TikTok Shop sellers

1. One-liner

Auto-builds the evidence pack the moment a TikTok Shop refund dispute fires, and flags repeat refund-and-keep abusers.

2. Trend signal — why now?

TikTok Shop is the fastest-growing US retail channel and the dispute system is bleeding small sellers dry — three things converged in the last 12 months:

• The seller base exploded. US TikTok Shop hit ~475,000 shops (up ~5,000% from mid-2023), and the platform onboarded 1.2M+ new merchants pushing US GMV toward $23–27B in 2026. Most of these are 1–3 person operations with no ops staff.
• The dispute clock is brutal and the math is upside-down. Sellers must submit evidence within 24 hours of a dispute notification or TikTok auto-rules for the buyer at the seller’s expense. Refund-without-return is auto-approved on low-value orders (≤$10), and a $50 item return effectively costs ~$65.70 once you stack product cost, forward + return shipping, lost affiliate commission, and the FBT return-handling fee. A May 2026 policy change made it worse: reject a return TikTok deems valid and you must still ship the item back to the buyer within 3 business days and eat the outbound cost.
• Evidence quality decides outcomes, and small sellers can’t assemble it in time. Industry guides report sellers who submit timestamped screenshots, packing/unpacking videos, and delivery proof see materially higher approval (one source cites ~87% higher approval with proper documentation). The losing sellers usually lose not because they were wrong, but because they submitted the right evidence in the wrong format — or missed the 24h window entirely.

Real seller voice is unambiguous: “I cannot be paying £8.50 per return for these £1 items!” and “if that’s how #tiktokshop is gonna allow sellers to do business I want NO PART OF IT!” Negative-balance penalties then cascade — at -$250 TikTok blocks Flash Deals and coupons, at -$500 it suspends ads, and below that the whole account gets restricted. Refunds aren’t just a cost line; they choke the seller’s growth engine.

Provenance:

• Signal 1 (demand): Sellers publicly furious about refund-and-keep losses and per-return cost; verbatim quotes — https://www.webgility.com/blog/guide-on-tiktok-shop-return — 2026-06
• Signal 2 (feasibility): 24-hour evidence deadline + evidence-format determines win rate; cheap multimodal AI (vision + OCR + LLM) can now assemble timestamped packs from a seller’s order data and media in seconds — https://macetric.com/tiktok-shop-dispute-resolution-seller-guide/ — 2026-06
• Signal 3 (economic): US TikTok Shop ~475K shops, $23–27B 2026 GMV, 1.2M new merchants — money and merchant volume both surging — https://branvas.com/blogs/news/tiktok-shop-statistics — 2026-06 Category: Platform shift

3. The opportunity

The dispute is a timed evidence game and small sellers show up unarmed. TikTok will side with the seller when the seller presses with a clean, on-format evidence pack inside 24 hours — and TikTok will even ban abusive buyers when sellers push investigations with evidence. But a one-person shop running livestreams and packing orders does not see the dispute notification, find the right order, pull the delivery scan, screenshot the live listing, and trim the packing clip — all within 24 hours, repeatedly, for $9 items. So they eat it.

Incumbents exist but aim elsewhere. Chargeflow is a broad cross-platform chargeback engine (bank-side, not TikTok return-abuse-native). 4Seller is a general seller suite where appeals are one feature among dozens. Agentative’s Dispute Defense Agent is the closest, but it’s young and pitched as a generic “audit all disputes” agent. None is laser-built for the specific small-seller pain: (a) beat the 24h clock automatically, (b) assemble evidence in TikTok’s exact accepted format, and (c) connect the dots on repeat refund-and-keep abusers across orders so the seller can trigger a buyer investigation. That third leg — abuser intelligence — is the un-owned wedge. The broad tools fight one dispute at a time; they don’t tell you that buyer “jess_xx” has now refunded-and-kept four times across three of your SKUs.

A focused team can win the narrow shop: “We defend TikTok refund disputes for you, in the format TikTok accepts, before the clock runs out — and we name your serial abusers.”

4. Target market

• Primary customer: US TikTok Shop sellers doing roughly $10K–$300K/year GMV — 1–3 person shops in beauty, home, apparel, supplements, accessories. Not the agency-managed mega-shops (they have ops staff); the founder-operator who packs their own boxes.
• Why they buy: “I’m losing 5–15% of revenue to refunds I can’t fight, the 24-hour window always catches me mid-livestream, and I have no idea which buyers are scamming me on repeat.” Every lost dispute is real cash plus a step toward the -$250/-$500 penalty cliffs.
• Rough TAM reasoning: ~475K US shops. Even if only the ~150–200K that are active, revenue-generating, and dispute-exposed are addressable, capturing 1% at ~$45/mo is ~$800K–$1M ARR; 3–4% is a $3–4M ARR business. Tight, bootstrappable, not a unicorn — exactly the target.
• Why now for them: The May 2026 ship-back rule and the merchant flood mean dispute pain is at its peak and the cohort of inexperienced new sellers (1.2M added) is the largest it’s ever been. The pain is acute, fresh, and growing.

5. Product sketch (MVP)

• Dispute watcher: Connects to the seller’s TikTok Shop account, watches for new refund/return disputes, and starts a countdown timer against the 24-hour deadline with push/SMS alerts.
• Auto-evidence pack: The instant a dispute lands, pulls the order record, carrier delivery scan/proof, the live product-listing screenshot, and the seller’s pre-shipment packing photo/clip — and assembles them into the exact format TikTok’s appeal flow accepts.
• One-tap submit / review-and-send: Seller reviews the generated pack on their phone and submits in one tap; or enables auto-submit for high-confidence cases.
• Win/lose triage: Scores each dispute as likely-winnable vs. not, so the seller spends effort only where evidence can actually flip the outcome.
• Repeat-abuser radar: Tracks refund-and-keep behavior per buyer across all the seller’s orders, surfaces serial abusers, and pre-drafts the buyer-investigation request to TikTok.
• Packing-clip capture: A simple “film as you pack” mobile prompt so the seller always has timestamped proof on file before disputes ever happen.
• Recovery dashboard: Running tally of disputes fought, won, $ recovered, and distance to the -$250/-$500 penalty cliffs.

6. AI angle — what’s load-bearing

Remove the AI and this is a manual checklist nobody has time to run. AI is doing the actual work:

• Vision + OCR read the packing photo/clip and the delivery scan to confirm the right SKU shipped and was delivered, and extract the timestamps TikTok wants.
• An LLM agent maps each dispute’s reason code to the specific evidence combination that wins that code, then writes the appeal narrative in TikTok’s accepted structure — the “right evidence in the right format” that drives the approval lift.
• Pattern detection clusters buyer behavior across orders to flag serial refund-and-keep abusers — a judgment call no static rule makes well.

The product is fundamentally “assemble and argue the strongest defensible case against a 24-hour clock, automatically.” That’s an AI job.

7. Localization angle

N/A — this is a US-first play. The wedge is TikTok Shop’s US dispute mechanics and the US small-seller flood, not a language/payment-rail arbitrage. Expansion is platform breadth (TikTok Shop UK/SEA, then Amazon/Shopify return-abuse) rather than geographic localization. Worth noting TikTok Shop UK has the same dispute structure, so a second market is a config change, not a rebuild — but the beachhead is US.

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

• Pricing: $39–$59/mo flat for the core defender (call it $45 baseline), with a higher $99/mo tier for higher-volume shops (more disputes, auto-submit, abuser investigations). Optionally a small success-fee variant (e.g. 10% of recovered amount) for sellers who prefer pay-for-performance — but flat SaaS is the spine.
• ACV: ~$540/year at the $45 tier; ~$700–$900 blended once a chunk upgrade to the $99 tier.
• Rough math to $1M ARR: ~1,550 sellers × $45/mo × 12 ≈ $840K, plus tier upgrades and success fees clear $1M. That’s ~0.3–1% of the active US shop base — very reachable.
• Rough math to $5M ARR: ~6,000–7,000 paying shops blended at ~$70/mo ACV, OR fewer shops + meaningful success-fee revenue + a second platform (UK or Amazon). Requires retention to hold and TikTok API access to stay open.
• Expansion path: Start at flat SaaS → add success-fee on recoveries → add packing-clip storage/insurance-style add-ons → expand to second platform (TikTok UK, then Amazon A-to-z claims) for the same shop, lifting ACV without new CAC.

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

• TikTok Shop seller subreddits and Facebook groups. There are large, active communities of US TikTok Shop sellers venting about exactly this. Show up with a free “Dispute Cost Calculator” + a teardown of how to win a refund-without-return appeal in the format TikTok accepts, then offer the tool. These groups are where the buyer literally complains in their own words — convert the complaint thread directly.
• Cold-DM the complainers. Scrape sellers publicly posting refund-loss horror stories on TikTok/X/Reddit, send a 60-second Loom showing their type of dispute auto-defended in the 24h window. The pain is fresh and specific; a 3–5% reply rate on a few hundred targeted DMs seeds the first cohort.
• TikTok Shop creator/seller coaches and agencies. Dozens of mid-size creators sell “how to win on TikTok Shop” courses and run seller Discords. Affiliate/rev-share them — they already own the audience and refunds are a top complaint in their communities.
• Content moat on “TikTok Shop refund dispute” SEO. Sellers Google the exact panic phrase (“TikTok shop refund without return appeal”, “missed 24 hour dispute”). Own those pages with genuinely useful playbooks → tool CTA. Slower, but compounding and high-intent.

10. Build complexity — justification

Medium. The AI pieces are off-the-shelf (vision/OCR + LLM agent + a clustering step) and the UI is a standard web + lightweight mobile-capture stack — that part is weeks, not months. The real work and risk is the TikTok Shop integration: dispute-event ingestion, order/delivery data, and submitting appeals in the accepted format, plus staying inside TikTok’s API/partner terms. A small team can ship a credible v1 in ~3–4 months, with the integration being the long pole.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Helping sellers assemble truthful evidence and fight disputes is legitimate; must avoid fabricated/edited evidence (off-platform screenshots get rejected anyway).
Ethical — no harm / dark patterns	✅	Defends sellers against genuine abuse and surfaces real repeat abusers; not about denying legitimate refunds.
Market exists (evidence above)	✅	475K shops, public seller fury, 24h deadline, incumbents already monetizing the adjacent problem.
1–5 person team can build this	✅	Off-the-shelf AI + standard stack; integration is the only heavy lift.
Launchable with <$50K / ₹40L	✅	No capex; main cost is dev time + inference.

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Hair-on-fire for the cohort — direct cash loss, 24h panic, cascading account penalties. Not 18+ because many losses are small-ticket and some sellers absorb rather than fight.
Demand evidence	15	12/15	Strong: explosive seller base, verbatim seller fury, existing paid incumbents in the adjacent space. Slightly capped — most quotes are about the policy, fewer are “I’d pay for a tool.”
Build feasibility	15	11/15	AI/UI easy; TikTok integration + appeal-submission format is the gnarly part and carries platform-terms risk.
Distribution clarity	15	12/15	Named communities, named complainers to DM, coach affiliates, high-intent SEO. Concrete first-100 path.
Revenue mechanics	15	11/15	Clear pricing benchmarked to incumbents; $1M path needs <1% of base. Capped by churn risk (sellers who quit TikTok) and willingness-to-pay at the very low end.
Time to first revenue	10	8/10	Can pre-sell off a Loom + manual-assisted pilot in weeks; full auto-submit takes longer.
Defensibility	10	5/10	Execution + accumulating abuser-pattern data is a soft moat, but incumbents exist and TikTok could absorb the feature. Platform dependency is the core risk.
Total	100	75/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy (TikTok integration + AI agent) · operations-heavy (dispute-format expertise, abuser logic, seller support).

Key assumptions to validate (3–5)

1. Assumption: TikTok Shop’s API/partner access permits ingesting dispute events and submitting appeals (or a reliable seller-authorized path exists). How to test: Review TikTok Shop Partner/developer docs and talk to 3 existing seller-tool vendors about what’s actually permitted before writing a line of integration code.
2. Assumption: Small sellers will pay ~$45/mo to win disputes that are often sub-$50 each. How to test: Pre-sell to 20–30 sellers from the complaint threads with a “we’ll fight your next 10 disputes manually for $45” pilot; measure conversion and stated WTP.
3. Assumption: Auto-assembled evidence packs actually lift win rates meaningfully. How to test: Run 50 real disputes manually-assisted, compare win rate to the seller’s prior baseline; need a clear lift (e.g. +20–30pts) to justify the product.
4. Assumption: Repeat-abuser detection produces action TikTok honors (buyer investigations/bans). How to test: Submit a handful of evidenced repeat-abuser cases and confirm TikTok acts.

Risk flags

1. Platform dependency (severe): This lives or dies on TikTok Shop API access and policy. TikTok could restrict third-party dispute submission, change the appeal format overnight, or build the feature in-house. This is the #1 kill risk and caps defensibility.
2. Market timing / channel fragility: TikTok’s US regulatory status has been volatile; a forced sale or ban would gut the customer base. Mitigant: architecture that ports to Amazon/Shopify return-abuse keeps the team alive if TikTok wobbles.
3. Low-ticket WTP: Sellers losing $9 disputes may balk at $45/mo unless the recovered-$ and saved-time story is vivid. Success-fee pricing is the hedge.
4. Incumbent acceleration: Agentative/Chargeflow could pivot harder into the small-seller TikTok niche. The abuser-intelligence wedge and small-seller focus are the differentiators to defend.

14. Structured verdict

Score:                  75/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder who can stomach a platform integration, paired with someone who lives in TikTok Shop seller communities
Time to revenue:        6–10 weeks to first paid pilot (manual-assisted), ~3–4 months to auto v1
Capital to launch:      $15–30K ($ dev time + inference + a small pilot)
Top 3 assumptions to validate first:
  1. TikTok Shop API permits dispute ingestion + appeal submission — confirm in partner docs / with existing vendors before building
  2. Sellers pay ~$45/mo — pre-sell a manual pilot to 20–30 complainers
  3. Assembled evidence lifts win rate +20–30pts — measure against seller baseline on 50 real disputes
Kill criteria:
  - Abandon if TikTok prohibits third-party dispute submission with no seller-authorized workaround
  - Abandon if <15% of 30 pre-sell targets convert to a paid pilot
  - Abandon if manual-assisted evidence packs don't beat the seller's baseline win rate by a meaningful margin

15. Next step — 1-week validation sprint

• Day 1–2: Read TikTok Shop Partner/developer + appeal-format docs end to end; DM 5 existing TikTok seller-tool vendors to learn what dispute access is actually permitted. Map the legal/technical path before anything else.
• Day 3–4: Pull 30 sellers from refund-complaint threads (Reddit/FB/X). Offer a concrete pilot: “Send me your next 10 disputes; I’ll hand-assemble winning evidence packs in the 24h window for $45.” Track replies and paid sign-ups.
• Day 5: Decide go/no-go on a falsifiable bar: ≥5 of 30 sellers say yes and ≥3 actually pay for the pilot. Below that, the WTP isn’t there — revisit pricing or kill. Above it, start the integration with paying pilots feeding real disputes.