TallyLive — commission reconciler for TikTok Shop sellers

1. One-liner

Matches every clawback and refund to the creator who earned it, then tells SEA TikTok Shop sellers exactly what to pay.

2. Trend signal — why now?

TikTok Shop in Southeast Asia turned into a real-money channel in 2025–2026, and the affiliate economics got complicated enough that the native tools can no longer keep up.

• The channel is huge and live-selling-driven. Indonesia alone has ~515,000 TikTok shops and ~$4.8–5B GMV in 2026, with live-selling driving roughly 60% of sales and live sellers seeing 84% more sales YoY. (resourcera, intura)
• Affiliate commissions are now the core growth motor — and they’re being clawed back. Commission rates run 3–30% by category; the effective take after stacked fees has climbed (one breakdown shows a 15% headline rate becoming 26.6% all-in). Commissions are clawed back from a creator’s pending balance when a buyer returns inside the 15–31 day settlement window, but become permanently non-refundable once paid out. (dashboardly)
• The native Seller Center can’t reconcile it. Per TikTok’s own settlement-report docs and third-party guides, the CSV export “doesn’t attribute affiliate clawbacks to specific creators” and “cannot show who generated the most net profit after returns, COGS, and stacked fees.” Sellers are told to “run actual payout statements against profit calculations separately.” Each payout “bundles dozens of moving parts,” and handling it manually “quickly turns into a full-time job.” (dashboardly payout guide, linkmybooks)
• May-2026 rule changes raised the stakes. TikTok Shop Indonesia introduced minimum per-category commission rates and Creator Health Rating requirements in May 2026 — sellers now negotiate and pay more creators, more often, under more rules. (desakarangbendo, gineehub)

If the only signal were “TikTok Shop is big,” this wouldn’t be ready. The signal that makes it ready is that the platform’s own financial reports are admitted-by-everyone unfit to answer “what do I actually owe each creator this cycle, net of clawbacks?”

Provenance:

• Signal 1 (Demand): Native Seller Center CSV “doesn’t attribute affiliate clawbacks to specific creators”; manual reconciliation “turns into a full-time job” — https://www.dashboardly.io/post/tiktok-shop-payout-explained-when-how-tiktok-pays-sellers — 2026-06-08
• Signal 2 (Feasibility): TikTok Shop exposes downloadable settlement statements (Finances > Statements, SKU-level Excel) + Partner Center API; clawback rules are deterministic ((Revenue − Refunds) × rate) — https://seller-us.tiktok.com/university/essay?knowledge_id=2336057241700098 — 2026-06-08
• Signal 3 (Economic): Indonesia ~515K shops, ~$4.8–5B GMV; commissions 3–30%; effective take climbing to 26.6%; May-2026 minimum-commission + CHR rules expand affiliate spend — https://www.dashboardly.io/post/tiktok-shop-affiliate-commissions-2026-payouts-clawbacks-profit-math — 2026-06-08 Category: Tech-unlock

3. The opportunity

TikTok Shop built a brilliant growth dashboard and a barely-usable finance report. The gap between them is where sellers bleed money.

A seller running 80–300 creators at negotiated rates faces three losses the platform won’t compute for them:

1. Overpayment on returned orders. In Indonesia and the Philippines, many sellers top up creators directly via bank transfer for negotiated flat fees and bonuses outside TikTok’s in-platform commission flow. When those orders refund after payout, the platform recovers nothing — the seller ate a commission on a sale that no longer exists and has no per-creator record telling them so.
2. Disputes with creators. “You underpaid me last cycle” vs. “that order got returned” — with no reconciled statement, every dispute is a manual dig through a 2,000-row CSV.
3. Scaling the wrong creators. Native reports rank by GMV, not net-of-clawback profit, so sellers keep paying high-return-rate creators who look like top performers.

The incumbent here isn’t a competitor — it’s the Seller Center finance tab plus a spreadsheet. That’s a 10× target. Existing third-party tools (below) attack the analytics and bookkeeping edges of this; none produces the operational, creator-by-creator settlement statement a seller can pay against and recover overpayments from.

4. Target market

• Primary customer: Owner-operators and finance/ops leads at TikTok Shop seller brands in Indonesia and the Philippines doing roughly $30K–$500K/month GMV, running an active affiliate/creator program (50–300 creators), with 1–10 staff. The sweet spot is the brand that has graduated past “post and pray” into managed creator collaborations but isn’t big enough to have a finance team.
• Why they buy (in their words): “Each payout bundles dozens of moving parts… without automation it’s easy to misreport”; native reports “can’t tell you who generated the most net profit after returns.” They feel it every settlement cycle — weekly or fortnightly — when the bank deposit is ~67% of reported GMV and they can’t explain the gap per creator.
• Rough TAM reasoning: ~515K shops in Indonesia + a large Philippines base. Conservatively, even 3–5% are programmatic affiliate sellers with reconciliation pain = 15,000–25,000 reachable accounts in just two countries, before Thailand/Vietnam/Malaysia.
• Why now for them: May-2026 rule changes (minimum commissions, CHR) pushed more spend into the affiliate channel and raised the cost of paying the wrong creators. Clawback mechanics that were a rounding error at 10 creators become a real monthly leak at 200.

5. Product sketch (MVP)

• One-click statement ingest: seller uploads (or auto-syncs via Partner Center API) their TikTok Shop settlement + affiliate-order exports; TallyLive normalizes the multi-tab, multi-format Excel into a clean ledger.
• Per-creator reconciliation: every order, refund, and clawback attributed to the exact creator who drove it — including the deterministic (Revenue − Refunds) × rate math the native report hides.
• “What to pay this cycle” statement: a payable, per-creator net figure for sellers who settle creators off-platform, with a downloadable/shareable statement each creator can verify (kills disputes).
• Overpayment recovery flags: highlights commissions already paid on orders that later refunded post-payout — the permanently-lost money — so the seller can claw it back on the next cycle’s bonus or stop the leak.
• Net-profit-per-creator leaderboard: ranks creators by profit after returns and stacked fees, not vanity GMV, so spend flows to creators who actually convert and stick.
• Return-rate watchlist: flags creators whose orders refund above a threshold before the seller renews their collaboration.
• Localized: Bahasa Indonesia and Filipino-English UI; PPh 23 (2%/4%) tax-withholding handling baked into the Indonesian payable figure.

6. AI angle — what’s load-bearing

AI does the dirty ingestion and disambiguation work that makes this a product instead of a macro:

• Statement normalization across changing, inconsistent exports. TikTok’s CSV/Excel schemas vary by country, settlement type, and change over time. An LLM-assisted parser maps “raw columns spread across hundreds of rows” to a stable internal ledger and adapts when TikTok shifts the format — the brittle part that has killed every spreadsheet template.
• Creator-to-payment matching. Off-platform bank transfers, in-platform commissions, and negotiated bonuses must be tied back to one creator identity despite name/handle mismatches — a fuzzy-matching/entity-resolution job.
• Plain-language explanation. “Why is my deposit ₹X lower than my GMV?” answered per creator in the seller’s language.

Remove the AI and you’re back to a fragile spreadsheet that breaks the next time TikTok renames a column. The reconciliation logic is deterministic (that’s the moat-building part); the ingestion and matching is where AI is load-bearing.

7. Localization angle

SEA-first is the wedge, not a nice-to-have:

• Off-platform creator settlement is a regional norm. In Indonesia/Philippines many brands pay creators via direct bank transfer / e-wallet for negotiated deals, so the platform’s in-app commission ledger is incomplete — exactly the gap a US-focused tool ignores.
• Tax rails: Indonesian PPh 23 withholding (2% with NPWP, 4% without) must be in the payable figure; a generic global tool won’t model it.
• Language: Bahasa Indonesia + Filipino-English UI and creator-facing statements.
• Local pricing: a Rp 300K–1.5M/mo ($20–95) tier works where a $99–299 US SaaS price would not.
• Distribution: seller communities live in WhatsApp/Telegram groups and local MCNs, not on G2.

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

• Pricing: tiered by creator count / GMV. Starter ~$25/mo (≤50 creators), Growth ~$69/mo (≤200), Pro ~$149/mo (unlimited + multi-shop + API sync). Annual discount to lock in.
• ACV: blended ~$700–900/yr as sellers land on Growth and expand.
• To $1M ARR: ~1,300 sellers at ~$65/mo blended. Out of 15K–25K reachable affiliate sellers in just ID+PH, that’s 5–9% penetration — achievable.
• To $5M ARR: expand to Thailand/Vietnam/Malaysia, add a per-recovered-dollar success fee on flagged overpayments (sellers happily share upside on money they’d otherwise lose), and an MCN/agency multi-client tier. ~5,500 sellers blended at ~$75/mo, or fewer with success-fee upside.
• Expansion path: creator count → multi-shop → success fee on recoveries → agency seats managing many brands.

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

• Free “Clawback Leak Report.” Seller uploads one settlement export; TallyLive returns a one-page per-creator breakdown showing exactly how much they overpaid on returned orders last cycle. This is the demo and the hook — it quantifies pain in their own numbers. Run it as the lead magnet everywhere below.
• TikTok Shop seller Telegram/WhatsApp/Facebook groups in ID & PH. These are large, active, and full of “why is my payout so low?” posts. Drop the free leak report; convert the ones who reply with a number that shocks them.
• MCNs and TikTok Shop Partner agencies. They manage creator payouts for dozens of brands and feel reconciliation pain at 10× volume. Land 3–5 agencies as design partners; each brings a portfolio of sellers.
• Local TikTok Shop “cuan” creators/educators. The Indonesian seller-education niche (buzzerpanel, gineehub-style accounts) monetizes tutorials; sponsor a “how to stop overpaying creators” walkthrough that ends in the free leak report.
• Top-100 seller lists (FindNiche). Scrape the public top-seller lists, DM a personalized leak report to the mid-tier (the ones big enough to hurt, small enough to lack finance staff).

If I couldn’t picture customer #1–100, this would be a PASS. I can: it’s “upload one export, see your leak, start paying right.” That’s a two-week sprint, not a content calendar.

10. Build complexity — justification

Medium. The reconciliation engine is deterministic business logic over a known (if messy) data model — not research. The genuinely hard parts are (a) robust ingestion of inconsistent, evolving TikTok exports across countries and (b) creator entity-resolution across on- and off-platform payments. Off-the-shelf LLM parsing + a rules engine handle both; no custom models, no hardware, no regulatory approval. A technical pair ships a credible v1 (manual upload, ID+PH, the leak report + payable statement) in 8–12 weeks; API auto-sync and success-fee tooling follow.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Read-only on the seller’s own financial exports; no platform ToS breach (no scraping of others’ data).
Ethical — no harm / dark patterns	✅	Reduces overpayment and creator disputes; helps both sides see honest numbers.
Market exists (evidence above)	✅	515K ID shops, documented native-tool gap, surging affiliate spend.
1–5 person team can build this	✅	Technical pair, off-the-shelf stack.
Launchable with <$50K / ₹40L	✅	No capex; cost is build time + cheap inference.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Felt every settlement cycle; real money leaks. Not existential like an account suspension, so not 17+.
Demand evidence	15	12/15	Strong documented native-tool gaps + huge channel, but no direct “I’d pay $X” quotes yet — hence Medium confidence.
Build feasibility	15	12/15	Deterministic logic + LLM ingestion; messy evolving exports are the only real risk. 8–12 wks.
Distribution clarity	15	12/15	Named channels (seller Telegram/FB groups, MCNs, top-seller lists) + a self-qualifying free leak report.
Revenue mechanics	15	11/15	Clear tiered SaaS + success-fee upside; SEA wallets cap price, so volume-dependent.
Time to first revenue	10	8/10	Leak report → paid in weeks; trial-to-paid funnel is short.
Defensibility	10	6/10	Clawback rule engine + per-creator ledger creates switching cost and compounding data; copyable but accumulates.
Total	100	77/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy (ingestion + entity resolution) · operations-heavy (SEA seller-community GTM, agency partnerships). A local-language operator co-founder materially de-risks distribution.

Key assumptions to validate (3–5)

1. Assumption: Mid-size ID/PH sellers settle a meaningful share of creators off-platform, so reconciliation is genuinely incomplete in the native tool. How to test: 20 seller interviews in ID/PH seller groups — ask how they pay creators today and how they track returns.
2. Assumption: The “free leak report” reliably surfaces a number big enough to convert (i.e., overpayment leakage is non-trivial). How to test: run real settlement exports from 10 volunteer sellers; measure median quantified overpayment.
3. Assumption: Sellers will pay $25–149/mo for this (not expect it free). How to test: pre-sell annual at a discount to the first 15 leak-report recipients who react strongly.
4. Assumption: TikTok export formats are stable enough across ID/PH to parse reliably. How to test: collect 15+ real exports across countries/settlement types; measure parser coverage.

Risk flags

1. Platform dependency: TikTok could ship native per-creator reconciliation, or restrict Partner Center API access. Mitigation: lead with manual upload (no API dependency), and the off-platform-payment reconciliation is something TikTok structurally won’t build.
2. Demand evidence is indirect: the pain is documented by guide-writers and incumbents, not yet by verbatim “I’d pay” quotes — the #1 thing to validate before building.
3. SEA willingness-to-pay: thin wallets cap pricing; success-fee model must carry meaningful revenue.
4. Incumbent encroachment: Dashboardly (per-creator profit analytics) or SettleBridge/LinkMyBooks (bookkeeping reconciliation) could extend into operational creator settlement. Move fast on the SEA + off-platform-payment wedge they don’t own.

14. Structured verdict

Score:                  77/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + SEA-local ops/distribution co-founder
Time to revenue:        6–10 weeks (free leak report → paid)
Capital to launch:      $8–15K (build time + inference)
Top 3 assumptions to validate first:
  1. Off-platform creator settlement is common in ID/PH → 20 seller interviews
  2. Free leak report surfaces convert-worthy overpayment numbers → run 10 real exports
  3. $25–149/mo willingness-to-pay → pre-sell annual to first 15 strong reactors
Kill criteria:
  - Abandon if <30% of 20 interviewed sellers settle any creators off-platform (native tool would suffice)
  - Abandon if median quantified overpayment in the leak report is too small to motivate payment
  - Abandon if TikTok ships native per-creator clawback attribution before v1 launch

15. Next step — 1-week validation sprint

• Day 1–2: Join 8–10 TikTok Shop seller Telegram/WhatsApp/FB groups in Indonesia and the Philippines. Post the offer: “Send me one settlement export, I’ll send back a per-creator overpayment breakdown, free.” Collect real exports.
• Day 3–4: Hand-build the leak report for 8–10 volunteers (manual + scripts — no product yet). Record: median quantified overpayment, parser coverage across formats, and how many settle creators off-platform.
• Day 5: Pre-sell. Offer annual access at a discount to everyone whose leak report shocked them. Go if ≥5 of ~10 pre-pay or commit a card; no-go if <2 — that’s a falsifiable line, not a vibe.