PotongBack — withholding reconciler for Indonesian marketplace sellers

1. One-liner

Turns every marketplace tax slip into a filed credit so Indonesian sellers stop losing 0.5% to the taxman.

2. Trend signal — why now?

On 1 August 2026 Indonesia’s tax office (DJP) forces Shopee, Tokopedia, Lazada, and Blibli to withhold 0.5% PPh Pasal 22 of gross revenue from every domestic seller and to issue an electronic bukti potong (withholding proof) per transaction. This is not a proposal — it’s a dated, mandatory event that flips on for millions of sellers at once. Two distinct pains land the same week:

1. Sellers above IDR 500M/year get money skimmed at source. To not eat it, they must collect every bukti potong across every marketplace, reconcile marketplace report ↔ bank payout ↔ own bookkeeping, and enter the total as a tax credit on their annual SPT 1770/1771. Miss a slip → lose the credit → effectively double-taxed. A mismatch between the marketplace report and the bank → audit trigger.
2. Sellers under IDR 500M must generate and upload a Surat Pernyataan (statement letter, PMK 164/2023 format) with a valid NPWP/NIK to each marketplace to stay exempt — and re-file the moment they cross the threshold mid-year, or the withholding kicks in the next month.

The Indonesian trade and accounting press is already in dread mode about the reconciliation, not the tax rate itself. temanakuntan spells out the exact three-step monthly routine and warns that “discrepancies between marketplace reports and bank records could trigger audits.” ddtc, bisapajak, and Bloomberg Technoz are all running seller how-to explainers weeks before the deadline. The demand is being manufactured by the regulator on a fixed date.

Provenance:

• Signal 1 (Demand): Indonesian accounting/trade press running seller-panic explainers on the monthly 3-way reconciliation (“small mismatch triggers audit”); existing paid tooling (Mekari Jurnal Rp299K/mo) proves willingness to pay — https://temanakuntan.com/article/188/pajak-marketplace-aturan-dan-cara-umkm-aman , https://bisapajak.id/pajak-e-commerce-1-agustus-2026/ — 2026-07-01
• Signal 2 (Feasibility): Cheap Bahasa-capable LLMs + marketplace CSV/PDF export parsing + bank-statement OCR are all off-the-shelf in 2026; the bukti potong is a structured electronic doc, and the Surat Pernyataan has a fixed government template — https://www.jurnal.id/en/ — 2026-06
• Signal 3 (Economic): Indonesia e-commerce ~USD 104B in 2026 growing 15%+ CAGR; DJP designates 4 marketplaces as PPh-22 collectors effective 1 Aug 2026, a state-forced compliance event across a market of millions of sellers — https://www.taxspoc.com/articles/indonesia-digital-tax-marketplace-collection-august-2026-2/ , https://news.ddtc.co.id/berita/nasional/1820505/merchant-cukup-buat-1-surat-pernyataan-omzet-untuk-semua-marketplace — 2026-07-01 Category: Regulatory arbitrage

3. The opportunity

The incumbents (Mekari Jurnal, Accurate, BukuWarung, Kledo) are general bookkeeping apps. They record transactions and, at best, sync marketplace orders. None is built around the workflow that was born on 1 Aug 2026: catching every withholding slip and converting it into a claimed tax credit.

That’s the gap. A seller on Jurnal still has to manually chase down each bukti potong, eyeball whether the marketplace’s stated gross matches what actually hit their bank, and hand-enter the PPh-22 total into their SPT. The 0.5% withheld is real cash the seller has already paid the government — if they don’t claim it correctly they lose it. On IDR 800M annual omzet that’s IDR 4,000,000 (~USD 250) of the seller’s own money left on the table, every year, if the credit isn’t filed cleanly.

PotongBack does one narrow thing 10× better than a general ledger: it owns the withholding-to-credit loop. Pull the marketplace exports and the bank statement, auto-match them line by line, collect every bukti potong into one place, flag the mismatches that trigger audits before the seller files, and emit a filing-ready PPh-22 credit summary — or, for the small seller, the correct Surat Pernyataan. It’s not trying to replace the accountant’s ledger. It’s the thing that sits between the marketplace and the SPT and makes sure no rupiah of withheld tax goes unclaimed.

4. Target market

• Primary customer: Indonesian individual / small-UMKM online sellers doing roughly IDR 300M–3B/year across 2+ marketplaces (Shopee, Tokopedia/TikTok Shop, Lazada, Blibli). The sweet spot is the seller just over — or nervously near — the IDR 500M exemption line, because they’re the ones now getting withheld and the ones who feel the double-tax risk most.
• Why they buy (in their words): “Aku dipotong 0,5% tiap bulan, tapi gak ngerti cara ngeklaim balik pas lapor SPT” — I’m getting cut 0.5% every month but I don’t understand how to claim it back at filing. And: reconciling three sources by hand every month is dreadful, and one wrong number invites an audit.
• Rough TAM reasoning: Indonesia has millions of marketplace sellers; even the slice above IDR 500M (the ones actively withheld) is in the hundreds of thousands, and the near-threshold cohort that must file exemption letters is far larger. Capturing tens of thousands of paying sellers is a comfortable sub-$5M-ARR business without needing the whole market.
• Why now for them: The withholding switches on 1 Aug 2026. Before that date this product had no reason to exist. After it, every affected seller has a recurring, money-at-stake chore they didn’t have last quarter.

5. Product sketch (MVP)

• Connect Shopee / Tokopedia / Lazada / Blibli seller accounts (start with CSV/PDF export upload; API where available) and pull the monthly transaction + fee + bukti potong data.
• Upload or connect a bank statement; auto-match marketplace-reported gross and net payouts against actual bank deposits, line by line.
• One dashboard showing, per month per marketplace: gross omzet, 0.5% PPh-22 withheld, fees deducted, and net that should have hit the bank — with every mismatch flagged in red before the seller files.
• Auto-collect and archive every bukti potong into a single, searchable vault so nothing gets lost at year-end.
• Generate a filing-ready PPh-22 credit summary mapped to the SPT 1770/1771 fields, so the seller (or their accountant) enters the credit correctly and recovers the withheld cash.
• For sub-IDR-500M sellers: one-tap Surat Pernyataan generator in the official PMK 164/2023 format, with a running omzet tracker that warns before they cross the threshold and re-files when they do.
• Bahasa-first UI, WhatsApp reminders at each marketplace’s monthly report-release date.

6. AI angle — what’s load-bearing

Load-bearing in two places. First, document extraction and normalization: every marketplace exports a differently-shaped report and issues bukti potong in slightly different electronic formats, and bank statements are messy PDFs — an LLM/vision layer normalizes all of it into one reconcilable ledger without hand-built parsers per format that break every time a marketplace tweaks its export. Second, mismatch reasoning and explanation: the AI classifies why a marketplace-vs-bank number differs (fee, refund, hold, split payout, timing) and explains it to a non-accountant seller in plain Bahasa, and drafts the plain-language notes an auditor would want. Strip the AI out and you’re back to hand-built brittle parsers plus a seller staring at three spreadsheets — which is exactly the status quo this replaces.

7. Localization angle

This is inherently Indonesia-first and can’t be a generic global product. It’s welded to one country’s tax code (PPh Pasal 22, PMK 37/2025, PMK 164/2023), one set of marketplaces (Shopee/Tokopedia/Lazada/Blibli/TikTok Shop), one filing system (SPT 1770/1771, Coretax), one language (Bahasa), and one price band (Rp99K–299K/mo works where $49/mo would be absurd). The regulatory specificity is the moat and the wedge — a global reconciliation SaaS can’t touch this without rebuilding for the local rules. Adjacent SEA markets with parallel platform-withholding moves (e.g. the Philippines, Vietnam) are the expansion path, each a fresh localized rebuild rather than a toggle.

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

• Pricing: Rp99,000/mo (single marketplace, sub-500M / exemption tier) → Rp199,000/mo core (multi-marketplace reconciliation + credit summary) → Rp399,000/mo pro (bank sync, accountant export, multiple stores). Annual filing-season add-on possible.
• ACV: ~Rp199K/mo core ≈ IDR 2.4M/year (~USD 150) blended.
• Rough math to $1M ARR: ~USD 1M ≈ IDR 16B. At IDR 2.4M ACV → ~6,700 paying sellers. In a market of hundreds of thousands of withheld sellers, that’s low-single-digit penetration.
• Rough math to $5M ARR: ~33,000 paying sellers, or the same ~6,700 at a higher blended ACV via the pro tier + an accountant/multi-store plan + a second SEA market. Requires proving retention past the first filing season.
• Expansion path: ACV grows via bank sync, multi-store, accountant seat licensing (small tax firms managing dozens of seller clients — a natural B2B2C upsell), and adjacent-country editions.

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

• Ride the panic content. Indonesian accounting bloggers and tax micro-influencers (temanakuntan, bisapajak, ddtc followers) are already publishing “how not to get cut / how to claim back” explainers. Sponsor or co-produce the “how to actually claim your PPh-22 credit” explainer with 3–5 of them and put PotongBack as the one-click tool at the end. This audience is self-selected withheld sellers, arriving warm, right at the pain moment.
• Free Surat Pernyataan generator as the top of funnel. Ship the exemption-letter generator free and ungated. Every sub-500M seller needs it before 1 Aug; it’s shareable, it captures NPWP/omzet context, and it converts the ones who later cross the threshold into paid reconciliation users.
• Seller Facebook/Telegram/WhatsApp groups. Indonesian marketplace sellers cluster in large “Komunitas Seller Shopee/Tokopedia” groups. Drop a genuinely useful reconciliation checklist + the free tool; these groups spread tools fast when the pain is universal and dated.
• Accountant channel. Small tax-prep firms are about to be buried in seller SPTs with PPh-22 credits. Offer them a multi-client console — they onboard their whole seller book at once.

10. Build complexity — justification

Medium. Off-the-shelf: Bahasa LLM + vision extraction, standard web stack, bank-statement OCR, the fixed government Surat Pernyataan template. Custom work: per-marketplace export ingestion (formats drift and must be maintained), the reconciliation matching engine, and mapping outputs to the SPT 1770/1771 fields correctly — the last requires real Indonesian tax-domain accuracy, which is where a wrong answer is dangerous. A small team with a local tax advisor ships a credible v1 in ~10–14 weeks, since the deadline forces a tight, well-defined scope.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Helping sellers file correctly; not tax advice if positioned as prep/reconciliation with accountant hand-off
Ethical — no harm / dark patterns	✅	Recovers money sellers are legally owed; exemption letter must not be used to under-declare (guardrail: omzet tracker warns at threshold)
Market exists (evidence above)	✅	State-forced event 1 Aug 2026 across millions of sellers; incumbents charge for adjacent tooling
1–5 person team can build this	✅	~10–14 weeks with a tax advisor
Launchable with <$50K / ₹40L	✅	Content + community GTM, off-the-shelf AI

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Real money lost + audit risk, recurring monthly, but the seller can limp along in a spreadsheet — not quite hair-on-fire for the smallest sellers
Demand evidence	15	12/15	Strong: forced event + active seller-panic press + existing paid adjacent tools. Docked because no direct “I’d pay for this exact product” quote yet
Build feasibility	15	11/15	Doable in ~12 weeks but per-marketplace ingestion + SPT-field accuracy is fiddly and unforgiving
Distribution clarity	15	12/15	Warm, dated, self-selected audience via panic content + free letter tool + seller groups
Revenue mechanics	15	12/15	Clear pricing benchmarked to Jurnal; ~6,700 sellers for $1M ARR is realistic; retention past filing season is the open question
Time to first revenue	10	8/10	Deadline creates urgency; free tool → paid conversion inside weeks of 1 Aug launch
Defensibility	10	3/10	Regulatory-knowledge + ingestion maintenance is a real but soft moat; a funded incumbent (Mekari) could bolt this on. Speed and niche focus, not a durable wall
Total	100	74/100

13. Qualitative modifiers

Founder-fit tags

domain-expertise-required · technical-heavy — needs genuine Indonesian PPh/SPT knowledge (or a local tax advisor as co-founder) plus solid extraction/reconciliation engineering.

Key assumptions to validate (3–5)

1. Assumption: Withheld sellers will pay ~Rp199K/mo to recover the credit and avoid audit-triggering mismatches. How to test: Pre-sell to 30 sellers in Shopee/Tokopedia seller communities with a paid waitlist before the 1 Aug switch-on.
2. Assumption: Marketplace exports + bukti potong are structured/consistent enough to reconcile reliably. How to test: Get 10 real sellers’ actual August exports and bank statements; measure auto-match accuracy.
3. Assumption: Retention holds after the first annual filing (not a one-shot seasonal buy). How to test: Track monthly-active reconciliation usage between filing seasons in the first cohort.
4. Assumption: The accountant multi-client channel converts. How to test: Pitch 10 small tax-prep firms on a client console before build-out.

Risk flags

1. Platform dependency: Reliant on marketplace export formats/APIs that Shopee et al. can change unilaterally, breaking ingestion. Mitigation: LLM-based extraction over hand-parsers, fast turnaround on format drift.
2. Incumbent bolt-on: Mekari Jurnal / Accurate / BukuWarung already have the sellers and DJP integration; any of them could ship a PPh-22 credit module and compress the niche. Speed and singular focus are the only defense.
3. Regulatory drift / marketplace absorption: DJP could push more of the reconciliation into the marketplaces or Coretax itself (e.g. marketplaces pre-compute the credit), shrinking the manual gap this product fills. Market-timing risk cuts both ways.
4. Ethics guardrail: The Surat Pernyataan feature must actively discourage under-declaration; a tool that helped sellers falsely claim exemption is both illegal for them and reputationally toxic.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + Indonesian tax-domain co-founder/advisor
Time to revenue:        6–10 weeks (launch timed to the 1 Aug 2026 switch-on)
Capital to launch:      ₹8–15 lakh / $10–18K
Top 3 assumptions to validate first:
  1. Withheld sellers pay ~Rp199K/mo — paid waitlist of 30 pre-1-Aug
  2. Exports + bukti potong reconcile reliably — 10 real August datasets, measure match accuracy
  3. Retention survives past first filing — track between-season MAU in cohort 1
Kill criteria:
  - Abandon if <10% of 50 target sellers convert on the paid waitlist before 1 Aug
  - Abandon if auto-match accuracy on real marketplace+bank data stays below ~90%
  - Abandon if Mekari or a marketplace ships a native PPh-22 credit/reconciliation feature before your v1

15. Next step — 1-week validation sprint

• Day 1–2: Ship the free Surat Pernyataan generator + a landing page with the reconciliation-and-credit pitch and a Rp199K/mo paid-waitlist button. Post the free tool into 5 large Shopee/Tokopedia seller communities.
• Day 3–4: DM 50 sellers who posted about the 0.5% cut; collect 10 real sets of August marketplace exports + bank statements; hand-run the reconciliation to measure how consistent the data actually is.
• Day 5: Go/no-go. Green light only if ≥5 sellers put down a paid-waitlist deposit AND the hand-run reconciliation matches cleanly on ≥90% of lines across the 10 datasets. Anything less = the data’s too messy or the pain isn’t pay-worthy yet — hold.

The falsifiable result: a paid-deposit count and a measured line-match accuracy, not “sellers seemed interested.”