StopajTrace

1. One-liner

Rebuilds which marketplace withheld your 1% stopaj, so your accountant can actually claim the offset.

2. Trend signal — why now?

Since 1 January 2025, every Turkish marketplace — Trendyol, Hepsiburada, N11, Amazon Türkiye, Çiçeksepeti, PttAVM — is required to withhold 1% stopaj from each payout to its sellers and remit it to the tax office on the seller’s behalf. This is not a tax. It is an advance against the seller’s income or corporate tax, and the seller is legally entitled to offset it — quarterly on the geçici (provisional) return, and annually on the gelir/kurumlar return. Anything left over after offset is refundable in cash.

The catch is that the offset only happens if somebody claims it. And claiming it requires proving which withholding belongs to which period and which payer. That is where it falls apart.

The mechanism has a specific, documented defect. When a seller logs into the Dijital Vergi Dairesi to see the withholdings filed against their tax number, the withholding company’s tax number and trade name are displayed masked with asterisks. A practicing Turkish CPA writing in muhasebetr put it directly: “Dijital Vergi Dairesi’nde kesinti yapan firmaların vergi numarası ve unvanların yıldızlı (maskelenmiş) şekilde gösterilmesi nedeniyle, kesintinin hangi platform veya hangi işlemden kaynaklandığını tespit etmek zorlaşmaktadır.” The seller can see that money was withheld. They cannot see by whom or against what.

The same author notes sellers cannot obtain a formal stopaj certificate from the platforms at all — they are left reconciling from bank statements by hand, which he calls a labour problem for accountants managing volume.

Meanwhile the recovery path for returns is worse. Turkish e-commerce return rates are high, and the guidance from multiple Turkish tax practices is consistent: if a product is returned after the withholding has been declared to the tax office, the platform makes no correction. The seller must recover it through their own offset or refund claim, and must hold a return invoice (for business buyers) or a gider pusulası (for consumers), plus proof the platform reported via BTRANS. One CPA lists four scenarios where the stopaj is simply lost: return processed without documentation, no invoice or expense slip issued, no deduction claimed on the return, or the platform never notified the authority.

Then the volume. Trendyol alone runs 125,000+ sellers. Turkish online shopping volume passed 2 trillion TL in 2026, roughly 60–70% of it through marketplaces. The instruction sellers get from their accountants is to forward the monthly hakediş ve kesinti raporları from every platform — with the warning attached: “aksi hâlde mahsup hakkınızı kullanamayabilirsiniz” (otherwise you may not be able to use your offset right). A seller on four marketplaces has four different report formats to pull, every month, forever.

Nobody is selling the reconstruction. The Turkish marketplace-integration vendors — PraPazar, ConnectProf, Zeisoft — sync products, stock, orders and invoices into accounting systems like Logo, Mikro and Eta. They move the order. They do not rebuild the withholding trail. Paraşüt at 150 TL/month does pre-accounting and e-invoicing with marketplace integrations, but the offset claim is not its job.

So the money is withheld automatically and reclaimed manually. That asymmetry is the whole business.

Provenance:
  - Signal 1 (demand): Turkish CPAs publicly warning sellers they will lose their offset right if they don't forward monthly platform withholding reports; documented four failure modes where stopaj on returns becomes unrecoverable — https://umutakpinar.com/e-ticarette-iade-edilen-urunlerde-stopaj-kesintisi-nasil-geri-alinir/ — 2026-08-26
  - Signal 2 (feasibility): Dijital Vergi Dairesi masks the withholding company's tax number and trade name, forcing manual reconstruction from bank statements — a matching problem, now cheap to automate — https://www.muhasebetr.com/yazarlarimiz/berkaysenol/005/ — 2026-08-26
  - Signal 3 (economic): 1% withholding applies across all Turkish marketplaces since Jan 2025; Trendyol alone has 125,000+ sellers, marketplace volume 60–70% of a 2T TL online market — https://www.dopigo.com/2026-e-ticaret-istatistikleri-turkiye-ve-dunya-verileri/ — 2026-08-26
  Category: Regulatory arbitrage

3. The opportunity

The gap is between automatic deduction and manual recovery.

Every peer in this workflow is optimized for the wrong direction. The marketplace withholds automatically — it has to, it’s the withholding agent. The tax office receives automatically. But the seller’s claim is a manual assembly job, and the one system that should make it easy (Dijital Vergi Dairesi) masks exactly the field you need to do the matching.

The incumbents each solve an adjacent problem and stop short:

• Marketplace integrators (PraPazar, ConnectProf, Zeisoft, Wisersell). They own the API connections into Trendyol, Hepsiburada and N11 already. They pull orders, stock, shipments, commission invoices. They do not produce a period-matched withholding file, because their buyer is the operations manager, not the accountant, and the offset happens in a system they don’t touch.
• Pre-accounting SaaS (Paraşüt, Bizim Hesap, Mükellef). Cheap, well-distributed, marketplace-integrated. But they’re built around issuing invoices and tracking receivables. The stopaj offset is a once-a-quarter, once-a-year event handled by the seller’s mali müşavir on a different system entirely.
• The mali müşavir (independent CPA). Willing but under-supplied. They receive whatever the client forwards. If the client is on four marketplaces and forwards two, the CPA claims two. There is no tool that tells the CPA what’s missing.

Nobody owns the artifact: a per-period, per-platform withholding ledger that reconciles to the seller’s declared revenue and hands the CPA a claim-ready file. That is what StopajTrace sells.

The 10× is not intelligence, it’s completeness. A seller doing 6M TL/year through marketplaces has roughly 60,000 TL withheld annually. If they claim 70% of it because two platforms’ reports never got forwarded, and returns eat another slice undocumented, they’re leaving 15–20K TL a year on the table — every year, silently, with no error message. The product’s job is to make the leak visible and then close it.

4. Target market

Primary customer: Turkish marketplace sellers — limited şirket (Ltd. Şti.) or şahıs şirketi — doing 3M–50M TL annual GMV across 2 or more marketplaces, typically 1–15 staff, with an external mali müşavir rather than an in-house accountant. Categories skew apparel, home goods, cosmetics, electronics accessories — the high-return categories where the post-declaration return problem bites hardest.

Secondary customer: the mali müşavir practice itself — small CPA firms in Istanbul, Izmir, Bursa carrying 20–80 e-commerce clients. They feel the pain as labour, not as lost money, and they’re a better distribution channel than the sellers.

Why they buy, in their words: The recurring instruction from Turkish accountants to sellers is to send the monthly hakediş ve kesinti raporları or lose the offset right. Sellers know they’re supposed to. Practitioners describe manual extraction from bank statements as the actual method, and describe the tax portal’s masked payer fields as making it hard to determine which platform a deduction came from. Nobody is disputing the pain — the profession is writing about it.

Rough TAM reasoning: Trendyol reports 125,000+ sellers. Total Turkish marketplace sellers across all platforms is meaningfully larger, but most are hobby-scale and never clear a threshold worth automating. The realistic serviceable segment is sellers with enough GMV that 1% is real money and enough platforms that reconciliation is genuinely hard: call it 25,000–40,000 businesses, plus a few thousand CPA practices serving them. At 750 TL/month, capturing 1,200 of them is roughly 10.8M TL ARR (~$260K). That’s the shape of the business — modest in dollars, healthy in a Turkish cost base.

Why now for them: The 1% has been running since January 2025. The first full annual cycle closed with the 2025 kurumlar/gelir returns filed in 2026 — which is the moment sellers found out what they failed to claim. Second cycle is now accruing. This is the year the problem stops being theoretical.

5. Product sketch (MVP)

• Multi-marketplace payout ingestion. Connect Trendyol, Hepsiburada, N11, Amazon Türkiye and Çiçeksepeti seller accounts; pull settlement/hakediş statements automatically each month instead of the seller downloading five files.
• Withholding extraction and normalization. Every platform formats its deduction lines differently. Parse each into one schema: gross base, commission, shipping, service fees, and the stopaj line — with shipping and bank charges correctly excluded from the withholding base per the rules.
• Masked-payer reconstruction. Import the seller’s Dijital Vergi Dairesi withholding list — where the payer is starred out — and match each masked entry against the platform settlement data by amount, date and period. Output: a named, sourced ledger showing which marketplace filed which deduction.
• Gap alarm. Flag the withholdings visible at the tax office that don’t match any ingested platform statement — the ones the seller would otherwise never claim — and the platform statements with no corresponding filing (the BTRANS failure case).
• Return-exposure tracker. Separate returns processed before declaration (platform self-corrects, no action) from returns after declaration (seller must claim), and list exactly which documents each one needs — return invoice or gider pusulası.
• Quarterly claim pack. One file per geçici vergi period and one for the annual return: period-matched withholding totals, per-platform breakdown, supporting statements, and a delta line showing claimed vs. available.
• CPA handoff mode. A read-only workspace where the mali müşavir sees all their StopajTrace clients, with a per-client “ready / missing documents” status.

6. AI angle — what’s load-bearing

Two places, both real work.

The masked-payer match is the core. The tax office shows an amount, a period and a starred-out payer. The platform statements show payouts, deductions and dates that don’t line up one-to-one — a single monthly filing aggregates many payouts, timing shifts across period boundaries, and returns net against later cycles. This is a fuzzy many-to-many reconciliation against an adversarially incomplete reference. Rules alone break on the edge cases; a model that scores candidate matches, explains its reasoning, and flags low-confidence assignments for human review is doing the job a junior accountant does by hand today, in seconds instead of hours. Remove it and the product is a folder of CSVs.

Statement parsing is the second. Five marketplaces, each with its own hakediş format, each of which changes without notice. Hand-written parsers are a permanent maintenance tax; a model that reads a statement and maps columns to the normalized schema — with a validation pass that checks totals reconcile — is what keeps a 1–3 person team from drowning in format churn.

What’s not AI: the offset itself. That’s a fixed rule on a fixed form. Don’t pretend otherwise.

7. Localization angle

This is a Turkey-only product by construction, and that’s the point rather than a limitation.

The regulation (1% tevkifat on intermediary payments), the platforms (Trendyol, Hepsiburada, N11), the reporting artifacts (hakediş ve kesinti raporu, gider pusulası, BTRANS), the filing calendar (geçici vergi quarterly, annual kurumlar/gelir), and the portal defect (masked payer fields in Dijital Vergi Dairesi) are all specifically Turkish. A global product cannot be pointed at this.

Consequences for the build:

• Turkish-first UI and Turkish-first support. The buyer is a seller-operator or a mali müşavir; English is not the working language of this workflow.
• Local pricing. 500–1,500 TL/month is the right band — comparable to what sellers already pay for pre-accounting and integration tools, and trivially justified against a five-figure TL annual recovery.
• Local payment. Turkish card acquiring and e-fatura issuance for the subscription itself, because a seller who can’t get a proper invoice from you can’t expense you.
• The mali müşavir is the distribution channel. In Turkey the CPA is the trusted advisor on anything touching a beyanname. Building for them isn’t a nice-to-have; it’s the go-to-market.

The same shape — automatic withholding, manual recovery — recurs in other marketplace-withholding regimes, so there’s a template here later. But shipping is Turkey.

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

Pricing:

Tier	Price/mo	Who
Satıcı	500 TL	1–2 marketplaces, under 5M TL GMV
Satıcı Pro	900 TL	3+ marketplaces, return-exposure tracking, quarterly claim packs
Müşavir	4,500 TL	CPA practice, up to 25 client workspaces (+150 TL per extra client)

ACV: 9,000 TL for a seller ($215), 54,000 TL for a CPA practice ($1,300).

Rough math to $1M ARR (~42M TL): roughly 3,000 seller accounts at Pro, or a blend — 1,800 sellers plus 350 CPA practices gets there with a much smaller logo count, which is why the CPA channel matters. Realistically this is a 24–30 month path, not a 12-month one.

Rough math to $5M ARR: requires either regional expansion into a second withholding regime, or moving up the stack from reclaiming the withholding to owning the full marketplace tax file — VAT on commission invoices, e-arşiv issuance, the return VAT adjustments the CPAs already flag as commonly botched. The second is more natural: same data, same buyer, larger job.

Expansion path: starts per-seller, grows by (a) marketplaces connected, (b) GMV band, (c) CPA seats. The genuinely attractive expansion is CPA-led — one practice that adopts it brings 20–80 sellers who each become a billable workspace.

Unit economics note: gross margin is high — API pulls, parsing, storage. The real cost is Turkish-language support during the two annual filing crunches. Budget for a support hire earlier than a pure-SaaS model suggests.

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

1. The free stopaj audit as the entire top of funnel. Offer a one-off reconciliation: seller uploads last year’s Dijital Vergi Dairesi withholding list and their platform statements, gets back a report showing how much stopaj is visible at the tax office that they cannot trace to a platform — i.e. money they probably didn’t claim. This is a number, in TL, about their own business. Nothing converts like that. Target 300 audits to land the first 60–80 paying accounts.

2. Mali müşavir practices in Istanbul, Izmir and Bursa carrying e-commerce books. These are findable — TÜRMOB chamber directories list members by city, and the CPAs writing publicly about e-commerce stopaj (muhasebetr, LinkedIn, personal practice blogs) are self-identifying as the ones who care. Approach 200 practices with the audit run on two of their clients, free. A practice that sees 40K TL of unclaimed stopaj across two clients adopts for the rest. Each converted practice is worth 20–80 seller workspaces.

3. The e-commerce seller communities where the complaint already lives. Turkish seller Facebook groups, Trendyol/Hepsiburada seller forums, and the e-ticaret YouTube/Instagram advisor circuit. Don’t advertise — publish the reconciliation methodology, including how to do it by hand. The manual version is tedious enough that publishing it sells the automated version.

4. Timed to the filing calendar. Geçici vergi quarters and the annual return in Q1–Q2 are the moments this becomes urgent. Run the audit campaign 6 weeks ahead of each deadline; go quiet in between. Seasonality is a feature — it tells you exactly when to spend.

5. Integration-vendor partnerships as a later lever. PraPazar, Zeisoft and similar already hold the marketplace connections and the seller relationships. They have no reason to build the offset ledger and a good reason to resell it. Not a launch channel — a month-9 channel, after the audit report has proof.

10. Build complexity — justification

Low. Five marketplace seller-API integrations for settlement data, an LLM-backed statement parser with a totals-reconcile validation pass, a matching engine against the tax portal export, and a reporting layer that emits claim packs. No custom models, no infrastructure novelty, no hardware.

The two real costs are (a) the marketplace API work, which is well-trodden — multiple Turkish vendors already do it and Trendyol publishes developer docs — and (b) getting the matching logic genuinely right on returns and period boundaries, which is domain work, not engineering difficulty. A technical founder with a Turkish CPA advisor ships a credible v1 in 10–12 weeks. The first version can even skip live API integration and accept uploaded statements — that’s a 5-week MVP and it’s enough to run the free-audit funnel.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Helps sellers claim an offset they’re statutorily entitled to. Reads platform data with the seller’s own credentials/consent. Not tax advice — output goes to a licensed mali müşavir who files.
Ethical — no harm / dark patterns	✅	Recovers legitimately owed money. No aggressive positions, no evasion.
Market exists (evidence above)	✅	1% withholding live since Jan 2025 across all Turkish marketplaces; Trendyol alone 125K+ sellers; CPAs publicly documenting the offset-loss problem.
1–5 person team can build this	✅	10–12 weeks for two people. Solo-buildable if the CPA advisor is engaged.
Launchable with <$50K / ₹40L	✅	Turkish cost base. Under $20K to a paying v1.

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Real money, recurring, quantifiable — a mid-size seller leaves five figures TL/year unclaimed. Docked because it’s a silent loss: no penalty, no notice, nothing breaks. Sellers who don’t know they’re losing it don’t feel urgency until you show them the number. That’s a demand-creation cost, not a hair-on-fire pull.
Demand evidence	15	12/15	Strong indirect evidence: multiple independent Turkish CPAs publishing warnings, a documented portal defect, explicit practitioner complaints about manual bank-statement extraction, large seller counts. Docked for what I could not find — direct seller complaint threads. The profession is talking; I couldn’t verify the sellers are.
Build feasibility	15	13/15	Off-the-shelf everywhere. Marketplace APIs are documented and multiple vendors have proven the integration path. Upload-only v1 in ~5 weeks.
Distribution clarity	15	12/15	The free stopaj audit is a genuinely strong wedge — it returns a TL figure about the prospect’s own business. The CPA channel is named, enumerable via chamber directories, and has correct incentives. Docked because CPA firms are conservative buyers and the seller-direct motion still needs testing.
Revenue mechanics	15	11/15	Pricing is anchored against real comparables (Paraşüt 150 TL, Bizim Hesap 870 TL) and trivially justified by recovery amounts. Docked hard on currency: TL revenue with lira volatility makes dollar-denominated ARR targets genuinely unstable, and $1M ARR needs ~3,000 accounts, which is a lot of logos for this segment.
Time to first revenue	8	8/10	The audit converts fast because it produces a number. Realistic first payment 6–8 weeks post-launch. Not faster, because the buyer wants to see the claim survive one filing cycle.
Defensibility	4	4/10	Honest score. The matching logic is the moat and it’s a months-long lead, not a permanent one. Any marketplace integrator can bolt this on — they already hold the connections and the customers. Real defense is CPA-practice lock-in and accumulated per-platform parsing edge cases. Thin.
Total	100	76/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · sales-heavy

Needs someone who can build five marketplace integrations and a matching engine, paired with someone who can sell to Turkish CPA practices in Turkish. A Turkish mali müşavir as co-founder or first hire is close to mandatory — not for the code, for the credibility and the channel.

Key assumptions to validate

1. Assumption: A meaningful share of marketplace sellers are actually under-claiming their stopaj — the loss is real, not theoretical. How to test: Run the free audit manually on 15 real sellers across 2+ marketplaces. Measure claimed vs. available on their last filed return. If the median gap is under 5%, the product has no story.
2. Assumption: The Dijital Vergi Dairesi withholding list can be exported in a form usable for matching, and the masking is on payer identity only — not amounts or periods. How to test: Get three sellers to pull their own export and inspect the fields directly. This is a hard dependency; if amounts or periods are also obscured, the matching approach dies and the product becomes platform-statement-only.
3. Assumption: Mali müşavir practices will adopt a client-facing tool rather than treating reconciliation as billable manual work they’d rather keep. How to test: 20 structured conversations with e-commerce-heavy practices. Ask what they currently bill for this. If it’s a profitable line item, they’re a competitor, not a channel.
4. Assumption: Marketplace settlement data is retrievable via seller API access at sufficient granularity to isolate the stopaj line. How to test: Build against Trendyol’s published developer API first; confirm the deduction breakdown is exposed and not merely a net payout figure.

Risk flags

1. Platform dependency (high). Five marketplace APIs, each of which can restrict access, change formats, or decide seller-data extraction isn’t permitted. Mitigation: the upload-based path must remain a first-class mode permanently, not a fallback.
2. Regulatory reversal (medium-high). The cleanest fix to this whole problem is the tax authority unmasking the payer fields — which is exactly what the CPA in the source article publicly recommended. If Dijital Vergi Dairesi ships transparent withholding reporting, the core matching value evaporates overnight. This is a product built on a bureaucratic defect, and defects get patched.
3. Incumbent absorption (medium). PraPazar, Zeisoft and the pre-accounting vendors already own the connections and the customers. A stopaj-report feature is a quarter of work for them once the category is proven. The 6-month head start is real; the 24-month one isn’t.
4. Currency (medium). Lira-denominated ARR against dollar-denominated infrastructure costs. Price in TL with periodic revisions, or the margin erodes without a single churned customer.
5. Demand-creation cost (medium). The customer doesn’t wake up looking for this. Every sale starts by proving a loss they didn’t know they had. That’s a longer, more expensive funnel than a product people search for.

14. Structured verdict

Score:                  76/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + Turkish mali müşavir co-founder or first hire
Time to revenue:        6–8 weeks post-launch
Capital to launch:      $15–20K (₺600K–800K)
Top 3 assumptions to validate first:
  1. Median unclaimed stopaj gap ≥5% — manual audit of 15 real multi-marketplace sellers
  2. Dijital Vergi Dairesi export masks payer identity only, not amounts/periods — inspect 3 real exports
  3. CPA practices treat this as a tool to adopt, not billable work to protect — 20 practice interviews
Kill criteria:
  - Abandon if the median unclaimed gap across 15 audited sellers is under 5% of withheld stopaj
  - Abandon if the tax office publishes unmasked payer detail in withholding reports
  - Abandon if fewer than 3 of 20 CPA practices will pilot after seeing a live audit on their own clients
  - Abandon if marketplace settlement data proves unavailable at deduction-line granularity on both API and export paths

15. Next step — 1-week validation sprint

Day 1–2 — Confirm the artifact. Recruit three cooperating marketplace sellers on 2+ platforms. Have each pull their Dijital Vergi Dairesi withholding export and their platform hakediş statements for the same period. Confirm exactly what the masking hides. If periods and amounts are intact, the matching approach is viable — that’s the single most load-bearing fact in this proposal.

Day 3–4 — Measure the leak by hand. For those same three sellers plus a dozen more recruited through a Turkish seller community, manually reconcile last year: withholding visible at the tax office vs. withholding actually claimed on the filed return. Do it in a spreadsheet. The output is one number per seller — the gap. This is the product’s pitch, so if it can’t be produced manually it can’t be automated.

Day 5 — Take it to the channel. Walk the results into 8–10 mali müşavir practices in Istanbul with e-commerce clients. Not a demo — a finding: here is what your peers’ clients are leaving unclaimed. Ask two questions: would you use a tool that produced this, and do you currently bill for doing it manually.

Go / no-go: proceed only if (a) the median gap across audited sellers is ≥5% of withheld stopaj, (b) the tax portal export preserves amounts and periods, and (c) at least 3 of 10 CPA practices commit to a pilot. Any one of those failing means the idea is a different, smaller product — or not one at all.