SB StartupBasket
All ideas
69 /100 VALIDATE Medium complexity

Tutanak — penalty objection desk for Trendyol sellers

Turns Trendyol's own shipment history into a filed penalty objection with evidence, inside the window, and tracks the refund.

Retail SaaS Turkey TurkeyMarketplace-sellerPlatform-penaltyAI-agentSMBSolo-builderCompliance-driven
— views
Evaluation Scores
69/100

VALIDATE

Overall Score

16
Problem
11
Demand
12
Build
10
Distrib.
9
Revenue
7
Time
4
Defense

Tutanak — penalty objection desk for Trendyol sellers

1. One-liner

Turns Trendyol’s own shipment history into a filed penalty objection with evidence, inside the window, and tracks the refund.

2. Trend signal — why now?

Trendyol fines its sellers by invoice. A late handover, an “unsupplied” cancellation, a wrong item, a customer-satisfaction claim: each one produces a ceza faturası that is deducted from the seller’s next payout without asking. Alongside the money, each incident adds ihlal puanı (violation points) on a ladder that suspends the store for 2 days at 12 points, 5 days at 24, 7 days at 36 and closes it indefinitely at 48 (birfatura, 14 Oct 2025). The seller has a short objection window, the objection is a form in the seller panel that wants documents attached, and the alternative is a live-chat lottery.

Four things line up in the last fifteen months:

1. Trendyol reworked the penalty schedule on 2 June 2025. Penalties for unsupplied, defective, incomplete, wrong and late shipments moved to tiers computed from the net sale price of the affected items, with a flat 300 TL for severely defective goods and 70 TL for a label-price mismatch; the full table lives in Section 9.2 of the intermediary agreement (Ecomtrex, 30 Apr 2025). The 2021 schedule topped out at 300 TL per incident (TicaretEhli, 21 Aug 2021). A 2026 tracking vendor now quotes late-shipment penalties as “50–250 TL tiered” per package (HeyvAql, fetched 7 Sep 2026). Higher, price-linked penalties on a marketplace that sold over a billion items in 2025 is a bigger pot to reclaim.

2. The complaints are dated, specific, and about the carrier, not the seller. On 6 February 2026 a seller switched to Trendyol Express; 66 orders were collected, could not be scanned at the branch, came back as returns, and on 18 February Trendyol invoiced him 10,500 TL for “tedarik edememe”. His complaint has 559 supporters and no recorded reply (Şikayetvar, 26 Feb 2026). On 9 April 2026 another seller wrote that Trendyol’s own cargo did not turn up to a booked pickup, his store was restricted, penalties applied, and support told him it “could not be resolved” (Şikayetvar, 9 Apr 2026). A third was charged both a late penalty and an unsupplied penalty on the same two order numbers, discovered on 28 January and 4 February 2026 (Şikayetvar, 24 Feb 2026). The seller-panel complaint page shows 422 entries; the store-closure page shows 97, including a seller doing 3–4 million TL a month who got two consecutive 30-day closures with no explanation on 28 August 2026 (Şikayetvar).

3. The objection procedure exists, is documented by sellers, and is manual. A February 2026 seller guide gives the exact path (Finansal İşlemler → Cezai İşlem İtirazı), the evidence Trendyol accepts (the signed desi fişi / teslim tutanağı from the cargo branch, the carrier’s movement log for the tracking number, the integrator’s data-transmission logs for API-caused delays), a 7 business day window after the deduction, a 1–3 day turnaround, and refund onto the next payout; the author claims about 90% success with proper documents (Tıkla Kazan, 14 Feb 2026). Other secondary sources describe a 3-business-day window before the penalty is applied (birfatura). The 3-day figure is not Trendyol’s invention: Turkey’s e-commerce intermediary regulation of 29 December 2022 says a marketplace may only restrict, suspend or terminate a seller on objective criteria in the contract, and outside fraud or public-order cases must first request an explanation and give three business days for it (Lexology summary; Lexpera consolidated text).

4. The API exposes the timeline and the deduction, but not the objection. Trendyol’s shipment-package service returns agreedDeliveryDate, extendedAgreedDeliveryDate, cargoProviderName, cargoTrackingNumber and a packageHistory array of status changes with timestamps (developers.trendyol.com). The “other financials” service returns DeductionInvoices with a description and amount (developers.trendyol.com). A compensation-tickets endpoint reads the status of Trendyol Express loss claims but cannot create them (developers.trendyol.com). There is no penalty or objection endpoint in the published catalogue (llms.txt index). Read-yes, write-no: the product is the countdown, the evidence pack and the draft, not an integration.

The incumbent proves the wallet and stops at the edge. HeyvAql audits nine deduction lines against settlement invoices, flags late-shipment penalties and return-satisfaction fees, and charges a variable monthly fee by order volume. Its own page describes tracking and alerting; it does not draft, file or manage objections (heyvaql.com). Integrators charge 200–2,000 TL a month for order flow and stop at “Shipped” (Sentos from 800 TL/mo; EticaretRadarı comparison). KarPanel gives the profit calculator away free and sells inventory and order analysis (karpanel.app). Nobody sells the objection.

Provenance:

3. The opportunity

Every marketplace-tooling vendor in Turkey lives inside Trendyol’s API. That is why none of them sells this. The objection is a form with attachments in the seller panel. It cannot be submitted by API, so the integrators and the deduction trackers treat it as out of scope and stop at an alert. The seller is left with a red line on a statement, a countdown they do not know is running, and the job of finding a desi fişi from eleven days ago.

The gap has three parts, and each is a feature:

  • Attribution. Trendyol’s penalty says “late” or “unsupplied”. The seller’s defence is “the carrier did not come”, “the branch could not scan”, “the integrator’s status call failed”, or “you charged me twice for one order”. All four are visible in data the seller already has: the package history timestamps, the cargo provider on the package, the extension dates, the deduction description, and the integrator’s own logs. Today a human reconstructs this per order, per complaint, by phone, “every time to a different representative from the beginning” (Onur, 26 Feb 2026).
  • The clock. Objection windows are measured in business days from a deduction the seller may not open for a week. Statement lines are not push notifications. Whoever runs the countdown wins the refund.
  • The document. A Cezai İşlem İtirazı that gets approved is a short, dated, evidence-first Turkish statement with the right attachments. That is a drafting job, and it is the same job every time.

Incumbents cannot follow without changing what they are. HeyvAql’s product is a control tower with a buybox bot bolted on; adding a manual-filing workflow with document collection is a different company. Integrators sell to the warehouse. Trendyol itself has no incentive to make objections easier, and the regulation only obliges it to accept them.

4. Target market

  • Primary customer: The owner or ops lead of a Trendyol store doing 300–5,000 orders a month: small brands, wholesalers moving to retail, dropshippers running an integrator. Above this band a company has an ops person who already fights every penalty; below it the penalties are small enough to swallow. Turkey only for v1. Hepsiburada runs a parallel system with a 2-business-day handover rule, per-day late fines and an objection that “requires documentable evidence” (pazaryeridestek, 2026), so the second marketplace is the same product with a different login.
  • Why they buy: In their words. “Yaklaşık iki haftadır durumu anlatmaya çalışıyorum. Her aradığımda farklı bir müşteri temsilcisi telefona çıktığı için her defasında yaşadığım sorunu en baştan anlattım” (Onur, 26 Feb 2026). “Hem gecikme hem de tedarik edememe adı altında iki kez ceza uygulanması nedeniyle mağdur edilmiş bulunmaktayım” (Veli, 24 Feb 2026). “Trendyol’un kendi kargosu, zamanında randevusu olmasına rağmen ürünlerimi almaya gelmedi” (Fatih, 9 Apr 2026). “Kesinti tamamen müşteri beyanına göre, bizden hiçbir savunma istenmeden yapıldı” (karakayaspor, 29 Jul 2026). “Aylık ortalama 3-4 milyon TL ciro yapan satıcı olarak ardışık kapatma kararlarına anlam veremiyorum” (Efe, 28 Aug 2026). “Bir ürün yüklemeden 48 puan ihlal aldığım ve mağazamın kapatıldığını öğrendim” (Hanife, 13 Mar 2026).
  • Rough TAM reasoning: Trendyol reported over 250,000 business partners and more than one billion items sold in 2025 (Anadolu Agency, Jul 2026). If one seller in ten sits in the 300–5,000 order band, that is 25,000 stores that already pay an integrator and would recognise the pain in one sentence. Two thousand paying stores is the $1M ARR case below.
  • Why now for them: Penalties are now price-linked, the objection window is short, Trendyol Express is being pushed onto sellers (Onur’s whole complaint is his first week on it), and the closure ladder ends at 48 points. A seller who eats five penalties a week is not losing money, they are walking toward a suspension they cannot see coming.

5. Product sketch (MVP)

  • Connect once. Seller pastes their Trendyol API key. Tutanak pulls shipment packages with their full status history, agreed delivery dates and cargo providers, plus settlements and other financials, going back 90 days.
  • Penalty ledger. Every deduction line from the statement, matched to its order, with the amount, Trendyol’s stated reason, and the seller’s own timeline for that package side by side.
  • Cause verdict. For each penalty, a one-line call: seller-late, carrier no-pickup, branch scan failure, integration status failure, duplicate penalty on one order, or customer-claim with no defence requested. Each verdict names the timestamps it rests on.
  • Objection countdown. A running clock per penalty in business days, with Turkish public holidays, from the day the deduction appeared. Red at two days left. WhatsApp reminder.
  • Evidence pack. A checklist per objection: which document Trendyol expects for this cause (desi fişi, carrier movement log, integrator log), a photo upload for the paper ones, and the package-history extract already filled in.
  • Draft objection. A ready-to-paste Cezai İşlem İtirazı in Turkish, dated, order numbers listed, cause stated, evidence referenced, tone matched to what gets approved. The seller pastes it into the panel form and attaches the pack.
  • Refund watch. Tutanak watches the next payouts for the reversal and marks each objection won, lost or silent.
  • Points ladder. A live count of violation points against the 12/24/36/48 thresholds and the projected date of the next suspension at the current run rate.

6. AI angle — what’s load-bearing

Two jobs are AI, and without them the product is a spreadsheet.

First, cause attribution across messy evidence. The structured timeline gets you halfway: if packageHistory shows Picking and Invoiced before agreedDeliveryDate and Shipped after, the carrier held it. But the deciding evidence is unstructured: a photographed desi fişi with a stamp and a date, a carrier’s tracking page, an integrator’s log dump, Trendyol’s free-text deduction description. Reading those, aligning them to the order timeline, and deciding which of six causes applies is a vision-plus-language job that a rules engine cannot do and a human does slowly.

Second, the objection itself. The approval rate depends on a short, specific, evidence-first Turkish statement that cites the right timestamps and attaches the right document. A model drafts that from the verdict and the pack in seconds, in the register Trendyol’s reviewers respond to, and rewrites it when the seller adds a fact. Remove the model and each objection is back to a 20-minute writing job, which is exactly why sellers do not file them.

7. Localization angle (if any)

This is a Turkey-only product by construction, and that is the wedge, not a limitation.

  • Language: everything in Turkish, including the drafted objection, the evidence names (desi fişi, teslim tutanağı, hakediş) and the panel path. A global marketplace tool would have to learn all of it.
  • Calendar: business-day windows need Turkish public and religious holidays baked in. Miss one and the objection is late.
  • Pricing: TL pricing with monthly rebasing. Turkish sellers are used to integrator subscriptions that move with inflation; annual USD pricing would fail.
  • Channels: Turkish sellers organise on WhatsApp and Telegram help groups (wpgruplari.com.tr; t.me/trendyolsatkazan), on pazaryeriforum.com, and complain in public on Şikayetvar with store IDs attached. All three are reachable without ads.
  • Regulation: the ETAHS regulation’s three-business-day explanation right is the legal hook for the countdown feature and for the eventual “you restricted me without an explanation request” escalation.

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

  • Pricing: three TL tiers by monthly order volume, rebased quarterly. Starter 799 TL/mo up to 500 orders, Growth 1,999 TL/mo up to 2,000 orders, Scale 4,999 TL/mo above that. Optional success fee of 15% on recovered penalties for sellers who prefer to pay from the refund. At roughly 40 TL to the dollar those are about $20, $50 and $125 a month; the FX assumption is stated in section 13 and is the weakest number here.
  • ACV: blended around 2,000 TL/mo, roughly $600 a year.
  • Rough math to $1M ARR: about 1,700 paying stores at that blend. Against 25,000 stores in the target band, that is a 7% share.
  • Rough math to $5M ARR: Hepsiburada and N11 as second and third marketplaces on the same account (each adds a penalty regime, not a new customer), the Scale tier growing as sellers cross 5,000 orders, and a 20% success-fee mix on recovered money. Five million dollars needs roughly 6,000 stores across three marketplaces plus success fees, which is aggressive for one country and is why this is a VALIDATE, not a GO.
  • Expansion path: per-marketplace add-on; Trendyol Express compensation-ticket tracking as a second module (the read endpoint exists); return-rejection drafting (createClaimIssue is a write endpoint that takes attachments) as a third.

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

  • Şikayetvar is a lead list with store IDs. The Trendyol seller-panel and store-closure pages carry 500+ dated 2026 complaints, many with store names or numbers. Run every penalty-related complaint through the product’s free “am I inside the window?” check, reply with a short comment offering the evidence pack, and DM the ones who reply. Expect 5–8% to take a call; 30 sellers in the first month.
  • Free penalty-ledger audit in the WhatsApp and Telegram help groups. Post one worked example a week: a real (anonymised) double-penalty, the two timestamps that prove it, the objection text, the refund line. Offer a free 90-day ledger pull to anyone who sends an API key. Convert 15% of audits to Starter.
  • Integrator co-sell. Sentos, Dopigo, Entegra and the smaller integrators have the customers and no objection feature. Offer a referral fee per paying store and a co-branded “ceza itirazı” page. One integrator with 2,000 Trendyol stores is worth more than any ad budget.
  • pazaryeriforum.com and YouTube seller channels. Answer every penalty thread with the specific evidence Trendyol accepts for that cause, and link the free checklist. Sponsor two Trendyol-seller YouTube channels with a “we got 10,500 TL back” case study once one exists.
  • The Hepsiburada mirror. The same seller sells on both. When a Trendyol customer asks for Hepsiburada, ship it; that is the first upsell, not a new acquisition.

10. Build complexity — justification

Medium. The Trendyol read APIs are public, documented and stable; pulling packages, histories, settlements and other financials is a week. The attribution logic is a matter of aligning timestamps and reading documents with a vision-capable model. Turkish drafting is prompt work plus a small approved-objection example set that the first 30 customers provide. Business-day logic with Turkish holidays is a library. The hard part is not code: it is learning, from real outcomes, which evidence Trendyol’s reviewers actually accept per cause, and that needs paying customers, not months. Two people, 8–10 weeks to a v1 that files its first objection.

11. Gating checklist

GatePass?Note
Legal in target market✅Seller’s own data via seller’s own API key; filing an objection is a right the regulation guarantees
Ethical — no harm / dark patterns✅Product argues from evidence; fabricated documents are a stated kill (Trendyol closes the store for them)
Market exists (evidence above)✅500+ dated complaints, 559 supporters on one, HeyvAql already charging for the detection half
1–5 person team can build this✅Read APIs + document reading + Turkish drafting + countdown
Launchable with <$50K / ₹40L✅Two people, inference cost, no hardware

12. Feasibility score

AxisWeightScoreNotes
Problem intensity2016/20Money deducted without a defence, a suspension ladder, and a window the seller does not know is running. Sellers with 3–4M TL/mo turnover are getting closed. Not 17+ because a 25 TL late fine on its own is swallowed; the pain is the ladder, not the line.
Demand evidence1511/15Dated 2026 complaints with amounts and store IDs; a paid incumbent for the detection half; integrator spend of 200–2,000 TL/mo as a wallet baseline. No search-volume data and no seller has been asked to pay for the objection specifically.
Build feasibility1512/15Public read APIs, no write API needed, standard stack. Document reading and Turkish drafting are prompt work. Learning what reviewers accept takes live customers.
Distribution clarity1510/15Şikayetvar gives named leads; WhatsApp/Telegram groups and integrator co-sell are real channels. Conversion from a public complaint to a paying subscription is untested.
Revenue mechanics159/15TL pricing under inflation, a $600 blended ACV, and 1,700 stores for $1M. Works, but the FX assumption and churn after a seller’s penalty run ends are both guesses.
Time to first revenue107/10A seller inside a 7-day window with a 10,500 TL invoice will pay this week. Reaching that seller in week one of launch is the uncertainty.
Defensibility104/10HeyvAql can add a drafter. Trendyol could auto-approve carrier-caused objections. The moat is the approved-objection corpus and per-cause evidence knowledge, which compounds but slowly.
Total10069/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required (Turkish-speaking; has sold on Trendyol or run ops for someone who has)

Key assumptions to validate (3–5)

  1. Assumption: Objections filed with the right document for the cause get approved often enough to be worth a subscription. How to test: File 30 objections for 10 pilot sellers over 3 weeks; log cause, evidence attached, outcome. Need ≥50% approval on carrier-caused and duplicate-penalty cases.
  2. Assumption: The seller band at 300–5,000 orders sees enough penalty money per month to pay 799–1,999 TL. How to test: The free 90-day ledger pull on 50 stores; measure median monthly penalty total. Need median ≥3× the Starter price.
  3. Assumption: The objection window is really 3–7 business days and is enforced. How to test: Read Section 9.2 of the intermediary agreement from a pilot seller’s panel; confirm the window with three sellers who have filed.
  4. Assumption: TL pricing at roughly 40/USD holds enough for the $1M math. How to test: Price in TL, rebase quarterly, and measure churn on each rebase.
  5. Assumption: Trendyol does not block the product’s API key pattern. How to test: The API is seller-scoped; integrators use it the same way. Confirm the terms of the developer program allow a read-only third-party tool.

Risk flags

  1. Platform dependency: One marketplace, one API, one objection form. Trendyol can change the panel form, the window, or auto-approve carrier cases and remove the job. Hepsiburada as second marketplace is the hedge, not a solution.
  2. Currency: Every price in this plan is in TL. The dollar math is an assumption stated in the open.
  3. Incumbent adjacency: HeyvAql already has the API connection and the customer. A “draft objection” button from them is the month-6 threat.
  4. Reviewer discretion: Several sellers report that carrier delays after handover are “not accepted as a valid reason”. The product must be honest about which causes win; a tool that files hopeless objections becomes an “Operasyonel Süreç İyileştirme” warning for the seller.

14. Structured verdict

Score:                  69/100
Verdict:                VALIDATE
Confidence:             Medium
Best-fit builder:       Turkish-speaking technical founder who has run a Trendyol store or its ops; a partner who lives in seller WhatsApp groups
Time to revenue:        6–10 weeks
Capital to launch:      $8–15K (two people for ten weeks, inference, a Turkish company registration)
Top 3 assumptions to validate first:
  1. Approval rate — file 30 objections for 10 pilots, need ≥50% on carrier and duplicate causes
  2. Penalty volume per store — free 90-day ledger on 50 stores, need median ≥3× Starter price
  3. The window — read Section 9.2 from a pilot panel, confirm with three sellers who filed
Kill criteria:
  - Abandon if fewer than 15 of 50 free-ledger sellers show ≥2,400 TL/month in penalties
  - Abandon if approval on carrier-caused objections is under 30% across the first 30 filings
  - Abandon if Trendyol ships an in-panel automatic reversal for carrier no-pickup before v1

15. Next step — 1-week validation sprint

  • Day 1–2: Pull the 30 most recent penalty complaints from Şikayetvar’s seller-panel and store-closure pages. Message every one with a store ID: offer a free ledger pull and objection draft. Get 10 API keys. Read Section 9.2 from the first panel that opens.
  • Day 3–4: Run the ledger on all 10. Hand-build the timeline for every penalty, classify the cause, draft the objection, and file it with the seller on a call. Log window, evidence used, and whether the panel accepted the submission.
  • Day 5: Decide go / no-go on two numbers: median monthly penalty total across the 10 stores (need ≥2,400 TL), and count of filed objections that came back approved or “under review” by day 5 (need ≥5 of the first 15). Below either, this is a feature for HeyvAql, not a company.

Interested in a detailed proposal?

Get a deep-dive with market research, competitive analysis, and implementation roadmap.

Contact us

info@startupbasket.ai