Marketplace Commissions vs Your Own Store: The Real 2026 Math for Turkish Sellers

Almost every seller we talk to knows their commission rate by heart and almost none of them know their take rate. Those are different numbers, and the gap between them is where the margin goes. The commission is one line on the settlement report. Underneath it sit four more deductions that are contractual, automatic and rarely modelled before a pricing decision is made. This guide reconstructs the full stack with published 2026 figures, then gives you the calculation that actually decides whether to build your own store.

Layer one, commission. Trendyol publishes category rates and the spread is enormous: around 5% on digital gift cards, 7% on smartphones, 7.50% on laptops, 8% on televisions and consoles, 15.25% on packaged food and pet food, 16.50% on children's toys, 20.34% on home textiles and furniture, 21.36% on clothing, bags and lighting, 22.37% on jewellery and belts, and 23.39% on footwear. Two mechanics matter more than the headline number. First, commission is calculated on the VAT-excluded sale price and then VAT is added on top of the commission invoice, so the cash effect is larger than the percentage suggests. Second, the rate is not fixed to you: seller level (4 and 5 earn published reductions of roughly 2.5-7% and 2.5-10% in selected categories), the women entrepreneur programme (published bands of 1.80-5.50%), and campaign periods all move it. Read your own rate in the seller panel under agreement details, not from any blog including this one.

Layer two, the per-order fixed fee. This one is invisible in percentage thinking and brutal on low-priced items. Trendyol charges a platform service fee per shipment: 13.19 TL including VAT on standard dispatch, 5.99 TL including VAT when the order goes out same-day. Hepsiburada charges a transaction fee in the region of 8 TL plus VAT per order in 2026. On a 1,200 TL jacket that fee is noise. On a 90 TL phone case it is 15% of the sale price on its own, before commission touches it. This single mechanic is why low-ticket marketplace catalogues quietly lose money at volume while the dashboard shows growth.

Layer three, shipping and returns. Commission is charged on the product price, not the shipping line, which sounds like good news until you look at who absorbs the desi-based freight cost and the return leg. Returns are the expensive half: return freight, repackaging, and the inspection labour on goods that come back unsellable. From 1 January 2026 this got structurally heavier. The amended Distance Contracts Regulation, published in the Official Gazette on 24 May 2025, removed the withdrawal-right exception that used to cover mobile phones, smartwatches, tablets and computers, and tightened the rules so that return costs are not pushed onto the consumer when the seller has named the carrier in the pre-contract information. If you sell electronics, your 2026 return reserve is not last year's number. Model it explicitly and talk to your own lawyer about your specific obligations, because this paragraph is a summary of published commentary, not legal advice.

Layer four, the discretionary spend that stops being discretionary. Marketplace ads and campaign participation start as a growth lever and end as rent. Once competitors in your category bid, visibility without ad spend collapses, and campaign periods carry mandatory discount floors as the price of admission to the traffic. There is no published percentage for this because it is auction-driven and category-specific, which is exactly why it belongs in your model as a real, measured line rather than a rounding error. Pull three months of settlement reports and calculate ad spend plus campaign discount as a percentage of gross sales. That number is usually the biggest surprise in this entire exercise.

Add the layers for a realistic apparel case and the take rate lands in the 30-40% band: roughly 21% commission, a fixed fee, freight, several points of ad spend, campaign discounts, and a return rate that in fashion is structurally high. We are deliberately giving you a band rather than a fake precise number, because the honest version of this calculation uses your settlement reports, not our assumptions. Export three months, sum every deduction line, divide by gross sales. That single percentage is the number your pricing should be built on.

Now the other side of the ledger, honestly. That take rate buys real things: demand you did not have to generate, consumer trust in the checkout and the return process, logistics infrastructure, payment handling, and a search surface where people arrive already intending to buy. A new brand with no audience and no search volume genuinely cannot replicate that, and telling you otherwise would be selling you a website you are not ready to use. The marketplace is not overpriced in the abstract. It is overpriced for the specific orders where the customer already knew your brand and would have bought from you anyway.

Your own store's cost structure is short by comparison. Payment processing runs around 1-2% for single-payment domestic card transactions on the published bands (iyzico around 1.19-1.95%, PayTR around 0.99-1.49%, with the wider Turkish market spanning roughly 1.29-3.49% depending on volume, sector and instalment mix). Hosting and infrastructure for a mid-volume store is a few hundred TL a month on modern platforms. The build is a one-time capital cost. And then there is the line that is not optional: acquiring the traffic the marketplace used to hand you. That is the whole cost of owning your channel, and any plan that ignores it fails in month two.

So here is the break-even that matters. Do not compare commission against build cost, that framing produces the wrong answer in both directions. Compare on repeat customers. Every order from a returning customer on a marketplace pays the full take rate for demand generation you already performed yourself. Take your monthly repeat-order count, multiply by average order value, multiply by your measured take rate. That is your monthly loss to the marketplace on demand you already own. Divide the build cost by that number and you have the payback in months. When we run this with sellers, the pattern is consistent: brands with a 25%+ repeat rate pay back a custom store inside two quarters, and brands where nearly every order is a first-time stranger do not pay it back at all yet.

The answer for most sellers is not either-or, it is a deliberate split. Keep the marketplace as a discovery and acquisition channel, and build your own store as the place returning customers, campaign traffic, WhatsApp orders, QR codes on packaging and your Instagram bio all land. Route first purchases through whichever channel wins them, then move the relationship. Practically that means an insert in every marketplace parcel that gives a reason to buy direct next time, your own store carrying the full catalogue including items the marketplace makes unprofitable, and a real CRM so you can measure repeat rate rather than guess it. That is the version where you stop paying acquisition rates on customers you already earned.

What to build, concretely, if the math says go. A store that loads in under a second on mobile, because Turkish e-commerce traffic is mobile-dominant and every second of delay is measurable conversion loss. Domestic payment infrastructure with instalments, because instalment support is not a nice-to-have in the Turkish market. Cargo integration with the carriers you already use, so fulfilment does not become a second manual process. Distance-selling compliance built into the checkout, including the 2026 withdrawal-right changes. Product data structured so that search engines and AI assistants can read it without executing JavaScript. And a stack you own outright, so the channel you just bought cannot be repriced on you next January.