Marketplace Management

Marketplace Integration: Integrator or Custom API?

Share
Marketplace Integration: Integrator or Custom API?

Integrator or custom API for Trendyol, Hepsiburada, N11 and Amazon? Five decision questions, API limits, a hypothetical cost example and a contract checklist.

For a business selling on Trendyol, Hepsiburada, N11 and Amazon at the same time, there are three ways to keep stock, price, product and order data in sync across channels: a ready-made integrator panel, an API integration you build yourself, or a mix of the two. This article sets out the questions that should decide it, what you need to know about marketplace APIs, and a checklist to run through before you sign anything.

Three Routes: Integrator, Custom Integration and the Hybrid Model

A ready-made integrator is a software service, usually on a monthly or annual subscription, that manages several marketplaces from one panel. Product transfer, stock and price updates and order import come ready; you only enter your marketplace account credentials.

A custom integration is software built for you that connects directly to the marketplaces' APIs and talks to your ERP or store system. The business rules are written around your processes; maintenance and monitoring are your responsibility, or your software partner's.

The hybrid model leaves everyday work (listing products, the order screen) to an integrator and solves only your own flows (stock from the ERP, special pricing rules, invoice sequence) with custom software. We covered the general framework in off-the-shelf or custom software; here we apply it to marketplace integration.

What to Know About Marketplace APIs

Whichever route you choose, the same API rules apply underneath. Taking Trendyol's developer documentation as an example:

  • Credentials belong to the seller. Requests are made with a supplierId, an API key and an API secret; only the account's master user can see them. Each supplierId has its own key, and production and test environments use different credentials. Keys should never be shared on open platforms such as GitHub.
  • Requests identify themselves. A User-Agent header is mandatory on every request; requests without one are blocked with a 403 error. The format is described as "{seller ID} - {integrator company name}" if you use an integrator, or "{seller ID} - Self Integration" if you use your own software.
  • Request volume is limited. You can send at most 50 requests to the same endpoint within 10 seconds; the 51st request returns a 429 error.
  • Batch operations complete asynchronously. Operations such as product creation and stock-price updates return a batch request ID, and you check the result separately with that ID. Batch status appears as IN_PROGRESS or COMPLETED and item status as SUCCESS or FAILED. In other words, "request sent" is not the same as "stock updated".
  • Versions get switched off. Trendyol's developer page says Product V2 and Order V2 services are live and that V1 services will be deactivated as of 15 October 2026. Whichever route you choose, confirm that your integration uses V2.

Amazon's Selling Partner API follows similar logic but different rules: requests are limited with a token bucket algorithm, limits vary by operation and by seller-application pair, and exceeding them returns 429. Amazon notes that even a well-behaved application may see an occasional 429; it recommends a back-off strategy for retries, notifications instead of frequent polling, and batch operations such as the Feeds and Reports APIs. These details show that the software has to be more than a plain "send the stock" loop. For the current limits and required fields on Hepsiburada and N11, check each platform's seller developer documentation.

Strengths and Weaknesses

Ready-Made Integrator

Strengths: a fast start, no maintenance burden on the seller, the provider handling updates when an API version changes, and one screen for several channels. If you have few channels, a standard product structure and either no ERP or a simple one, it is often the most sensible starting point.

Weaknesses: the rules are limited to the options the provider offers. If you need a different price rule per channel, buffer stock, stock separated by ERP warehouse or a special invoice sequence, you may not find them in the panel. The management of your account (connected keys, data history) also depends on a third party; if you want to change provider, you may have to rebuild your product mappings.

Custom Integration

Strengths: all the business rules are in your hands; stock can flow event-based straight from your ERP, price can be calculated per channel, and invoice and shipping steps fit the order chain exactly. We explained the reasons in our articles on multichannel stock and price synchronisation and ERP integration.

Weaknesses: the upfront cost is higher and the work is never finished. Marketplaces switch off versions, add fields and change limits; like the V1 shutdown above, someone has to watch each change and adapt. Error handling (429s, failed batch items, duplicate orders), monitoring and reconciliation are part of the software too. For more on choosing a method, see our API integration article.

Five Questions to Decide

  • How many channels and products do you manage? Few channels and few products point to an integrator; many channels and a large catalogue with variants point to a custom flow.
  • Where does stock originate? If stock is born in an ERP or warehouse system and delays cost you sales, you need an event-based flow.
  • Do you have channel-specific rules? Do per-channel prices, campaign prices and buffer stock fit the options in the panel?
  • Who catches errors? When a failed batch item or a duplicate order is noticed, who is responsible and how fast do they act?
  • What is your exit strategy? If you leave the provider, how do product mappings, order history and settings come back to you?

Cost Comparison: A Hypothetical Three Years

Real prices vary widely by provider, product count and scope; every figure below is an assumption, used only to show the logic of the calculation.

Hypothetical example

Assume an integrator subscription of 6,000 TL a month: 36 months × 6,000 = 216,000 TL. For a custom integration, assume 250,000 TL for the initial build and 60,000 TL a year for maintenance and monitoring: 250,000 + (3 × 60,000) = 430,000 TL. On these assumptions the custom solution costs more after three years, but what falls outside the comparison can change the picture. If a rule the integrator cannot handle causes 40 oversold orders a month, and each costs on average 300 TL in cancellations, penalties or lost sales, the monthly loss is 12,000 TL and over three years 432,000 TL. So the decision is not subscription versus build fee, but the total cost each route creates.

The rate limit is a cost too. If an endpoint takes one product per request and the limit of 50 requests per 10 seconds applies, you can send at most 5 requests a second. Updating all 12,000 products through that endpoint takes 12,000 ÷ 5 = 2,400 seconds, about 40 minutes (hypothetical; check the documentation to see whether the endpoint accepts multiple items). You can spread price or stock changes over that time, but at the peak of a campaign that delay means stock that cannot keep up with orders.

A Staged Roadmap

You do not have to decide in one go. The sequence that tends to work in practice is:

  1. Month 1: Open at least one marketplace through a standard integrator and keep product mappings tidy.
  2. Months 2-3: Collect problems such as delayed stock updates, duplicate orders and failed batch items in one list, and write down the estimated cost of each.
  3. Month 4: Review the list. If most problems are solved with panel settings, stay with the integrator; if a few recurring, expensive problems go beyond the panel, move only that flow to custom software.
  4. After that: Set up monitoring and weekly reconciliation for the flow you moved, and assign someone to follow version shutdowns.

This approach stops you investing up front in software for a need you do not yet know, and lets your own data show the real requirement.

Contract and Setup Checklist

  • Make sure you own the API keys; give the provider only the permissions it needs and agree that you can rotate the keys when you leave.
  • Get in writing that the integration uses V2 services on Trendyol and how quickly the provider updates when versions change.
  • Ask how many minutes, in the worst case, a stock or price update takes to reach the marketplace, and how failed batch operations are reported.
  • Find out how retries and back-off behave on a 429 error.
  • Verify that you can export the product mapping table (SKU to marketplace product code).
  • Define a weekly reconciliation step that compares the order count in the marketplace panel with the count in the integration system.
  • Name the person responsible for errors and delays, and the response time.
  • For custom software, ask for maintenance, monitoring and version tracking as a separate line item in the quote.

Conclusion

The right route is often not either-or but staged. Open the channels with a standard integrator at the start, record the rules and the errors, and when the flows it cannot handle become countable, move only those to custom software. Whichever route you choose, keeping the API keys, the product mappings and the reconciliation report in your own hands is what lets you reverse the decision later.

Sources: Trendyol Developers: Getting Started, Authorization (credentials, User-Agent, request limit), Check Batch Request Result; Amazon Selling Partner API: Usage plans and rate limits (2026). The worked example is hypothetical.

More

Related posts