Rethinking multi-marketplace publishing

One of the most important tools in the product is the integration with multiple marketplaces such as Amazon, Mercado Livre, Shein, and TikTok Shop. Customers often migrate to Magazord because of it, but there was a critical challenge affecting both Magazord's operations and the ecommerces' own.

Sending a product to a marketplace demanded too much effort. It had to be done product-by-product for each marketplace, and every marketplace had a different flow. Because of this, Magazord had to run many training sessions and received a high volume of support tickets around it.

The stakeholders had concluded that the marketplace product, one of the platform's biggest revenue generators, was simply "too hard to use," and that the fix was to add a new step inside product registration.

The product lead endorsed it, largely because competitors had done something similar. By the time it reached the product manager, the direction had hardened into "build this," and the ask that landed on me was essentially "create a prototype for it."

Following my intuition, I knew this approach wouldn't solve the problem, so I decided to investigate further.

I also read all the documentation and training material available to understand the process, and realized that most of the information needed to publish a product already existed in the system.

With this in hand, I started a conversation with the PM team, and the core problem became clear:

All marketplaces work around product category, and the Magazord system only discovers which information it needs to ask for after the user inserts the marketplace category into the system.

We developed a new source inside the system where the customer can tell the system:

Every time I create a product with this category, map it to these marketplace categories.

By setting up this source, every product is one step closer to being sent, and the system already knows which inputs are necessary to publish it.

By this point I had gathered enough information to design a better flow and solve this last part. Talking with development first, I validated the idea of reusing every piece of information already entered during product registration inside this new flow.

With that, I recreated the flow with a single product to check whether it was easy, while already thinking about scale. After prototyping something I thought was easy to use, I kept adding more products to stress its boundaries, and after around 50 products with different categories it started to become harder to use.

One important thing I took into consideration was the shape of the marketplaces: nearly 50% became a select-and-publish flow, 40% asked for some additional information, and 10% were more complex, which were Amazon, Mercado Livre, and Shopee.

Some edge cases started to appear, but none of them broke the flow I had set up.

With a solid high-fidelity prototype built using Claude Code, I was able to add business rules, marketplace constraints, edge cases, and interactions to create a realistic experience. Then I started an initiative to travel across the state and collect feedback from different customers for a final validation.

Most customers were struck by how much of a game-changer it would be for their operation, because this new way was faster and easier to operate across many marketplaces at once. Some even talked about reallocating their marketplace team, since with this new process only one or two people would be needed to run the operational tasks.

After coming back to Magazord with such great news, I started conversations with the technical team to understand the API constraints and how to handle certain scenarios. For example, with Mercado Livre there is a problem when sending more than one product per flow, so I suggested generating an async queue to space out the submissions and prevent the issue.

These conversations were also used to split, plan, and prioritize how this large project should be delivered. We created an effort-versus-impact matrix, which the product manager then owned.

The internal team also felt the impact, with less training time and higher satisfaction. Training dropped from 2h30 to 30 minutes, because the flow is now the same for every marketplace.

With a simpler process to handle, more customers started to activate more marketplaces within their ecommerces, and the support team saw marketplace-related tickets drop from 7% to 4%.

Changing the perspective on this project was really hard, because many stakeholders were demanding a solution without understanding the problem. At first, that put me in an uncomfortable position, but the results, and becoming an example to be followed inside the team, made me proud of how I approached it.