instandart
#amazon #business_site #web_development #api_integration

Project description

Through a complex analysis of various data, our system helps Amazon sellers not only with controlling and forecasting their stocks more effectively, but also help them simplify one of the most complicated and intricate parts of their work, letting them get remuneration which they have lost before.

Project summary and fact sheet

Location:
Worldwid
Team size:
project manager, 3-5 developers,  QA, 2 support specialist
Industry:
e-commerce, seller tools, inventory management
Expertise:
e-commerce development, SaaS, Big data, Amazon API integration
Business goals:
reduce client’s losses, prevent fee overcharges
Technologies:
Python/Django, MySQL, Redis, RabbitMQ, Docker, JQuery, Vue.js, NPM, Pandas

Client and goals

The majority of Amazon sellers participate in the FBA (Fulfillment By Amazon) program, which makes selling and shipping easier as Amazon itself provides and controls the supply chain using its fulfillment centers. But even with a well-defined and established Amazon business process, the goods can often get damaged or lost during the delivery. In these cases sellers can submit a request to receive the reimbursement from Amazon. There are more than 20 scenarios in which Amazon is required to pay compensation to the sellers. In accordance with these defined cases, reimbursement requests are created and fulfilled automatically, but many of them often slip through the cracks and Amazon simply “forgets” about them. If the seller has a high flow of goods, the amount which Amazon owes to the seller may become significant.  

In order to guarantee their money back, sellers are forced to send requests for reimbursement manually which leads to large volumes of routine transactions, such as requests for Amazon reports, export, and conversion of data, manual file editing etc. It also becomes necessary to use third-party programs such as Excel to analyze and sort data. Overall, these steps prolong the reimbursement process and make it ineffective.

Problem/Challenge

There are many services that make the reimbursement process automated to some extent. However, they work with specific types of cases and still involve a substantial amount of manual work from the users which becomes overwhelming and ineffective when dealing with large amounts of product flow.

Our client, the US company, offers an integrated online service of process automation for Amazon sellers. The company requested a high-productive solution that would make possible the following:

  • To request Amazon reports via API automatically;
  • To analyze a received data and to create completely automated requests for a refund based on analysis of that received data for all possible cases: lost/damaged/not returned to seller etc;
  • To cross-check Amazon fees to prevent fee overcharge;
  • To analyze the inventory balance, as well as control and predict sales for certain time periods;
  • To generate shipment labels with 2D Barcodes and shipment pack file;
  • To provide simple and convenient case management;
  • The solution must be compatible and integrated with the rest of the subsystems of Amazon process automation (repricing, restocking, listing management etc), included in the client’s service.

Solution

In the course of the development, we relied on our significant expertise in creating process automation systems for Amazon. For back-end implementation, we used Python as our main language, along with Django framework for its proven capabilities such as rapid development, pragmatism, maintainability, clean design, and website security, in regards to similar tasks. 

Anticipating working with a large amount of data that is generated by Amazon reports, we used Pandas. To ensure that the system has a high load capacity, we deployed a multiple droplet system (more than 20 droplets), deployed over the docker swarm via the load balancer, database replications, and shardings for high performance, RabbitMQ with Pika for multiprocessing (can handle more than 300 consumers simultaneously). For the front-end, we used JQuery, Vue.js, which allowed us to develop a convenient and laconic user interface. 

Development

Core project team consisted of 4-6 people (depending on the project stage) such as 1 project manager and 2-3 back-end senior developers, 1 front-end developer, and 1 Q&A. We also had a support team working in close collaboration with our customers. Over 100 experienced Amazon sellers contributed their data for debugging, and their suggestions and experiences have been incorporated into additional functions and system improvements.  

The team maintained coherence, demonstrating professionalism and high levels of interest in solving a client’s problems. For developers who were experienced in using API and Amazon marketplace processes, it was an excellent challenge to keep them highly motivated and involved. 

Result/Conclusion

System development started more than 2 years ago and has been ongoing as of today. The first working version was presented to the client 2 months after the beginning of the project. 

System development implemented more than 10 unique sellers reports that were not possible for the standard Amazon console capabilities, with some ideas and developments creating the base for another Amazon system automation product (‘Case Management’ ™).

The outcome of the project resulted in a powerful, convenient cloud solution that added another “blade” to the integrated service that can be kind of a “Swiss army knife” for Amazon sellers.

The system has proven its effectiveness in a short time after the roll-out: one of the customers who is specialized in sea shipping was able to claim and receive reimbursement from Amazon for over $10k for lost products in the first weeks of using our client’s system.

contacts-bg

Got a project?

Want to discuss your project with us and make your ideas come true?
We're always just one click away.

We use cookies to give you the best possible experience on our website. If you continue without changing your settings, we presume that you accept receiving all of the cookies on our site read more.
Accept