Back to case studies

Our goal

Our goal was to build a system able to buy products from a bulk seller and sell them on various marketplace simultaneously.

Technical Challenges and solutions

1.Database normalization

In dropshipping business, databases powers the system. Our project has one database that combined all the information about orders, clients, and products. When bulk vendors change the price of the product, the system updated the whole thing. We’ve got a high number of deadlocks as other functions were making requests to the database.

To overcome the issue we have done database normalisation. Our team created a new database with a clear structure that contained only prices for every item. Therefore, updates were performed separately, not overloading the main database.

Our task was to change the java code for functions dealing with prices in the system. We changed business logic and updated about 1 000 000 items on every marketplace. This challenge allowed us to play with logic and to accelerate connections between stores and system.

2. Change vendor logic

According to the business logic, when a client orders a product via the store, the system sends a request to the vendor assigned to this product. However, we figured out such an algorithm doesn’t bring maximum profit. Since we had vendors with similar product but lower price.

With new business logic, after the client makes an order, a system in a runtime monitors identical products from all the vendors and assign a request to the one having the lowest price.

3. Delete duplicated items from marketplaces

Since we had similar products assigned to different vendors, we were selling duplicate items in one store. As we changed business logic, we had to eliminate duplications.

The challenge was to dig in the database and save dependencies product names. For example: 'SomeBrandName' has relation to 'SomeNameBrand Inc'. This task was done using SQL. We added two new tables which stored relation between brand names.

• - first table stores «VendorBrandName», «brandId»
• - second table stores «BrandName», «brandId»
• - new column «brandId» we added to our inventory table

Next step was actually to merge existing items by having brand names relations.

We came up with the new business logic to decide which items to keep and which to delete. After determining we removed duplicate items from marketplaces and linked them in our system with remaining ones.