How our self-learning model doubled the throughput of incoming products in just a few months

Photo by Joshua Sortino on Unsplash
Success Stories
How our self-learning model doubled the throughput of incoming products in just a few months


A large materials marketplace for the architecture and design industry


Efficient product data onboarding presented a considerable challenge for our client, as the process often stretched over weeks, and in some cases, months. Data flowed through various channels, including emails and file share drives, making it a convoluted and time-consuming ordeal. The complexity of the task grew exponentially as our client expanded their operations into Europe and Japan. With a significant number of SKUs to manage, the manual processing of data was not only slow but also financially burdensome. As SKUs continued to accumulate, the inefficiency of the manual approach became increasingly obvious.


Our team spent the first few weeks learning about our client's data stack, the data quality and their onboarding process. We then used a combination of automation and human validation to drill down to the products that needed to be onboarded.

The automation was done by our pre-trained model, and as is the case with any ML model, the first few passes required some amount of human validation to correct errors, if any. Every correction was fed back into the model, which then further trained itself. With more and more iterations, the model fine-tuned itself to the client’s data, and got better and better at producing accurate output.

With this kind of self-training, reliance on human validation decreased, and accuracy of the model increased, giving our client a highly efficient way to process product data.

Solution Plus

The team provided our client with a crucial recommendation to establish a brand portal, following which suppliers could efficiently channel their product details through a unified source, eliminating the need for file shares and emails. This strategic move resolved the issue of sifting through vast amounts of product data that did not fall within our client's business domain. The new portal enabled an automated and streamlined onboarding process, significantly reducing the backlog of products. Manual processing, which was slow and costly, was replaced with a more efficient system, ensuring that as SKUs accumulated, the process remained agile and cost-effective.


Our intelligent automation improved the throughput of incoming products from the brand portal. In just a few months, we were able to increase the number of SKUs processed by 100%, effectively doubling the throughput.

Share this post

More Stories