Case studies

30% Sales Increase AI/ML Solution by Codvo Transforms Product Categorization

About the client

Our client is a leading lease-to-own in-cart payment platform for retailers and consumers across the United States. The company has a consumer-centric business strategy and strives to extend credit to all consumer classes.

With an efficient lease application and approval process, the client provides transparent and tailored payment terms and award-winning customer service to its consumers. Instant credit extensibility while purchasing products online or in-stores makes the company a preferred platform.


Business Challenge

As discussed above, the company offers a no-credit-required lease-to-own in-cart payment option, allowing consumers to make monthly payments to pay for their purchase rather than paying for it in full at the time of checkout. This feature contributes to a more seamless checkout experience for end users and significantly increases sales volume for retailers and merchant websites.

The company also provides many retailers with information about hundreds of products, subject to change with time. The project goal was to create a dependable and scalable ML (machine learning) based solution that, when given product attributes, can successfully predict whether a product is leasable.

Our Approach and Solution

For the solution to correctly classify products as leasable or non-leasable, the product information has to be up-to-date, and the metadata needs to be exhaustive. Our team was confident that products and metadata collection required large-scale distributed scraping across the web and other sources. This data was cleaned and then used to train, tune, and validate a deep-learning recurrent neural network capable of classifying products into over 2000 categories. The results map against leasable or non-leasable criteria decided by subject matter experts and businesses to determine if the product is leasable.

The accuracy of our ML solution in production was around 95% in the first six weeks of implementation. The accuracy improves with time and should reach 97-98% within a year of implementation.

As the next step, this solution was containerized and made available to end users as a Chrome extension. The chrome extension instantly analyzes products in the cart on a merchant website, identifies if the product is leasable, and provides instant credit to consumers for buying a product using virtual credit card capability. For the same client, we have also done a project on virtual credit cards (VCC), and details are available on our website.

Tech Stack

The tech stack used for this project included: Python, TensorFlow, Docker, Jenkins, Selenium, Flask.


Business Impact

Without ML based solution for instant lease validation for products, the experience of consumers using the payment platform of our client improved.
The ML based in-cart payment solution was highly scalable, easy to use, and efficient, leading to increased adoption and traffic on merchant websites.
Due to increased traffic, the client was able to generate orders worth $3M+ in just two weeks of its launch. 
Increased traffic leads to more led to improved quality of lead generation at an exponential rate.
At the end of 6 months, the solution helped the client increase sales by 30%.