At a Glance:
The Challenge:
- Predicting whether a customer who made a purchase, will make another one in the next 24-48 hours
- Saving significant resources on multiple coinciding shipments to customers
The Solution:
- ShopTJC’s analysts built a predictive analytics model on the Pecan platform to predict the likelihood of making an additional purchase
- The model was built without prior ML experience
- A production-grade model was reached in a week
The Impact:
- 2000-3000 shipments saved per month
- 6% decrease in shipping costs
- High levels of customer satisfaction
- Expansion of Pecan to new use cases and geographies
About ShopTJC Ltd.
ShopTJC Ltd. (The Jewellery Channel) is a UK-based e-commerce retailer specializing in high-quality jewelry, gemstones, and lifestyle products, addressing a market of over 30 million homes in the United Kingdom and the Republic of Ireland. They offer approximately 15,000 items through their website, making thousands of shipments every month. In addition, ShopTJC Ltd. operates through two extremely popular TV shopping channels in the region.
The Challenge
ShopTJC Ltd. serves a large number of loyal customers who enjoy making multiple separate purchases in short time frames. Since ShopTJC Ltd. prioritizes the customer experience, they aim to commit to a short delivery time. This means they can’t hold all their customer orders for 48-72 hours, while waiting to see if customers make additional purchases.
However, if customers do make additional purchases, sending multiple packages to the same address is an inefficient use of resources. ShopTJC Ltd. needed a solution that could help them predict whether a customer who just made a purchase will make another purchase in the next 24-48 hours, allowing for the consolidation of shipments.
The Solution
ShopTJC Ltd. chose Pecan AI as their predictive analytics platform, because they have strong data analysts and an abundance of data but no data science resources available.
ShopTJC’s analysts were able to build the model in Pecan without prior ML experience. Their model reached production grade in only a week and was deployed.
Pecan automatically prepared and restructured ShopTJC’s transactional history, product data and customer behavior. Then, after a model training process, the predictive model for “likelihood to purchase” was ready. The model predicts whether a customer who made a purchase in the past 24 hours, will make another purchase in the next 24 hours.
These predictions allow ShopTJC Ltd. to bolster their business operations. For the top 50% of customers identified by Pecan’s model, ShopTJC is able to group orders and ship them along with any additional orders that come in within a set time period for better user experience .
The Impact
With Pecan’s highly accurate model, TJC’s leadership is able to benefit from:
- 2000-3000 shipments saved per month, a 6% decrease in shipping costs – ShopTJC was able to save redundant shipments without unnecessarily holding many orders. Thousands of unnecessary monthly shipments have been reduced in the UK, with an expected annual decrease of 6%. This significant reduction in operational costs has increased their profitability.
- High levels of customer satisfaction – Pecan’s accurate modeling ensured shipments were not delayed for customers. ShopTJC was able to maintain a positive customer experience, while living up to their promise for fast deliveries.
- Strategic business planning support – The ability to predict purchase likelihood opens up new possibilities for strategic business and sales activities, allowing ShopTJC Ltd. to find new ways to meet and exceed their business goals.
- Predictive analytics outputs without ML resources – ShopTJC’s analysts were able to build the model on their own on the Pecan platform without prior ML experience. This enabled ShopTJC Ltd. to enjoy high quality predictive analytics without having to hire costly and scarce data science talent.
Future Expansion
The success of Pecan’s predictive model for purchase likelihood has encouraged ShopTJC Ltd. to expand their use of Pecan. This takes place in three ways:
1. Expanding the use of Pecan to a Churn model use case: developing a predictive Churn model. This model predicts which members of ShopTJC Ltd., their premium membership, will churn in the next 30 days and preemptively treat them.
2. Expanding the use of Pecan to a fraud prediction use case: developing a predictive fraud model to identify fraudulent activities and potential chargebacks.3. Expanding Pecan to new geographies: Pecan has developed a similar purchase likelihood prediction model for the US-based sister company of ShopTJC Ltd. called ShopLC.
