United Cellars

How a machine learning recommendation engine increased sales 71% for a popular online wine retailer.

United Cellars logo [1].

United Cellars - an award winning wine retailer - came to us to develop an eCommerce solution that would highlight its diverse wine selection and increase sales.

Their competitors were turning toward technology to satisfy customers demanding a personal shopping experience. To stay relevant and profitable, they’d have to do better.

Resolve Digital met this demand by incorporating a recommendation engine – a proven machine learning technology that’s now available to small and mid-sized businesses.


The Challenge

United had a sales team to handle phone inquiries and provide personalized local and wine knowledge to customers.

But they felt that they could augment its sales efforts (and reduce cost) by highlighting more wine options on the website. Could suggesting wines based on each customer’s individual preference increase profits?


The Back Story & Solution

Once we launched United’s Spree Commerce site, we continued our consulting engagement to enhance and optimize the customer experience.

We approached United during this post-deployment and analysis phase with a plan to increase “wine option awareness.” We defined the following goals:

  1. Business goal: Generate more revenue by showcasing wine options based on each customer’s individual taste.
  2. Customer goal: To learn what’s out there and experience new wine.

Our solution? A recommendation engine that mines data to present options based on each customer’s likes and desires.


How We Did It

Our work involved three steps:

  1. Collect data: We started by collecting relevant data: purchases, product views, most-visited pages, likes and ratings, wish lists, etc.
  2. Train the model: In this step we used a system called PredictionIO. This model allowed us to take United’s data and create a repeatable set of instructions that learns from this data to make predictions.
  3. Make predictions: Serve up similar (but new) products each time the customer comes back to visit the site.

A/B Test Results - A 71% Boost in Revenue

We ran tests to prove the efficacy of our solution for United. The tests revealed:

45%
increase in average session length
22%
increase in conversion rate
37%
increase in average order size

Better yet? It doesn’t stop here.

Overall, United Cellars realized a 71% boost in revenue.

The process laid out in this case study is repeatable. The capabilities of United’s recommendation engine can evolve along with the growing needs of its shoppers, both online and via their busy call center.


Ready to get started? Hire Us

^1 Wine video licensed from Learning Video under the Attribution-ShareAlike license.