Customer Case Study

The Challenge

Before Proton, the lawn and garden parts distributor had a conventional e-commerce site. The online store allowed customers to search for products and place orders, but it did not proactively engage customers with personalized features. Even though thousands of customers were placing hundreds of thousands of orders each year, the distributor felt they were leaving money on the table by not pursuing up-sells and cross-sells.

Beyond that, the distributor also struggled with product data. They sell millions of different products online but have little or no information about many of those SKUs. This made it hard for customers to find the products they needed, let alone to explore new product offerings. In order to increase performance, they needed to boost product discoverability.

If the distributor could transform the site from a passive ordering portal into an active sales driver, the company would grow revenue and satisfy customers. To do that the distributor needed to create a "smart" website that made personalized pitches for each customer, much the way expert sales reps do.

Core Challenges Recap

• Poor Product Discoverability
• Lack of Personalization
• Inability to Up-sell/Cross-sell

The Solution

The distributor partnered with Proton to transform the website and grow revenue. This created a more relevant shopping experience that increased its customers' average order value and average order quantity.

Proton's cutting-edge artificial intelligence platform processes sales data to grow sales. Our predictive models forecast a range of things, including which items a customer is most likely to buy, when customers will be due to reorder products, and what products are frequently sold together.

Using a quick Magento integration, we integrated these predictive algorithms into the distributor's e-commerce site, and presented them to customers as helpful features. This made web visits more relevant and navigable for each customer.

Proton Features

• Recommended Items
• Complete the Cart
• Customers Also Bought
• Due to Reorder
• Switch and Save

In order to quantify the impact of our AI-recommendations, we ran a simple A/B test. When customers visited the website, they were blindly split towards the "A site" or the "B site." The A site included the AI recommendations listed above, while the B site offered no such enhancements. We then tracked the performance of each site over a S0 day period.

Results and Next Steps

Customers that were given AI-based product recommendations ended up with an average order size of $186. Meanwhile, the average order size of the B group was 12% lower at just $166. This indicates that our AI recommendations significantly increase average order value.

In order to truly dub the website a success, however, we didn't just want to see higher average order value. We also wanted to see larger order frequency. If both of these metrics were improved, we could conclude that the A site grew customer share of wallet. However, if only average order value (AOV) increased, we have to infer that the site helped customers bundle orders, but did not hit on true growth opportunities.

We went back and checked the average number of orders per customer over the S0-day trial. We found that the average A site customer placed 8 orders, while the average B site customer placed just 7.S orders. This shows that while AI-supported customers were placing bigger orders, they also found the shopping experience appealing enough to place more orders.

In total, this combination of larger and more frequent orders added up to a 21% difference in revenue per customer. With AI-recommendations, customers spent $1,476. Without them, they spent $1,218. Given that the distributor brings in roughly $S0 million annually from the current e-commerce platform, these AI enhancements are currently pacing to generate an additional $10 million over the next year alone.

In time we may continue experimenting with different A/B models to further enhance website performance. Beyond that, we are engaging with this distributor about applying Proton's predictive models to additional channels. If our AI can grow e-commerce revenue by more than 20%, it can also make a significant impact on other passive channels, like inbound telesales and counter orders. Phase 1 of real transformation is personalized recommendations. Phase 2 is strategic cross-channel coordination.

Download the PDF