At Caravel, we’ve spoken before about the challenges facing digital retail: online conversion hovers around 3%, and even when customers do buy, 30% of purchases on average will be returned—an extra cost to brands and retailers.
Early on, our team was compelled to uncover the human behaviors and roadblocks feeding this retail challenge. We’ve distilled it down to a few key concerns.
First, the online shopping experience often falls short to build buyer confidence, which causes drop-off prior to purchase. In particular, buyer confidence lags in three key areas:
"Which of these products best fulfills my needs?"
"What products fit my personal tastes?"
"What size is right for me?"
When these questions aren’t addressed, customers either stall out before purchasing or purchase multiple items that they’ll need to return.
Customers often seek out resources and support to build that confidence, but it’s not always easy to find online. We found that brands have a high volume of pre-purchase questions coming through their traditional post-purchase support channels. In our own research, we found that 72% of customers want to engage with a brand pre-purchase through chat when shopping online.
Through this research, we concluded that if customers could talk to a brand representative pre-purchase, they would be more likely to convert and have a more positive experience. In this article, I'll share some of the approaches we've taken in designing Caravel to support this goal.
Ask the right question, at the right moment
A good sales associate doesn't bombard you the moment you walk into a store; rather, they approach the customer at the right place and time. This requires the agent to be aware of the customer’s browsing behavior and where they’re at in the customer journey. Proactive selling means that Caravel reaches out to the customer on key pages or in specific situations to assist them at the moment they need it, instead of interrupting their browsing or engaging too early.
Once the right time for assistance is identified, it’s time to enact the conversation. The first question we ask will set the tone of the conversation, define its parameters, and set the interaction up for success.
Caravel is designed to observe browsing behavior and trigger proactive responses based on where the customer is in their shopping journey. We then use contextually-relevant dialogue to be assistive to the customer but not assertive.
Caravel starts a conversation with a specific question, such as “Who are you shopping for?” This identifies the buyer and guides them towards the highest-value assistance we can offer.
Caravel uses traditional natural language processing methods and deep learning models for its Natural Language Understanding to translate it into structured data, which can then be fed into search queries for product discovery, size translations, etc.
The most valuable customer insights can come from natural back-and-forth conversation. Customers may change their minds, ask for more information, or make contradictory statements as they determine their purchase decisions.
Caravel intelligently constructs follow up statements that ask for specific data points it needs until it can confidently suggest a product or action. Lastly, we've designed our agent to make it known that it'll connect the customer with a human agent should they ever need more assistance. This tactic helps in expectation setting.
As Caravel launches on the platforms of its first customer cohort, the preliminary results have already begun to validate our approach. For one example customer alone, within the first 30-days of Caravel running on their eCommerce platform, we saw a:
- 64% increase in conversions
- 3x increase in add-to-cart rate in chat over Shopify average and
- 58% reduction in returns
With these results as motivation, we’re excited to continue building out Caravel as a leader in the conversational AI space.