E-tailer uses automated chat to save sales, maximize marketing spends

By Published on .

Reprints Reprints

When Anaheim, California-based set out to transform its business from a niche power products e-tailer into a leading mobile technology provider, it realized missed opportunities were not an option. The company had to close on leads generated by its search and other marketing expenditures and convert exit traffic from its Web site to actual sales.

“One of the things any marketer is concerned about is reducing cart abandonment,” said Jeff Gardner, the company’s CMO. “We wanted a solution to help with win-back … to keep people who ‘X’ out and get them to come back to continue their purchases.”

Serendipitously, Gardner stumbled on a solution while browsing the Web in search of laser printer products. When exiting one site, he was greeted by SMARTagent, an automated interactive chat product from technology company

SMARTagent was designed to launch chat windows when customers abandon shopping carts or order forms. Agents “manning” the chats are virtual.

Using artificial intelligence methodologies and natural language parsing, the SMARTagent technology invites departing visitors to engage in conversation and delivers scripted responses that correspond to their questions or comments. It can even “learn” the customer’s name if he or she divulges it, said CEO Glenn Russell. debuted SMARTagent on its Web site in November 2007. The implementation required minimal resources.

“We spent the most time crafting script responses and educating UpSellit on our business,” he said. “From an IT standpoint, it was a matter of taking snippets of java script code and putting them … on the pages where we wanted the agent to launch on exit.”

Then the real fun began. tested multiple variables, including chat messaging, agent personas and incentives to purchase to determine the most effective application of SMARTagent for its business.

According to Gardner, the tests revealed a few surprises. For example, in an A/B test of free shipping versus 10% off, the discount offer performed significantly better.

“That was something we hadn’t expected,” he said.

Also, female agent personas outperformed their male counterparts. “Maybe it’s something about trust,” Gardner surmised.

To date, has primarily used SMARTagent to answer online shoppers’ questions and to offer incentives that encourage at-risk customers to make immediate purchases.

But the technology can also be used to request and collect customer data—for example, for Web site registrations or e-mail opt-ins—by driving customers to an online form, including a form within the chat or collecting the e-mail address directly.

And, of course, chat transcripts can provide a wealth of information about the customer experience.

“We continually review and audit the chat logs to find out how the agent is engaging with customers, what kinds of questions are being asked and better ways to respond to commonly asked questions,” Russell said.

In the case of, chat logs revealed frequent requests for a popular product that the company had inadvertently omitted from its Web site. Correcting the mistake quickly not only improved the customer experience, but also reduced call center costs.

Gardner said has realized a 7% to 9% lift in revenue since deploying SMARTagent. And because the technology’s fee structure is performance-based, he can precisely quantify the return on investment.

This year, the company plans to roll out the technology across several new Web sites and to continue to optimize its use of SMARTagent on “We’ll … refine and improve on what we’ve already built,” Gardner concluded.

In this article:
Most Popular