March 02, 2005
Case Study

How to Convert More Online Shoppers with Live Chat Invites: Test Data From Earthlink

SUMMARY: Ouch! 70% of the consumers shopping at Earthlink's site for ISP services were leaving without purchasing, even when they saw broadband was available in their zip code for a "best offer" price. So, the sales team decided to test popping up a live chat request box in front of fleeing shoppers. Learn how they figured out how to test it without annoying everyone who got a pop. The tactic worked so well that Earthlink execs now consider live chat a key online selling tool. 15% of chats converted and 80% of consumers gave the chat a good or excellent rating. Here's the link for our Case Study featuring loads more data.
CHALLENGE

In a typical week, almost three million visitors go to Earthlink's site, many of them considering becoming a new customer.

Most of these ISP shoppers would immediately type in their zip code to see what types of services were available in their area. "We have a pretty good footprint nationally with 78% available for high speed Internet," says Donald Berryman, EVP Customer Support.

So, you'd expect most of these shoppers to be delighted with results and sign up for service. Not so. At least 70% of shoppers left without buying. "How can we shrink that 70% disappeared number?" Berryman's team wondered. They tried running site-only Best Offers and simplifying the check-out process, but nothing budged the needle too much.

Meanwhile, the customer support team reported adding live chat capabilities to their part of the site, reduced monthly phone call volume by almost 20%. Plus, customers clicking to chat loved it and tended to be an upscale demographic who could be upsold to more services.

Hearing this, the new account sales team yearned to pop up proactive chat invitations in front of shoppers.

On the other hand Earthlink's marketing department, like most, viewed chat pops with fear and loathing. Sending pop-ups to shoppers wasn't exactly in tune with Earthlink's brand. "Our culture was never to bother a customer," says Berryman.

CAMPAIGN

Jennifer Garrett, Director of Earthlink's Sales Call Center, got the OK to run a chat beta test in five very careful steps.

Step #1. Set measurement standards

Garrett worked with Earthlink's chat vendor to develop easily-readable measurement reports for her goals in three key areas:

o Conversion performance -- how many visitors qualified as "hot" enough shoppers that business rules would give them a pop-up; click rate for pop-invites; stick rate for chat sessions; ultimate buying conversion rate from chats.

o Shopper experience -- survey results showing how potential customers rated their chat experience.

o Operator performance -- based on typical call-center stats such as the average wait time, chat length, staffing requirements, etc.

Step #2. Select and train chat reps

Garrett had a ready pool of applicants for chat positions in her inbound call center. She chose the most eager ones with the best writing skills for the beta test.

Critical -- These reps already had two weeks of intensive training in all facets of Earthlink services, so they already knew the answers to most shopper questions. Now they had an additional three full days of training in chat-specific operations. Rules included:

o No chat shorthand -- all communications were to be professionally spelled out.

o Personal answers preferred -- although chat reps were provided with a full library of canned answers to common questions, they were encouraged to answer in their own words and create their own personal library of most-used answers.

This fits with Earthlink's brand campaign running on TV now featuring real-life employees competing for viewer votes to get parking spaces, vacation time, etc.

o Only run four chats at once maximum -- any more than this and response time slows too much.

o Focus 100% on chat -- while these reps could swap into call center roles if needed in a crisis, Garrett preferred to focus them on one mode of communication whenever possible.

o Pick up on chat requests in under a minute. Don't leave shoppers hanging too long.

Step #3. Launch click to chat offers on the shopping pages

First, Garrett added a "click here to chat" offer as a standard part of the site navigation on most left-hand site pages (link below to sample). It was a great way to get the reps used to answering shopper questions.

Step #4. Test business rules for chat pops

"We don't send chat until a customer requests it," notes Berryman. And, in fact, that pop-up invitation doesn't appear to all shoppers.

Instead the team created a set of business rules determining if and when a shopper might see a chat invitation pop. These included what the team suspected were the following warning signs that the customer was confused or uncertain and might welcome a helping hand:

o Repeat visits by an individual shopper to the same page more than once a day or multiple times in a week.

o A shopper "sitting on" a page for more than the page contents would seem to warrant (could range from five seconds to 30 seconds).

o A shopper abandoning an order form midway through entering their information in it. ("A lot of people start sign-ups but don't continue the process. Fallout is very heavy," notes Garrett.)

Important: Garrett only tested sending pops to a few business rules at first instead of sending pops to everything that moved. "We try not to be really aggressive about this. We didn't want them to think we were hovering over them or overwhelming them."

Plus, business rules were never set in stone. Garrett's team tweaked where and when they'd send a pop based on ongoing results from their metrics reports.

Step #5. Create inoffensive chat invitation pops

"We would nicely float the chat invite onto the side of the screen," says Garrett. "I don't want to call it a pop-up because it's not really intrusive."

The invitation was carefully copywritten to reduce the potential annoyance factor. Berryman describes typical copy, "We use very passive-type questions. 'Would you like help?' 'Is there anything I can do to help you?'"

Only after a shopper clicks to accept the invitation would the actual chat begin. So the shopper felt like they were in control and no sales rep was breathing down their mouse too heavily at them.

At the end of every chat, the shopper was presented with a quick five-to-seven question survey so they could score their chat experience. Questions were radio buttons to make it easier -- no typing.

RESULTS

15% of shoppers who interact with the chat pop-up wind up converting to buying services during their session. "We were really surprised," says Berryman. "It was outstanding. It told us we really needed that little extra push on the site because it increased our percent of conversion dramatically."

Here are a typical week's pop-chat metrics from Garrett:

-> Data on pop-up chat performance (does not refer to the "click to chat" icons)

o Average weekly site visitors -- 2,769,492 o Hot Lead Rate -- 4% (determined by our preset business rules such as being on a page for a certain number of seconds) o Invitation rate -- 57% (% of hot leads that we actually invited) o Acceptance rate -- 3.80% (% of invitation rate that accepted) -- this is actually above industry average o Interactive rate -- 81% (% of acceptance rate that was fully engaged and finished chat after starting the session) o Close rate -- 15% of interactive rate.

If you do the math, these numbers don't at first appear to be all that impressive -- roughly 292 new customers convert due to chat per week. But, when you consider the average new customer could be worth at base $240 per year to Earthlink, this incremental gain in conversions could make the company almost $6 million a year if they sent chat invites to all qualified visitors.

-> Post-chat survey results data

o Exit survey completion rate -- 12% (very high compared to industry benchmarking) o Excellent or good rating -- 80% (this is actually higher than our phone customer satisfaction in most cases) o Chat as preferred contact method -- 61%

-> Data on operator performance (call center-type metrics that measure efficiency)

o Average wait time -- 47 seconds (time it takes for agent to answer chat) o Average chat length -- 8 minutes 53 seconds o Abandonment rate -- 6% o Concurrent chat rate -- 2.43 (number of chats the average agent handles at the same time) o Agent utilization -- 80% (a measure of "staffed time to productive time") - anything above 80% is good.

Two key lessons from the above data for other marketers to consider:

Lesson #1. Chat as a preferred mode of communication

61% of chat users named it their preferred contact method. Garrett strongly suspects this will keep rising as a percent of all shoppers. "Chat is a mechanism they enjoy communicating through.

"I'm 33. I did not grow up chatting. The people who grew up chatting are all about to get jobs. They're the ones who will be buying via chat. It's probably an up-and-coming buying channel."

Lesson #2. Tweak business rules for incremental gains

"Over the last six months, we've increased sign-ups 25%-30% from chat just from managing business rules," says Garrett. "It's just like direct marketing; you'll do a bunch of different tests. If you stick it on there and let it sit, you're probably going to just get a bunch of tire-kickers. You have to work, work, work to understand the behavior of people visiting your site."

Useful links related to this article:

Samples of chat creative from Earthlink: http://www.marketingsherpa.com/earthlink/study.html

LivePerson the chat technology Earthlink uses: http://www.liveperson.com

Earthlink: http://www.earthlink.net


Improve Your Marketing

Join our thousands of weekly case study readers.

Enter your email below to receive MarketingSherpa news, updates, and promotions:

Note: Already a subscriber? Want to add a subscription?
Click Here to Manage Subscriptions