CHALLENGE
J.G. Wentworth's target marketplace is middle America... with a twist. The financial services firm specializes in clients who've won a structured settlement (perhaps a lawsuit or the lottery) and now want to cash out.

It's a bigger market than you may think -- hundreds of thousands of potential clients. But, that's still pretty niche compared to the millions of TV viewers J.G. Wentworth's TV ads reached. So, the firm's marketers were thrilled by the targeting possibilities of the Web.

Their goal: to get consumers to fill out online qualification forms so the sales team could focus on closing the best prospects, instead of dealing with a deluge of incoming unqualified calls.

J.G. Wentworth launched its first website, featuring three lead generation devices, in April 1997. Then, for the next six years, the marketing team tested loads of online ads.

Frustratingly, no matter how creative the team got, results were "blah-blah," says CEO Michael Goodman. "We invested an awful lot without the expected return."

CAMPAIGN
Goodman was fed up with online advertising creativity. He wanted results. So he hired a new online advertising team who were all crackerjack mathematicians.

Tired of flushing good money after bad, Goodman gave them a tight deadline. The new team had just six months to launch an entirely new online ad program that delivered higher quality leads to his call center at a lower cost than TV.

Step #1. Constructing Algorithms

"We looked at conversion rates up and down the pipeline," explains Goodman. The team discovered they needed to meticulously track 11 specific points during the online lead generation process to judge and optimize any ad campaign.

Final algorithm factors included:

- cost per ad impression

- basic click rate, landing page conversion and follow-up email campaign metrics

- cost per lead generated

- qualification-level of each lead and whether it needed further qualification steps prior to handing to agents

- typical agent phone time required to close lead

- which types of agents (skill set) work best for typical leads from that source

- best contact time for phone calls to/from qualified leads (i.e., call center activity by daypart)

- value of closed leads

Step #2. Testing media buys

After reviewing past campaign data, the team allocated roughly 80% of their media budget for predictably "sure thing" sites, and 20% for new tests. If all tests were winners, the team felt like failures for playing it too safe.

They knew they needed some failed tests to prove they were reaching out as aggressively as possible to discover new media placements. Only that aggression could turn the Web from a trickle to a river of leads.

Tests included sites ranging from Publishers Clearing House to Yahoo Automotive. They also tested: run of site versus hand-picked sections; behavioral targeting; contextual ads; PPC search with highly targeted keywords; PPC search with more general keywords; text links; co-registration email info offers.

Step #3. Testing ads and landing pages

The team mined J.G. Wentworth's years of broadcast and print ad tests to learn what wording worked for the target audience. Their new online ads often used the same exact wording, with headlines such as, "Stop waiting for your money! Get cash now."

However, for PC search so many copycat competitors were also running campaigns that the team decided to test a more branded approach, such as JG Wentworth: "Don't be fooled by pretenders. Get a quote from the industry leader!"

For graphical ads, the creative director often tested campaigns that matched the site the prospect was on or coming from. For example, the team created a "NASCAR-style landing page" for all clicks coming from ads on automotive sites. (See link below for sample.)

Every major part of the ad creative was tracked -- not just the ad in its entirety. For example, the team put several coded click links throughout Flash ads so they could determine which specific part of the ad instigated the click. (We know very few creative teams who do this - it's a great idea.)

Step #4. Creating lead response funnels

Let's face it, if your creative is blaring "Need cash now?" across the Web, you're going to get a heck of a lot of unqualified clicks.

The team also ran algorithms to determine which landing page and response process was best for leads coming from each media buy.

Then they created a back-end decision tree for both landing pages and email replies for each media buy. As much as possible was automated and frequently a/b tested.

Example: Leads coming from a source that generally sent a larger amount of unqualified prospects might see a landing page with more questions to answer. Or, those prospects might be sent a series of carefully timed emails to further qualify them as true leads prior to a rep spending time on the phone with them. (Link to sample email below.)



RESULTS
Internet-generated leads rose by 29.5% within six months, but that's to be expected as part of increasing media buys. The campaign metrics that really mattered were three critical cost savings:

- 32.8% decrease in cost per closed lead versus past Internet ads - 49% lower cost per closed lead than TV ads - 7%-10% decrease in call center costs

Plus, Goodman notes delightedly, "We run a 24/7 call center six and a half days per week. The [new algorithms] plot out activity dayparts for us. We're able to forecast call volume ebb and flow." This makes call center staffing and HR decisions far easier.

More results:

- Matching the ad creative and landing page to the style of the Web site the prospect is on can help improve results. The NASCAR-style landing page for automotive site-generated clicks has been a solid success.

- Although run of site media buys are often cheaper, they rarely perform well enough to be worth the upfront savings. The team feels their offer is too niche. PPC search campaigns have become so competitive that the team has ratcheted back on major engine spending. For example, the firm does not always advertise for its own name in Google although 10 competitors do. (J.G. Wentworth does have the #1 organic listing however.)

- Although contextual ads are sold by the same networks (Google, Yahoo/Overture) as search ads, the team notes "the rules don't apply. What's working on search engines is not necessarily working on contextual." Generally contextually-generated leads required more qualification steps. (We've heard this from many advertisers.)

- Behavioral networks can work, but the team warns against buying these ads blindly without investigating which sites your ad will appear on and how carefully the context is chosen. "There's a quick buck to be made [through unsavory practices]. That's what holds back the industry."

Final note: "Traditional media will always have a place," says Goodman. "But the Internet's potential is far stronger. It's a better way to target the audience and it's becoming better and better over time. That's the key."

Useful links related to this article

Creative samples: http://www.marketingsherpa.com/jgwent/study.html

DMi Partners - the interactive direct marketing firm who now run J.G. Wentworth's Internet campaigns and algorithms: http://www.dmipartners.com

J.G. Wentworth: http://www.jgwfunding.com