Follow the Money

Not long ago, direct marketers might have waited three to six months before they knew if an advertising campaign was producing desired results. But that was before e-commerce, sponsored search engine marketing, and most significantly, analytic software that captures information on how many consumers click on and look at an online listing, and which listings lead to the most sales.

The rise of web analytics meant that marketers who paid for a high position on a given keyword on a given search engine could now figure out the value of that keyword investment in terms of its return in conversions. Able to calculate return and plan marketing spending with a new level of precision, it wasn’t long before marketers started thinking beyond what happened offline, and about how analytics could be used to track customer behavior that originates with an online search all the way out to where most retail sales are transacted: in the call center or in stores.

A good starting point

New generations of web analytics are seeking to answer that question. “You can look at what people view on the web site, and what they ultimately end up purchasing. You can track if a keyword leads to an offline purchase of an item associated with the keyword. That is indicative of the kinds of capabilities people will ultimately be building,” says Forrester Research Inc. analyst Bob Chatham. “Web analytics is a good starting point for the profile of that kind of information.”

Staples Inc. has a wealth of transactional data on what customers buy online, from its catalog and in its stores. With a recent launch of Coremetrics Inc.’s LIVE profile system, which captures customer-specific online behavior over time, Staples aims to boost its integration of transactional behavior with customer behavior data for more insight into what online campaigns do to drive activity across channels. “We are working on the ability to see that a searcher came in off Google to the web site, for example, looked at a product, didn’t buy it then but walked into the store later to purchase it and put it on their Staples reward card. You want to be able to attribute those sales appropriately for the lead- generating entity,” says Colin Hynes, director of usability as Staples.

But as Hynes himself says, it’s still a challenge to stitch that cross-channel information together. As a first step beyond measuring the profitability of a search campaign solely in terms of direct conversions off a keyword, some companies are starting by seeking to quantify the value of keywords that produce conversions which lag the initial click on a search listing. Capturing these conversions, though they may happen days or weeks later, provides a more accurate assessment of keyword value. Some analytics packages can track such latent ROI, and some marketers are putting them to work. Such information can present the value of keywords in an entirely new light.

Matt Belkin, vice president of the best practices group at hosted analytics provider Omniture Inc., offers an example. “Say our company buys the term, ‘web analytics’ for our lead generation site, Omniture.com,” he says. “If we measure a click-though rate of 1%, that is a pretty bad click-through rate, so based on that, we might decide to stop bidding on ‘web analytics.’ But we really have no visibility as to whether that click-through eventually generates leads and revenue for us.”

The next level

Taking it to the next level would mean using analytics to measure the searcher’s behavior beyond the initial click, but still within the same user session. If searchers who did click through on “web analytics” view information on the site but leave without taking any action, the keyword has produced a click, but no conversion, and signs still point to an unsuccessful keyword campaign. Configure analytics to capture multi-session information, by customer—which Omniture’s SiteCatalyst tool does by means of cookies or other assigned identifiers—and a different view of the term’s marketing value emerges.

If a searcher clicks through on the keyword, returns seven days later and fills out an online request for further information, the keyword has ultimately resulted in a conversion. “Because we measure multiple sessions we can say that although it took seven days instead of happening in the initial session, the keyword has generated a lead for us,” says Belkin. Marketers in the search arena managing keyword investment to a targeted ROI typically aim for a target in excess of 10 times return on an average spend, he adds. “With this kind of visibility you can much more effectively manage your marketing dollars,” he adds.

Belkin grants that though it represents best practice, multi- session tracking and measurement is still in its infancy among online retailers. But if capturing the value of a click across more than one session more accurately quantifies the value of a keyword, it stands to reason the picture would become even clearer if search marketers could also measure any activity the keyword drives in the call center when searchers jump from the web to the toll free number displayed on the site. Some marketers are using analytics to do that as well, but the cross-channel link-up is more challenging than tracking activity that occurs online only.

A unique number

One Omniture customer, an online furniture retailer, knew that while browsing its web site played a critical role in the selection process, 50% of its business actually closed in its call center. To capture the web-to-phone crossover—and ultimately, the true vale of the keywords that brought searchers to the web site—Omniture established a unique user ID number for every visitor who entered the site that was tied to the keyword and the search engine that brought them to the site, then to the call center. The number appeared on the page when the visitor first entered the site and stayed with the visitor throughout the session. When a web visitor called the retailer’s call center, it was built into the call center agent’s script to ask for the unique customer number. The number was then fed back into Omniture’s SiteCatalyst analytics tool to close the loop and link any call center sale back to the keyword that led the visitor to the site.

The analytic data resulted in several changes to the retailer’s keyword campaign, Belkin says. “They had some keywords that absolutely were producing for them, but they had thought were not, so they had been taking money way from assets that were actually working for them. They’d been flying blind on 50% of their business,” he says.

Omniture’s system is one of number of ways analytics work to link the web to the call center to search campaigns. Retailers have used banks of phone numbers that are unique to designated, selected keywords to track results. Toshiba, for example, which in the past has manually managed about 50 toll-free numbers, is now working with Coremetrics on a system that will bring more automation to the process.

Automation is the only way a marketer looking to track a large keyword program could do so practically. But what happens when an online marketer has 10,000 or 20,000 keywords on both paid search engines? Even with automation, managing 40,000 unique toll free numbers would prove unwieldy. That’s an issue tackled by ClickPath, a web analytics provider whose system uses an algorithm to help shrink what might otherwise be the huge number of unique toll-free numbers needed to link a big keyword campaign to the call center into a smaller, more easily managed rotating pool.

Tracking popular keywords

“You might think that because you have 15,000 keywords on Google and 15,000 on Yahoo you need to have 15,000 toll free numbers on each engine to track those search campaigns to the call center. But we connect the call back to the keyword with a small pool of telephone numbers,” says Ted Carpenter, vice president of ClickPath, which is a unit of Who’s Calling Inc., a provider of telephonic sales lead management technology.

Guide to E-Commerce Technology

The hosted service tracks call center activity generated by campaigns where keywords number in the multiple thousands, without actually generating and tracking that same number of toll free numbers, by measuring the popularity of keywords in terms of how quickly and how often a keyword brings visitors to a site. Typically, says Carpenter, a small number of keywords in a large campaign will drive a very high percentage of the campaign’s traffic to the site. A larger number individually drive lower volume, but in the aggregate, account for substantial traffic.

As the analytics algorithm, with experience, begins to understand keyword popularity by measuring the volume of traffic the different terms bring in, it starts to dynamically generate telephone numbers from the pool for the most popular ones, and it inserts them into the web page on the site so they are visible to the users who came to that site through that keyword. The number, representing a call to action, stays visible to the user as he or she clicks through the site. Less popular keywords, when they actually do get a click and bring a searcher to the site, receive a dynamically generated number at that time from a rotation pool of phone numbers. After a keyword doesn’t get searched at all for a set amount of time, the toll-free number generated for it goes back into the pool for recycling and reuse.

Carpenter says that because users who come to the site through a paid search campaign and then contact the call center do so within the context of a unique toll free number, ClickPath can take that information and integrate it back into its web systems and match up those two interactions. “Then the web analytics portion of the product lets the site operator see all the calls that were made, and all the information surrounding those calls, with their online conversions,” he says.

ClickPath has yet to share any retailer’s experience with the system, which is a relatively recent launch, but a b2b client, payments processing equipment seller Merchant Warehouse, has been using it since January to track sales in its call center, where the majority of its sales are transacted, from people using the phone numbers dynamically generated through its online campaigns. Within the first week, Merchant Warehouse saw which keywords were performing and which weren’t, and it cut spending on underperforming terms, reallocating the funds to campaigns that were generating calls, which its model defines as conversion. By consistently identifying and spending on the keywords with the highest value and the lowest cost, the company cut its keyword costs by 75% over six months.

The elusive store information

Carpenter says one place he sees the system having application in online retail is where items have a high dollar value or are a high- consideration purchase, such as a home entertainment system. “We’ve got this swell of the Baby Boom generation that is used to using the phone,” he says. “For such purchases, they’re not about to forsake that for a strictly online transaction.”

While the analytics technology exists to track customers’ web search activity out to call centers, tracking the effects of online search out to store purchases remains more elusive. Consumers can shop anonymously in the retail environment, and in that situation, there’s no telling what advertising or marketing effort may have driven them into the store.

Online coupons for store redemption are one way to tie customers to campaigns, but they don’t always provide a complete view because customers who visit the store as the result of an online campaign don’t always remember to redeem the coupon. Loyalty or membership programs that provide ongoing benefits such as points or discounts to members afford analytics another opportunity to track the offline behavior of customers who are also online. In fact, Recreational Equipment Inc. has become so proficient at tracking members’ behavior across channels that it now purchases keywords for their ability to influence offline purchases as well their ability to generate conversions online, according to Forrester’s Chatham.

Programs that allow customers to register their purchases online provide another way to close the loop between web search and an eventual store purchase. Omniture, for example, worked with one client, a software publisher, on a program that assigned an identifying number to customers who downloaded a trial version of software. Eventual store purchases of the software could be tied back to the trial download when customers went online to register their purchased software. But again, not every customer chooses to register their purchase, so Omniture worked with the software publisher to understand what percentage of customers who bought the software did register. It then assigned a total value to the online search campaign by extrapolation.

A new CRM tool

If it seems as though analytics’ efforts to track the value of keyword campaigns across time and across channels goes beyond the realm of search marketing and folds in customer relationship management, post-sale support and other non-marketing retail disciplines, it’s because that is what’s required to connect the dots between channels. According to Forrester’s Chatham, “Web analytics is becoming customer analytics. Web analytics have the potential to grow beyond the web into a general customer analysis platform.”

That’s starting to roll out as marketers deploy analytics to build bridges between pairs of venues; for instance, tracking e-mail to web purchase, or web search to call center. “By breaking it down into pairs of channels at a time, it becomes a much more tractable problem,” he says.

Hynes says that in the past Staples has used means such as exit surveys attached to particular search results pages on Staples.com to ask visitors poised to exist from those pages questions on why they were leaving, such as whether they were leaving because they intended to purchase the item in a Staples store. Research Staples conducted to support its recent site redesign confirmed what the earlier research showed about the percentage of customers who searched online with the intention of buying in the store.

As Staples gets deeper into analysis of transactional data against customer data, “Our intention is to be able to attribute sales and not just the intention to buy, but seeing that the customer actually made it to the store and bought the product,” says Hynes. “That’s where we’ll be moving in the future, being able to pull those things together.”

mary@verticalwebmedia.com