You’re capturing recency, frequency and monetary data about your customers. You’re recording customer service contacts. You’re doing your best to capture point-of-sale information and correlate it with direct mail and telemarketing touch-point data. You’re trying to integrate your salesforce automation systems to keep everybody up to speed on customer contacts and customer desires.
Then along comes the World Wide Web, bringing with it something called Web analytics.
Web analytics is the art and science of tracking what individuals actually do on your website and how they respond to different stimuli (advertising, marketing and customer service). Web analytics allows you to get more granular in your knowledge of each customer. You are moving from the age of demographics, through the age of segmentation and into the age of behavioral analysis. The effect on the bottom line and on customer satisfaction is significant.
At the start, Web analytics consisted of log-file analysis: sifting through log files that Web servers created as a matter of course (see sidebar). These were never meant to be a business tool, just a transactional record of every file sent out to 0.
Marketing departments around the world have grown in intelligence in the past few years. They’ve started using Web analytics to measure advertising and marketing, rather than just sales results. The persuasion process has been instrumented and diagnosed. Small changes were made to see how they impacted a prospect’s ability to get from awareness, to interest, to intent and, finally, to purchase.
We’ve long played with shopping-cart analysis to upsell and cross-sell, but today’s eye-in-the-sky, over-the-shoulder surveillance gives you an edge you didn’t have before: the ability to classify your customers based on their behavior.
Sign up for a newsletter at any maternity or baby-focused website, and you’ll be asked when the baby is due or how old your infant is. The needs of a pregnant woman or a new mother change almost weekly. This industry knows that every woman will go through the same stages, so segmenting those women by month is simple.
It’s a bit more complex to segment your customers when they are buying wrenches. But Snap-on, Inc., purveyor of high-end automotive repair paraphernalia, knows it can be done. Sending the same sales rep to talk to the same automotive service technicians month after month made it clear that these customers move through an occupational arc, choosing their specialties along the way. Each specialty requires different mechanical and electronic implements to help them get the job done.
Detailed information on tools and their applications in the Snap-on online catalog coupled with online training turn those sales reps in the field into career counselors. By recommending an educational path to a technician and following his progress through the courses, a unique bond is formed. CRM? Sure, if you help your customers advance their skills and track their progress, you’re ready to fulfill their needs as those needs arise.
THE TECHNICAL EXPLANATION
The CMO heads over to the CIO and says, “Tell me what’s happening on my website.”
CIO: “What do you want to know?”
CMO: “Well, what have you got?”
CIO: “Well, what do you need?!”
CMO: “Whatever you can tell me!”
In an attempt to halt this circular madness, we start with “What have you got?” and then move on to “How do we extract insights from that data?” There are a variety of ways to collect data about what people are doing on a website. Here they are in the briefest terms possible:
Is your server serving quickly? Is it overloaded? Are your e-commerce applications running smoothly? Is your server sending out “404, File Not Found” messages? Performance monitors watch the technical side of things and give you a holler if something goes bump in the site.
Every Web server makes a log entry of every page it sends out. By using log file analysis tools, you can glean such interesting things as number of visits and page views, entry and exit pages, hourly usage, browser and operating system used, and usage by country. All of which come with the caveat: sort of. Each of these measurements has some inherent irregularity, so take them all with a good-sized grain of salt.
If the visitor comes to your site via a link on another page, your Web server will also record the URL of that previous page. This is enormously informative if you’re buying banner ad space or tracking the success of a pay-per-click campaign. Did banner ad A outperform B? Did the press release cause a stir? Are people finding you while searching for a word you didn’t expect? Interesting stuff.
Server logs have trouble confidently revealing what an individual did during a single visit, so tagged URLs were designed for session or clickpath analysis. When a home page is served, all of the links on the page are tagged with a sequential ID number. A link that normally looks like this: www.company.com/page2.html, is tagged by the server on the way out the door to look like this: www.company.com/page2.html?127, indicating that I’m the 127th person to visit today. When I click on that link, the server records that page2.html was requested by visitor number 127 and tags the links on the requested page with 127 as well.
Server logs can rarely recognize individuals from one day to the next, so someone came up with cookies. For our purposes, it’s enough to know that you can track an individual over time using a cookie, and you can correlate his traffic patterns with any personal information he may have divulged while on your site.
Information moves online from here to there inside small data packets. Your home page is divided into capsules of about 1,000 bytes of data, and off they go, finding their way to your visitor’s screen. Should the visitor type something into a form and hit the Submit button, the entered data gets packetized and sent back to you. By sniffing the packet on the way back, you not only know which Submit button this person clicked (top of the page, middle or bottom) but you also capture the entered data. Now you have a record of the click as well as the information that would normally go straight into an application database and require a good deal of extract, transfer and load effort to match up with the server activity information.
Web Beacon (cleardot.gif)
Web pages are often stored in temporary cache files on the visitor’s computer and in cache farms of ISPs to minimize the time it takes to go back to a previously viewed page. This minimizes the overall traffic on the Internet and improves the customer experience on your site. It also means that your server doesn’t always know when a page is viewed. Embedding a transparent, 1-pixel-by-1-pixel image designated as “no-cache” can help identify when each page is viewed. If such a graphic is served from a third-party server, the aggregate information about cross-server and cross-site traffic can be useful.
A small Java program watches every move you make. Every mouseover. Every keystroke. This method can collect the most complete picture of what an individual does on your site. The data can be used to play back a visit and can capture data entered into a form even if the form is not completed or the Clear button is used. Very powerful, somewhat spooky. Full disclosure and full opt-in are highly recommended.
If you’d like to look over peoples’ shoulders and ask them personal questions at the same time, the panel research method is the way to go. Each panelist agrees to have client-side surveillance software on his computer and is willing to respond to the occasional pop-up asking them why he made certain click choices or how he feels about the color of the site’s background, etc. You can also learn how he feels about the competition.
If you want to get up close and personal, there are a number of firms measuring what individuals actually look at on individual pages of your site. You’d be surprised at what people simply do not see online.
Gross Traffic Patterns
A handful of companies are learning what people do online by watching from the vantage point of the ISP. Millions of people looking at millions of sites per day reveals which sites are the most popular. For the price of admission, you can learn where people (in general) were before they came to your site and where they went afterward.
All of the above explains why the CIO is so quick to respond with, “What do you need?” There is so much data available (granted, some of it is rather expensive to gather), that you must stop asking for anything and figure out what you want in particular.
WHAT THEY DO INSTEAD OF WHO THEY ARE
Overstock.com is a closeout retailer offering discount, brand-name liquidation merchandise to the tune of a couple of hundred million dollars per year. CEO Patrick Byrne is a big proponent of clustering. He wants to know which of his customers are the same and how they differ. But he’s not a fan of demographics.
“When you add up everything you know about your online customers, you can get rid of education, income, children and the rest of the demographics,” Byrne said at the Web Analytics Summit of the eTail 2004 conference. “Demographics are foggy data. You can only guess and surmise why people who live in the same postal code might or might not shop the same way, want the same stuff or be willing to pay the same price.
“But behavioral data is sharp,” Byrne continued. “Customers’ behavior is a signpost to their needs, so we don’t cluster people based on who they are or what car they drive but by their needs as expressed through their behavior.”
How an individual uses a retail site provides more insight into how he or she might be persuaded to buy than does martial status, age or income. “Men on a diabetes website will click on an electronics ad seven times more often than the same ad on all other sites. Why? You’ll never know – you can only surmise. And it doesn’t matter! Just take note and make use of that knowledge.”
Byrne offered another example: If somebody buys electronics on their first trip to the site, they always buy electronics during the second visit. If a first-time buyer does not buy again within 45 days of the first purchase, they are lost forever.So what can you do with that information?Make sure the electronics buyer isn’t given too much of a discount on his next visit and send a special offer on the 44th day if he doesn’t come back at all.
CRM From the Other End of the Telescope
Web analytics gives us yet another window through which to view the customer.But Dianne Binford, director of consumer direct marketing at Nine West, wants to turn CRM on its head. “A 360-degree view of the customer is all well and good. But we also need to give the customer a 360-degree view of the brand,” Binford said,also at the eTail 2004 conference.
Integrated marketing means ensuring that the direct mail piece, the television ad, the in-store displays and the website all look alike. Binford is adamant that customers see a united front when they look at Nine West, no matter what angle, no matter what method and no matter what medium.
When you combine Overstock’s behavioral clustering, Snap-on’s hands-on approach to customer care and NineWest’s desire to integrate its countenance,you move into the realm of microbranding.
Microbranding is the art of determining what type of company an individual customer is most likely to respond well to and presenting that face to that customer.Every time that customer sees the brand,in whatever channel, it looks cohesive and familiar and is tailored to them at that point in time.
Amazon was first to offer books based on collaborative filtering. If you and 500other people like the same 10 books,then the chances are pretty good that you’ll like additional books those other 500 like.
We have reached the end of the practical road when it comes to data capture.You can now collect more data about an individual during a single website visit than you can realistically use.
Web analytics systems have become sophisticated enough to track every mouse movement, every click and every keystroke. Rather than merely recording which links customers click,you can now see which links they considered clicking when they hovered over them with their mouse. You can see how long it takes them to absorb a page.You can tell when their hand leaves the mouse – a sure sign that they took a Websurfing break.
For purposes of measuring the effectiveness of an individual page, this is an absolute gold mine. Every mouse twitch and every nuance of a customer’s visit to your website can be trapped and played back. But for the purposes of building a profile of who she is and what her needs might be, there is simply too much data.You need a balance between what you can record and what you can use.
Let The Computer Decide
One approach to this info glut is to put the power of the computer to work on the problem.
Let’s say you bid on a keyword or create a banner ad. People who click can be directed to one of several landing pages. You can then track which page is convincing more people to buy. Better,however, is to implement a system that creates multiple permutations to see which landing element combination (headline,body copy, photo, call to action) has the most positive impact.
Several vendors are offering optimization systems to create a testing matrix. For example, well-known provider Optimost says it can “generate up to 1 million permutations of the creative execution. Through sophisticated experimental design techniques, Optimost will then test certain combinations of variables to arrive at the optimal values along each parameter.”
In other words, it’s the computer’s job to figure out which variables matter, which variable values are improvements and which variables barely move the needle.The more people touching down on your landing page, the faster the system will show which combination of text, graphics and layouts is getting the most people to do what you want them to do.
A Web analytics system can chronicle how the majority of people are attracted to a website and feed results back to an advertising system so that only the most attractive promotions are reproduced. It’s Darwinism at the speed of light.
All The Way to Loyalty
More sophisticated systems can track how the majority of people are attracted o a website, how they wander through its pages and how they do or do not buy.This type of system can watch which promotions result in the most profitable sales. It can feed results back to an advertising system so that only the most profitable promotions are repeated. It can feed results back to the website design team for insights on how to improve navigation.It can help online merchandisers determine which website elements might have the largest positive impact on conversion rates.
It takes one more step, however, to turn this marketing machine into a customer relationship management engine.This near-perfect system needs to have a longer-term view of the outcome.Instead of being programmed to reproduce the promotions, navigation and merchandising with the most profitable sales, it would detect which generate the most profitable customers. It needs to look all the way downstream to lifetime value.There are some companies working on these solutions, but they aren’t quite ready for prime time.
Your customers are coming to your website for information, transactions and to get help. Are you making the most of the information they leave behind?
Gleaning Meaning – Asking the Next Question
You know the standard questions to ask: Why did we get a spike in traffic to this section of the site for a solid week? Why are people who click through from search engines dropping out on page three of a five-page process? How much impact can we have if we tweak this page? Where does our advertising spend start to experience diminishing returns? If I shorten the copy on a page by 20 percent, does that increase customer satisfaction? Do they buy more? How much less clutter can I publish and still make sales? What changes in process completion should have alarms and what should the threshold be?
There is no Web analytics magic here. The standard questions are good ones and they help with incremental, continuous improvement. That’s all well and good. But the best question is the next question. It’s all about intuition, correlation and curiosity.
A large retailer in the U.S. found only a small percentage of its site visitors were Macintosh users. It seemed silly to devote the necessary resources to redevelop the website for the express use of a group that made up less than 10 percent of visitors. As the discussion progressed about ceasing dual development to accommodate this small group, one of the technologists asked the next question: What percentage of Mac users are buyers and how does that compare to Windows users?
The answer was surprising. While Windows visitors exhibited a conversion rate around 2 percent, Mac users’ ratio was up around 20 percent. That represented a healthy chunk of overall revenue. The next question had been asked, and the dual development continued.
The popular British online betting company William Hill noticed that only a small percentage of site visitors were viewing special, additional content. This content was expensive to create and maintain and was on the chopping block for the new website rollout. Then somebody asked the next question: Was there something special about those visitors who did view the special content?
Yes, there was. They were responsible for an unusually high percentage of site revenue. They placed more and larger bets. Rather that trash the content, William Hill Emailed the rest of its customers to tell them about the value of that content. The result was increased revenue, higher customer satisfaction and no measurable increase in costs.
A Web analytics tool does not provide a wealth of answers. Instead, it is an unending source of questions. Every good analyst uses the tools to sift through the data, searching for patterns and asking the next question.
So what’s the next question for your website?