Yesterday, I read an article on increasing conversion rates. The author encouraged website owners to find ways to increase page views per visitor in order to increase conversion rates. Which got me to thinking about correlation and causation…

It is often true that visitors to e-commerce sites who look at more pages convert better. But it is not necessarily true that the extra page views cause the increased conversion — perhaps the visitor has a propensity to purchase and therefore looks around a little more. I would call this a “lurking variable” if I were a statistician.

Furthermore, e-commerce sites who try to cross-sell too much, and send their customers to new pages at the checkout stage may get higher page views but lower conversion. Asking the customer to leave the shopping cart to do anything other than “continue shopping” can be a mistake. Even wonderful, conversion-rich tactics like “View our privacy policy” next to the email field in the shopping cart should come up in a window that has no navigation — the customer reads it, closes it, and continues with the check-out process.

Of course, if you are making money based on advertising views, then get all the page views that you can.

Robbin
LunaMetrics

More and more web analytic companies have on-page analytics. If you haven’t seen them, they look just like your web pages in a coded (often color-coded) format so that you know where people are clicking. All the big packages (SiteCatalyst, HBX, Coremetrics) have them, as does WebTrends, and even intermediate packages like ClickTracks and Google/Urchin include on-page analytics. Note: this was not meant to be an exhaustive list. Also, some companies charge extra for their on-page analytics, so you might have the package and not have this feature.

They are very cool. There is no better way to determine if a link on a page that leads to a different spot on the same page is pulling its weight. They also make it easy to see where the visitor went next. Sure, you can pull a report to see where the visitor goes next, but a picture is worth a thousand words.

In some ways, though, on-page analytics are dangerous — as dangerous as Google AdWords used to be before Google implemented conversion tracking a couple of years ago. Before conversion tracking, Google customers would say, “Look how many clicks I got!” instead of evaluating which clicks turned into customers. On-page analytics have the same problem.

If you only look at your on-page reports (and some people do), you can see that after the visitor lands on, for example, the Socks for Tweens category page, she clicks more often on Toe Socks than on any other kind of socks. A winner, yes? Not always. It might be the case that Toe Socks are a real loser — visitors click to it and they don’t get the color choices they expected. So they leave. You’ll look at your on-page analytics and wonder why your most popular product has so few orders.

Furthermore, clicks are people, too, and you can’t tell how many of the clicks came from repeat visitors or new visitors. You can’t tell how many came from a customer who is using your on-site search engine. Or how many came from paid search vs. organic. In other words, you can’t filter them.

At the end of the day, they only show you what the average visitors clicks on. Not who the visitor is, what he buys, or where he leaves. They are incredibly valuable, but they are only one more tool in the analyst’s kit.

Robbin
LunaMetrics

It’s not really vacation yet (unless you started yours early.) But the phone isn’t ringing and inboxes are only filled with electronic greeting cards. What a great time for web analytics — time to do all those things you really want to get at.

I am still awed at the number of people who have 125 page reports for analytics. I train my customers (and myself!) to look at their dashboards every day — not their 125 page reports. Who has time for that? I run exception reports (now, how many people did touch “About Us” before they saw the “Thank You for Your Order page?”) all the time, and then I report selectively. And I look at my dashboards.

Last night, I created a press release for some conversion rate successes that one of my customers has been having. The CEO read the release and wrote back, “What are KPIs?” I had to laugh at that one — I had created an atmosphere where dashboards were “business as usual” in such a big way that he didn’t even know they had a fancy name.

So this is a great opportunity. You can go out into the conference room and eat more chocolate or spend a few minutes creating Key Performance Indicators. Put them into a dashboard that you look at every day. If you know that you’ll forget, have it emailed to you daily. If you don’t know how to do it, send me email and I’ll try to help you.

Robbin
LunaMetrics

WebTrends recently published a 34 page report on analytics (which you can get here.) This finding really amazed me:

In our own 2005 Web Marketing Confidence Report, we surveyed over 250 marketing professionals and learned that only 5% are currently “very confident” in the measurement of their results. In fact, 26% admitted to “flying blind”. Why? Because less than one out of four had a complete view of their performance metrics—including conversion, revenue, ROI—to gauge the success or failure of their marketing initiatives. The majority, a combined 52%, are still relying on clickthrough rates or have no campaign metrics at all.

For those of you who aren’t looking at your metrics every single day — don’t you wonder if you are shouting into the wind? (No one can hear you?) And if you respond, “Well, we have e-commerce so we can see that people are purchasing,” you are leaving lots of money on the table by not having analytics.

When I first got this blog, I was sure there were web analytics attached, and I just couldn’t find the right button. It drove me slightly crazy to blog and blog but not know — was anybody listening?

Eventually I realized that there was no button, I just had to install my own analytics. I chose a small, free, client-side package because a) I only want to measure traffic and marketing referrals and b) I need a client-side package in order to be able to measure this particular site.

Analytics are magical. They are instant feedback. When I installed analytics here, I learned that I wasn’t shouting into the wind. More interestingly, I learned that more of my audience wanted to hear about analytics, as opposed to just great ways to increase conversion rate, and so, I’ve started to concentrate on that topic a little more (as you see!)

Robbin
LunaMetrics

I don’t usually blog about other people’s blogs, but Eric Peterson of Jupiter Research fame has done a fabulous job of creating a web analytics forum on his blog. {It is now 16 months since I wrote this, and I should point out that Eric has given the forum to the WAA to administer and run. But as usual, I digress…} I set up the RSS yesterday and got fifteen threads fed through to me. (And that was just for one day.) I do wish that each conversation had only one thread, but I can handle the duplication given how involved the analysts are.

Robbin
LunaMetrics

I do an incredibly amount of work in analytics in order to increase conversion rates, and one of the things I do is help the tech-types create better reports. (Well, I help them because it helps me, but that’s teamwork, right?) So, for example, I helped one customer take his thousands of products and put them into about fifteen categories. That way, we can create a product report with the click of a mouse (and not get a list of meaningless part numbers.)

The report defaults to Product Category and Revenue. This week, one of the company’s employees created a third column, product page views. “Look,” he said in an email, “Product Category A gets hundreds of thousands of page views and very little revenue. Product Category B gets few page views and lots of revenue.”

In general, this is a great kind of analysis to do with one’s analytics — for each product, what is the purchase to view ratio? Often, we find that “hot” products do well only because they are seen often (e.g. they’re on the home page.) Notice that I wrote, “For each product.” However, the employee’s analysis was way off base. What he didn’t realize is that if a page includes 47 products, the number of category page views increases 47 times when a visitor looks at it once.

There is a more important point here, though. Whenever I have some big analytic pronouncement to make, I do two things. 1) I ask, does this make sense? In the case I described above, the numbers were way out of line. 2) I triangulate. If the data leads me to a strange or unexpected conclusion, can I find it in a different way? In this particular case, I looked at page views of individual products and saw the problem right away. But triangulation can take other forms. For example, maybe you do have a really awesome analytics package. Have you thought about using a little free one at the same time? That enables you to say, is this a problem with the package I am using, with how I am interpreting the report, or is this truly something that matters?

Recently, I did some keyword work for this same customer. They have one of those big functional analytics packages, SiteCatalyst by Omniture. But I still installed Google’s conversion tracking and waited a month (to collect data) before making any drastic decisions. The SiteCatalyst report gave me lots of data that I couldn’t get from the Google AdWords conversion tracking, such as conversion over multiple visits. Nonetheless, having the Google data made me feel really good about making an important AdWords decision, because I could tell from the Google data that the Omniture data was in the right ballpark.

But I guess that to truly triangulate, you need three sources of data….

Robbin
LunaMetrics

Just last week, I blogged about writing for the web. If we write as if the visitor is standing in front of us, we can have a more personal exchange (and convert more visitors into customers.)

Then a couple of days ago, I signed up to go to a usability conference. Today, the usability guys sent me back a form letter:

Robbin,

I think it’s great that you’ve decided to join us for the UIE
Roadshow. It’s shaping up to be a fun and exciting day, one that
you’ll find yourself talking about for months to come.

We’ve already had a bunch of people sign up. I’m expecting that
we’ll be near the room’s capacity in each city. It’s good that
you registered early.

Sound like a form letter to you? Right, it doesn’t sound anything like a form letter. They wrote it as if it were a personal letter, from them to me. How fabulous!

Wouldn’t it be great if we could write on our sites like we write to our friends? We sound so genuine that way — potential customers must just believe us.

Robbin
LunaMetrics

I write about exit surveys a lot because surveys are a great way to find out what visitors are really thinking (instead of playing guess-and-check all the time, the way that we web analysts do.) This time, I get to write something nice about a survey. (I know, you think I can only critize…)

My HP printer jammed up, so I finally got on the HP website for guidance. At some point, the site served up this message, “We’d like to ask you about your experience on our website. May we send you email in the next 24 hours?” The message asked for my email address and told me that my address would always be kept private. It was easy to answer “yes” because they didn’t insist that I drop everything to do their survey at that instant, and because they gave me privacy assurances at just the right moment.

When the email came, I had the opportunity to let it sit in my inbox until I was ready. I didn’t particularly love the survey — why can’t companies ask five good questions instead of 25 questions that all seem hard to understand? — but this time, they approached me in the right way.

Robbin Steif, CEO
LunaMetrics

This morning, I sat down to sign up for the UIE Usability Roadshow. I was very impressed at a number of features in their shopping cart. For example, as soon as they asked for my credit card, they included the words “Note: This is a secure server.” And then, when I got to the bottom of the page, it said, “You’re not quite done registering yet. One more step to go. Please proceed to the final step and review your order.” Talk about a friendly message that helps you understand where in the process you are!

But when I got to the final page, where I had to confirm my order, I didn’t get enough information. They told me the price and the city that the seminar would be in, but the dates weren’t there. “When is that seminar again?” I thought. So I had to go back to my email and find their message and open up another browser and find the date of the seminar, which wasn’t even in the same month that I had expected.

I’m sure it will still be a great seminar.

Robbin
LunaMetrics

Yesterday, I started working with a new customer. On her home page, she had text that read: New clothing by Company-A New Release! IRRESISTABLE!!! WOMENS MENS PANTS SHORTS sweaters turtlenecks mocks. CARDIGANS coming soon! (Words changed slightly to create a different product category.)

I called her and this was the analogy I gave her: Imagine if you walked into a department store and the salesperson said, “May I help you?” You answer, “No thanks, just looking.” Instead of accepting that, the salesperson says, “NO, NO, you MUST look at these AMAZING new products!!! You are going to LOVE them!! Now come with me!!”

The customer got it immediately. “I would run from that salesperson as soon as possible.” This morning, when I got on her site, I saw that she had already edited them to sound like a salesperson you would want to talk to.

I always come back to the same caution — whether we are writing error messages or product benefits, we should write in the same tone that we would use when speaking to a customer in person.

Robbin
LunaMetrics