Finally, you’re on the front page of Digg! You’re server is hopefully prepared to handle the ridiculous amount of traffic it is about to see today. You check back in the next day to see how many new orders you have. Think about the possibilities: 10,000 visitors with your 4% conversion rate. You finally bring up the report…nothing. In fact, your conversion rate is probably lower than usual.

Writing that social media whale that will bring in thousands of visitors to your site is hard. You have to research the hook that will be most successful for your niche, and then successfully market it to get on the front page of Digg, Reddit, Mixx, or other social media sites. Despite the opportunities social media gives you, I still hear the same complaint over and over: the traffic doesn’t convert.

It is easy to understand your conversion concerns with social media. You probably wonder, “what good is all of that traffic if 0.01% of it actually buys something or clicks on one of my ads? I’d rather have my resources used on strategies that will get people to buy more.” But what so many people fail to realize is: social media marketing is not about conversion, it is about links.

The links that you get from a successful social media campaign will help you rank higher for your targeted keywords that will bring in the traffic that actually converts. Don’t expect any of the thousands of visitors you receive from Digg to convert, but you should expect them to link to your site from their blogs and from their websites. These links can easily be enough to take you straight to the first position for your most important keywords. Next thing you know, more and more of your quality traffic is converting.

A great way to filter out all of this non-converting traffic is to setup a filter in your Google Analytics. First, create a new profile so you still have that main profile that shows everything. Next, add a custom filter to get rid of all of the social media traffic coming your way:

Now your new profile will display all of the traffic and conversion rates, minus all of the non-converting traffic from social media. See how your traffic from the Search Engines improves, and how it affects your goals. You’ll be amazed at the indirect benefit social media can create for you.

“You really mix it up, Robbin, don’t you?” LunaMetrician Jason Green joked as he unpacked box after box of Melitta’s Breakfast Blend javapods.

“Well, that’s what you guys drink,” I answered. But in fact, that wasn’t the main reason that I bought all Breakfast Blend. Melitta just didn’t have the information that I needed to distinguish between their products. And I saw the same problem with the user tests we did in December, for a completely different kind of customer.

The problem was that the Melitta site made it too hard for me to buy anything else. I would have loved to try some new stuff, but all those coffee words didn’t mean anything to me.

For example, one of the flavors was, “A Cafe Kind of Day.” The description is, “The forecast calls for smooth and satisfying. The 100% Colombian brew delivers a subtle, wine-like overtone from daybreak to nightfall. Each box contains 18 pods. Fits all Coffee Pod Brewers.”

Seems simple? Not to me. I want to hear, “This is the perfect cup of coffee when you have that mid-afternoon sleepy feeling.” Or, “Just awesome when you crave something as dark as espresso, but can’t get to your machine.” I need some way to differentiate this kind of coffee from “Breakfast Blend.” With Breakfast Blend, I understand one important feature: it is for drinking in the morning. All those other blends — I just couldn’t tell the difference.

This is the same issue I saw last week in user testing. We had a site with product after product, and they were so similar. The owner certainly understood how and when to make a selection, because he knows the product selection intimately. Some of the users, however, were overwhelmed. He needed to tell them when to choose each one — just like Melitta needed to tell me how to choose a flavor of coffee.

Next time, I am going to find a site that has compatible coffee pods, and that tells me what to buy, and why. And we’ll try something new.

Robbin

Do long search phrases convert better?

This was what I wanted to find out for a particular client, but it took some work. I used a regular expression in the Keywords Report of Google Analytics to filter by the number of terms in the Keyword Phrase. The exported results showed a clear increase in conversion rate as the number of search terms increased.

This client was doing far better with searchers who were using a lot of terms. They were being specific! They knew just what they were looking for and were ready to buy. This data put additional power behind recommendations concerning content, search engine optimization and paid search strategies.

1 .59%

2 .60%

3 .90%

4 1.17%

5 1.06%

6 1.22%

7 1.88%

8 3.33%

Even though there were a lot of people using long search phrases, this data was obscured. As the number of terms increased, the number of people searching for exactly that phrase decreased. This resulted in none of the individual phrases seeming to count for much. The so-called Long Tail.

You really have to dig to find these sorts of gems but they are invaluable in the pursuit of providing information that can be acted upon.

A tool for digging

The tool is a Regular Expression, a pattern matching language. If you’re not already familiar with it, there is a great series of articles right here on the LunaMetrics blog.

Here is what I used:

^([\\+*"*\\s*,*'*\\-*]*\\w+\\b\\s*[\\+*"*\\s*,*'*\\-*]*){3}$

It accounts for the most common characters I’ve found between words.

Steve (see comments) pointed out a great way to shorten my expression by using the \W character set. Here is what it looks like.

^(\\W*\\w+\\b\\W*){3}$

\W is shorthand for all non-word characters

How do I use it?

I know this may look like gibberish but keep reading — you don’t need to understand it to get some use from it.

In Google Analytics, go to Traffic Sources > Keywords and paste the Regular Expression into the box at the bottom of the data. Just change the {3} to whatever number of terms you want to see and click the GO button.

A brief look at the RegEx

Although this is not strictly a Regular Expression post, I feel obligated to include a basic glance at the different parts of the expression. Feel free to skip this if you just don’t care.

^ anchors the beginning of the match to the beginning of the string

( ) used to group a set of items together for a match

[\+*"*\s*,*'*\-*]* This group matches any number and any order of + ” , – ‘ and whitepace (\s). It is what handles all the characters that might end up separating different search terms.

\w+ Matches 1 or more alphanumeric characters (the \w is another pre-defined set of characters like \s)

\b Match for a word boundary. It forces the \w characters to be separated by something. Otherwise the expression will match any string of characters longer than {3}.

{3} Requires exactly 3 of the above sequence so it would match the phrase one two three but not one two three four

$ anchors the end of the match to the end of the string

Don’t Sweat the Small Stuff

You can’t account for every situation. For example, sometimes ‘ is meant as an apostrophe and sometimes “-” is used as a hyphen. In the end the impact is usually small – just 2-3% of the search phrases were affected in my case and they just get bumped to the next higher match instead. (For example, non-glare window would match at {3} instead of {2})

It is an interesting way to look at keyword data and maybe you’ll get some use from it– if you do, let me know.

“I really want to download that song,” my anti-war teenager whispered to me in the darkened theater. We were watching the (almost mandatory) music video about the National Guard that is showing here in the US on every screen, before every movie, it seems.

“Don’t you care that it’s about the US military?” I whispered back, incredulously.

“No,” she whispered in reply, “I just care that 3 Doors Down did it.”

Today is the end of the holiday weekend, and as she was getting ready to go back to school, I asked her if she had downloaded it. “I didn’t,” she answered, “They wanted my name and my email address. I’m just the right age, I’m sure they’ll try to recruit me. I’ll get a ton of spam from them.” I pointed out that she could easily just create an alternative email account and not worry about the spam. Which is exactly what she did.

The National Guard is really foolish. They have spent an incredible amount of money getting a truly great music video on movie screens nationwide. But instead of putting a wall in front of it (stopping lots and lots of people), they should give it away without asking for names. And they should make not just the music, but also the video, available for download (I only found the streaming video available.) It’s a great recruiting piece, and instead of stopping the conversion, they should let people take it and watch it and watch it. Isn’t that what you want, all your potential customers putting your advertising on their iPods? The US National Guard has that opportunity, but left as it is - email address required - lots of people will just leave the page (I’d like to see that page’s exit rate - I notice that they use Google Analytics), and plenty of others will just use an alternate address. One they will never look at again.

Now, that would make a great split test - because all that should matter is whether they get enough recruits…

So, I hit the wrong button and deleted this whole post. And I have never been so busy in my life. All I want to do is go lay down and sleep. So here is my post again on the 100% bounce rate, we’ll see if I know anything about WordPress slugs and if I can actually replace that page, or if it’s htaccess to the rescue…

My cell phone rang and I paused Grey’s Anatomy. “Oh, I’m so sorry,” the voice on the other end of the line said, “It must be evening out there on the East Coast.” He was the nicest guy in the world, with a really good business idea. And while we were talking, he asked me, “Why is my GA bounce rate 100%?”

I love problems like this. They are so black and white that the answers are almost always technical (and not human. We humans are messy.) So I instructed, “Tell me how many unique pages your analytics showed on the day when you had a 100% bounce rate. It’s on the left navigation, go to Content > Top Content.”

“Gee,” he said, “I can only see two pages. That can’t be right, we have hundreds of pages.”

So there was the problem. The site had very few pages tagged, and when the visitor leaves a tagged page and goes to an untagged page, the GA only knows that the visitor has left the site altogether — because for GA, if your page isn’t tagged, it might as well not exist. And since a bounce is a page view of one page and immediate exit without looking at another page on the site, the broken page tags artificially created a bounce. (I won’t go into the intricacies of the visit length not expiring yet.)

Here’s the opposite problem, which I saw today (while I was actually looking at a computer, instead of Grey’s Anatomy.) I showed one of our customers how his site was doing with the new GA Site Search. Now, one of the things that you can do is measure visits with site search and visits without site search along many metrics. But most of the time, it doesn’t make sense to use “bounce” in the same sentence as “visits with site search,” because most site search requires the visitor to hit the enter button and see two (usually different) pages — a page with search and a page with results. Ergo, no bounce at all for most on site search (depending on their architecture. This is actually an interesting problem, worth looking into in depth…)On the same topic: pages that get redirected in the browser (and not by the server) will often see their GA load, and then the user automatically gets two page views and whoops! no bounce, even if it is a crummy page.

So before we fall in love with bounce rate, we have to understand it well.

*****

Of course, I lost the comments, too. I have most of them in email, so will try to reproduce them here. Jacques commented that he would love to see how GA computes bounce and site search (I can’t find that one in my email). Daniel Shields wrote, “Bounce rate is a funny thing. In some instances it is a very important means to uncovering behaviors which indicate propensities. In other instances, it creates almost no value when improved upon. Context is ultimately where the difference is. As a metric which receives so much attention from a broad, utilitarian perspective, I find it almost completely useless in itself.

“I attribute my insolence mainly to the fact that the bounce rate is dependent on important factors such as the linguistic relationship between where and what your audience searches for. Slight modifications to search algorithms in any direction can drive volatility in bounce rate. However, when bounce is analyzed with respect to keywords or referring domains, it becomes extremely useful in a marketing metrics toolbox. This, however, deals more with the idea of bounce rate as a technical performance indicator. I like that.

“It would be nice to have alerts built in for systems like this. Maybe someone close to Google should say something?:-)”

Finally, Alan weighed in from Paris, “Hi there Robbin, Bonsoir Jacques,

“You really got me thinking about this bounce rate issue and I think I came up with a legitimate scenario.

“It is quite common for certain types of sites to use result pages on their websites as landing pages for their PPC campaigns. It is even possible to automate this in AdWords by using the keyword insertion tag in the destination URL, e.g. www.mysite.com/search.php?kw={keyword} whereby the ‘bought’ keyword in AW will automatically be placed in the URL (I would advise against this however because of LPQ, but that’s a different matter).

“If site search is enabled on such a site with the ‘kw’ query specified, then GA will count this visitor as having used the search functionality from their very first page-view and all the visitors that bounce will contribute to the bounce rate in the Site Search reports.

“I have a couple of customers who had a site search bounce rate significant enough that it can’t just be explained away by what I call the “freak-occurrence factor” (i.e. 7-8% bounce rate and above). I cross-segmented their google(cpc) traffic by landing page and saw that most of them were indeed search result pages with the site’s search query variable in the URI.

“I suggest you check out the Landing Page report and filter on the search query variable to see if this makes sense for your customers too.

“Hope this explanation makes sense.”

And speaking of hoping, I hope that I got everyone’s comments correct, please complain loudly if I didn’t.

This was the question that someone asked on the Yahoo WA Forum today. “I work with B2B sites that have 2-5K visits a month,” the writer commented. A lot of things I learned at the eMetrics summit won’t work for me, he worried. In particular, he commented on multivariate testing, and how much easier that is for a site with a lot of traffic.

So is he right? Are those techniques mostly for large sites?

I think that split and MVT is harder for small sites, but not impossible. And let’s not assume that the game ends there, either.

So, let’s say that your site gets 5000 visits/month, and 20% of those people touch the page you want to test (e.g. the home page). So that’s 1000 visits/month, or 33 visits/day. If the conversion rate on the page is 2% and you are looking for/hoping for an increase to 2.5%, i.e. a 25% increase in your conversion rate, you can run four combinations and complete your test in a month. (If you want to check my math, go use my favorite tool, the Website Optimizer Calculator. ) Four combinations can be 2×2 or 1×4. Or, in real-people-speak, that would be two elements, like a headline and a photo, with a control and variation that you want to test for each of them. Or that could be one element, like a piece of copy, with a control and three variations you are testing. Either way, it is four combinations.

Granted, not 400, but helpful nonetheless. I find myself doing tests like this all the time, because I just want to know {fill in the blanks}. I just need to figure something out.

And how about A/B testing? One of my colleagues pointed out to me that an A/B test, early in the process, might be a lot more valuable than a whole bunch of multivariate combinations, especially when you can look at a site and say, “Oh my, how awful. We really think we know how to fix this.”

And not all optimization is about MVT or A/B..N. How about user testing? I really feel that user testing never gets enough press. It isn’t exciting like MVT, but I learn so much from it. When I present it to the customer, s/he often says, “Oh right, usability,” but the truth is, we learn things about the offer (”I would never spend that much money,”), we learn about trust, (”Well, ok, it’s your credit card,” they say to me, “But I would never give that site my card. I just don’t trust them.”) We learn that customers can’t even understand what the site does.

And now that you know that you can use the Google Website Optimizer for even the smallest of sites, head on over to sign up for one of the GWO upcoming webinars. There are going to be two webinars,  GWO for Newbies and GWO for Intermediate/Advanced folks. Tom Leung, the Chief Executive Officer of Website Optimizer says, send him questions ahead of time, and I really believe he will address them.

Introduction to Website Optimizer (New or inexperienced users)
Tuesday, October 30th, 2007 10:00 - 11:00am PDT
Register to attend

Website Optimizer: Creating & Launching Experiments (Intermediate and
advanced users)
Thursday, November 1st, 2007 10:00 - 11:00am PDT
Register to attend

Welcome to the new Google Analytics! I wasn’t at the announcement a few minutes ago in DC, but we announced Urchin 6.0 software, plus event tracking for GA and site search. I’ll write about event tracking over time, but first I want to write about site search — the new capabilities GA has to track how visitors search within the content of your own site.

At first, I wasn’t overly impressed. What’s the big deal here? I thought. I can already give my content > top content report the name of the parameter that pulls out my onsite searches, like search= or query= or even s=, and from there, get a report of what people searched once they were on my site. So who cares?

So, this proves that I wasn’t a very good beta tester. I had all this functionality and didn’t play with it (or even bother to understand it) until our team went to Google Analytics Authorized Consultant training last week. And there I saw Phil, the product manager, demo the functionality that I have been ignoring. (Wow, how cool! I thought. What analyst in her right mind wouldn’t want that in her reports?)

Sure, once you have site search analytics, you can see what was searched for, more easily. But it doesn’t stop there. You can see conversion rate and e-commerce metrics by on-site search term, just like you can with organic search terms coming in from search engines. You can also see what pages people were on when they did a search (so you could correlate that with entrance pages, and when the correlation isn’t very good, start to hypothesize that the navigation isn’t working for you on that particular page. Some people are just born searchers, and you have to assume that a search that is on an entrance page is very often a personality issue — they love to search — not a navigation issue.)

You can figure out how search helped you or not. For example, in the screenshot below, you’ll notice that those who used search on our site stayed for over 9 minutes, and those who didn’t, stayed for less than 2 minutes:

You can work with destination pages. For example you can choose a search term from the search terms menu, and then drill down and segment by destination page. In other words, when people typed in “red shoes,” did they choose the important red shoes page with the cool new red clogs, or did they choose that awful red shoes page with the old red ratty cheap sneakers? Or did they just exit in droves (no good destination page, one might assume?)

You can do other things with the usage (and I am sorry to skip around, but am working to keep the ideas with the screen shots.) For example, you can look at bounce rate by search (notice how our bounce rate goes way down if the visitor uses search). Or you can look at conversion rate, or time on site, or a whole bunch of other metrics.

So welcome to the new Google Analytics. Soon I will write about the new ga.js and event tracking. (But before that, I have filter articles to finish and pictures to show.)

Here at LunaMetrics, our “About Us” page is our third most popular page - right behind our  main blog page and our homepage.

In fact, “About Us” is a pretty popular page for lots of sites. But some site owners treat it like a necessary evil, instead of a place that they can sell their product and services in the most subtle and wonderful of manners. So I wanted to evaluate a variety of About Us pages, and since my company is as guilty as everyone else, I’ll start with LunaMetrics.

When you touch the “About Us” page of a consulting firm like ours, what do you want to know? Well, you want to know where they are located. You want to know if you trust the people, or if they are just going to steal your money. You want to feel like the people there are going to be your partners. And you want success stories. At the end of the day, you want to know that this is a company that you will feel really good about hiring. (See? I said that we are guilty of not having a great About Us page. Soon, soon.)

Similarly, we have a local customer who advertises themselves as a Pittsburgh company. They perceive that it’s one of their key differentiators. “So where are those Three Rivers pictures?” I wrote in the expert analysis. “I wouldn’t even be offended if your technical site’s About Us page ended, ‘Go Steelers!’ — it sure would make me believe that you’re a Pittsburgh company.”

Big, big companies can use their About Us page to humanize their company. Almost two years ago, I heard someone from Travelocity talk about their brand makeover. Users perceived that they were “just software” and not real people - and in fact, if you can find the Travelocity About Us page, you can read their Customer Bill of Rights, one of the campaigns they instituted as part of their “just us people” rebranding. Similarly, I didn’t believe that the people at Quicken were “real” until I saw this snapsnot.

Sure, we little companies can use our About Us page to make us ourselves look bigger, more important. The Internet is the great equalizer. But I think (and test, test, test, right?) that we work too hard at having fancy photos and not hard enough to sound human.

About Us is the place to say, “Believe in us.” So the question is always - what will it take the customer to believe in a company?

I just wanted to exchange my theater tickets today.

So I called the Pittsburgh Public Theater, and had that standard box office conversation (”When can you get me good alternate tickets, how about this date, try that one.”) And oh, by the way, I said to him, where is the seating chart on your new site?

The box office guy pointed it out to me over the phone, and proudly indicated that I could see any seat’s view on the website. (In fact, you can play along at home.) “Put in your row and seat number,” he instructed, and that one was easy, I could see the boxes right there, begging me to fill them in. Immediately, the seat I was going to get lit up. “But wait,” I complained, “I thought I would be able to see the view from that seat!” Well, in fact, the box office guy explained, you can see the view. Just roll your mouse over your seat.

Now that I look at it again, I do see the little type with the instructions. But — where you sit in a theater is incredibly important. My best friend, a theater addict, taught me that seating is everything. For the person who really cares where she sits (and that is me, and a lot of people like me), this is a great opportunity to make the sale. A fabulous feature. Not one to hide with little type.

When you buy shoes online, you think that there might be an opportunity to see them from various angles. So, it might be somewhat intuitive to click on the shoe. But when you buy tickets online, do you expect to be able to see the view? No. The feature is so cool and so new - this site needs to make a much bigger deal of it.

OK, go ahead, tell me about the theaters around the world that are already doing this.

I would add, maybe they have tested it and found that I am wrong. But given that I can’t find any WA, I pretty much doubt that they have tested anything. Pretty website, though.

Don’t you ever wonder about the computations of that little calculator that Google gives you to figure out the length of a multivariate test?

I don’t have any insider knowledge. But I have studied it enough to understand certain issues (and many thanks to Dylan Lewis of the Web Analytics Wiki for confirming my suspicions wrt how it should work.) Specifically, you should need more data to “prove the same thing” if your control has a higher conversion rate, up to a conversion rate of 50%.

So let’s start: why does the GWO calculator ask you to input the conversion rate of your current page? Well, here’s why they care. If you hold everything else the same and tell the calculator that your current conversion rate is 4% instead of 3%, it will want a larger sample (translation: more pageviews, or more time to get those more pageviews) in order to get the statistical significance it needs.

So look at these two examples. All the variables are the same (sort of — I promise I will explain.) However, in the examples below, one conversion rate was 3% and the other is 4%. Notice, also (here is the explanation just promised) that I changed the expected increase in conversion rate. With the 4% test, I have it expected to increase by 25% (so that I will get a one point lift in my conversion — after all, .25*4=1) And with the 3% test, it’s expected to increase by 33.33% (because 3% times .33333 is also a one point lift):

So when the current conversion rate is higher, and you are looking for the same absolute expected improvement, the test takes longer, so that you can get more pageviews - i.e. get a greater sample size.

Why?

Why do we need more data to prove improvement with a highly-converting page than with a poorly converting page? Here, I will use a more extreme example: an absolute increase of one point is pretty low when you are looking at a page that converts at 25%. So we need lots of data to prove that a test will do better than the 25%. But a one point increase — that a whopping increase if your control page converts at 4% right now. So we can prove that our new test is better than our old control with just a little bit of data in that situation.

Here’s the really interesting part: when your control has a conversion rate of 50%, you need the most pageviews, i.e. time to get those pageviews. As you keep going beyond 50%, the time to run your tests starts to decrease. When you get to a conversion rate of 75% for your control, the time it takes for the test should mirror the time it takes at a control conversion rate of 25%. (It’s not perfectly exact for mathematical reasons that are too boring to go into here.) But check it out:

(notice that 25* 10% is a 2.5 point lift, and 75* 3.33333 is a 2.5 lift in conversion rate, also.)

Why?

Why does it all turn around at 50%? And I want to try to explain this without using ps and qs and little hats, since I’m not a statistician. So I won’t use fancy equations. Just simple ones.

All these equations that are behind all these kinds of calculators, they include two events: heads or tails. Conversions or non-conversions. They never say (to the extent that they talk), “Conversion is good.” Only people think that conversion is good and non-conversion is bad. (Those equations also include other stuff, but we don’t have to go there.) In fact, you have to have five conversions and five non-conversions for a combination to show up in the graphical area of the website optimizer (the area where the bars are green and grey and red.)

So when you start playing with conversion times non-conversion, you find out that they multiply out to the largest amount when they are both 50%. Right? .5*.5= .25 but if you now use a little 2% conversion rate instead, you have .02*.98 = .0196. That’s way lower than .25 (and remember — this is not sample size, but is one of the important parts of the sample size equation.)

My fourth grade teacher, Mrs. Petrowski, insisted that I learn all those math laws, and one of them was about “commutativity” — it doesn’t matter what the order is in multiplication, you still get the same answer, she lectured. So we can swap those numbers and say that the conversion rate is 98%, leaving the non-conversion rate to be 2%, and the product is still .0196.

So whether you have a 98% conversion rate or a 2% conversion rate — your sample size is going to be the same. (Remember that there is a lot of other junk that goes into the equations, but this basic principle should hold, even though I don’t have access to the innards of the calculator.) And from all this gobbledygook we learn:

• To prove that a test is 1% better than the control, you need more pageviews if the control has a high conversion rate than you would if the control had a low conversion rate.
• However, once the control has a conversion rate over 50%, you start needing fewer pageviews.
• This is a hard topic. If you didn’t understand, please comment and I will do my best.

Whew. This post took me at least two months to write. Many thanks toDylan, again; to Wendi Malley; to Tom Leung (whom I have driven crazy on this topic); and to EV, the GWO engineer who must be sorry he ever gave me his email address.

Everyone who thinks that change in conversion rate should be viewed as a PERCENT and not as an absolute lift in conversion is welcome to flame in the comments.

Robbin