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.

Yesterday, GA made one of their new features, Site Search analysis, available to everyone with a GA account. I started to set it up for my customers, and realized that I didn’t know what choices to make. (They give you choices, but don’t tell you the implications.)

So I did some testing, and here is what I found.

First, you’ll find the abilities to set up Search under the edit Profile settings:

Click the Edit and you’ll get this search setup:

Look to the bottom of the page to find it — most of the options will only appear after you click on the “Do Track Site Search” radio button.

Now, you have to find your query parameters. Go to your website and do a search. Look up into the URL address bar, find where your search term is, and what query parameter goes with it. For example, you might have typed in gifts — and your search term might look like this in the url: &word= gifts. Or maybe it looks like this, &search=gifts. In the first case, your query param is word, and in the second, it is search. No matter, just put that value into the Query Parameter (required) box.

But what if you have two on site searches? For example, our site has a regular search and a blog search. In that case, you can put your query parameters in with a comma, like this: word,search. However, the new search doesn’t seem to know how to separate them. You’ll get all the search terms from both, but they’ll be in one long list. You can disaggregate them yourself by working with the Start Pages Report (see it on the left, under the Site Search, which you can find in the Content Reports.) Or your can set up a new profile for the second search.

Now you have to decide, should you strip your query parameters out of GA?I did two profiles last night, one with and one without query params stripped out. I found that stripping out the query parameters here is like stripping them out in the Main Profile Settings (go look at that top screen shot in this post - see how you have the chance to Exclude Query Parameters?) This does a few things for you:

1. 1. It aggregates all your URIs (so that you don’t have to look at a million pageviews)
   2. It thereby reduces the number of unique pageviews (because www.mysite.com?search=Robbin is a different pageview from www.mysite.com?search=LunaMetrics). This makes it less likely that you will go over GA’s pageview limit and just get the dreaded (other) in your top content
   3. It enables you to take the query parameters out here without messing up your search. This is golden! One of the problems we always encounter is deleting query parameters and breaking our analytics. This is usually about breaking our goals, if, say, one goal is only distinguished from another with a query parameter. By taking them out here, you still achieve the first two points, above, without breaking your searchalytics.

Ok, now you are done with search terms, but not with all the other junk in your search uri. Maybe that’s all you have (for example, our wordpress search just has a search term in it), but you often will have lots of other stuff. The site I did the testing on had a little search box, and below it, allowed you to search by drop down topic, by country, and by kind of material (book, white paper, etc.) Each of those is a category. First, I did a few pretend searches to find out how those things show up in the URI. Then, I used the bottom half of the new search setup, where it asks, Do you use categories for site search? I changed the radio button to YES, and then entered each one, followed by a comma. Again, I stripped out my parameters. (I compared the two profiles that I set up, and noticed that, as would be expected, only the parameters I specified got stripped out.) Again, I noticed that GA doesn’t disaggregate the categories, so if they all come through as a number (the way mine did), you have to create a system (three digit number from one kind of category, two digit numbers from another), put them in different profiles, or perhaps the easiest, just use a filter to rewrite them. And then you’ll know what category they came from.

Before you are done, go on over to Justin’s blog, and read about Search Implementation. He talks about two interesting topics, how to do the search when you don’t have a query parameter, and what to do when you have multiple searches that are really the same thing, such as misspellings.

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

I have been working on this series, Filters for Google Analytics, for almost six months now. This is the last part of the series I am going to write (at least in the foreseeable future.) You can get the full thread at the bottom of this post.

So once again, I will start with the question a lady asked me on Friday morning: how does one create exclude filters that function as if they were AND filters? How, she asked, can I create a profile and put on it two different exclude filters that work together: Exclude the visit if the visitor is a new visitor AND her language is German? We can’t do that with two separate exclude filters, because as soon as the first filter works, the visit gets thrown out. Exclude filters are, by nature, “or” filters.

(If you are a little lost, and don’t remember how multiple exclude filters work like OR filters and multiple include filters working like AND filters, and how you have to work to make them do the opposite — see this post that I wrote recently.)

Well, I had to think about this one, and then realized that it was just a great practice for this series on filters for GA. The reason why: It is both a rewrite (custom advanced filters) and an exclude (custom filter). Here is the overview: you rewrite the two fields into one field, and then you do an exclude on that one field, making sure that it matches both pieces of data.

If that was clear as mud, or you are more of a step-by-step “please show me” person (like I am) — then this is a good place to keep reading.

First, you use a custom advanced filter to rewrite the data so both pieces of info — the visitor language and visitor type, which is GA-Speak for “new” vs “returning” visitor — are in the same field. Like this:

Now we have both pieces of data in the same field. If we could look at them, they would look like this, concatenated together on the same line.

1. en-us/New Visitor
2. de/Returning Visitor
3. da/New Visitor

and so on. They would be the language code followed by a slash and the visitor type. (Right? That’s the way I set it up, $A1 captures everything in visitor language settings, then a slash, then $B1 captures everything in Visitor Type.) Above, I choose three languages for my examples: English-US (en-us), German (that’s de) and Danish (da).

Next, we create an exclude filter. The hypothetical example was to exclude a new visitor whose language is German - the two excludes that must work together. This is the heart of our problem, exclude two different kinds of variables. So here is our exclude:

So what does this say? It say, go to Custom Field 1, where we now have a list of concatenated languages and visitors types (en/Returning Visitor, and so forth). If one of those lines in the list matches de (which stands for German - Deutch, right?) and has a slash and then the word new, it’s a match, so please exclude it.

And that’s how you can exclude two different fields in Google Analytics at the same time.

If you would like to read all the other posts in this series:

• Part 1: Get ready with profiles and ReGex
• Part 2a: Filters for GA, What have we here?
• Part 2b: Why does anyone use filters?
• Part 3a: Easy Filter Implementation
• Part 3b: Custom Implementation
• Part 3c: How do you filter those utm links you create?
• Part 3d: When do you use them?
• Part 3e: Include vs Exclude filters, how do they work?
• Part 4a: Custom Advanced Filters
• Part 4b: More on Custom Advanced Filters
• Part 4c: Cascading Custom Advanced Filters

I should be writing some cool post after three days of Website Optimizer and Google Analytics Authorized Consultant training. Or writing more about yesterday’s GA and Urchin announcements (I will, I promise.) But first, while they are still fresh, I wanted to publish all my pictures from Google Analytics and Website Optimizer training.

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.)

I spoke about implementation Google Analytics on Friday, and someone asked me, “How do I create Exclude AND filters?” This is a cool question, but it has to wait until I write the info below on how Google applies your filters (and how that differs on Include vs Exclude filters.) Note: you can read the GA documentation on this one, which is not terrible.

You might consider include multiple filters to be AND filters, when you think about the way that GA applies all of them to a profile. So if you create a profile where an include filter specifies Visitor City = Pittsburgh, and the another include filter specifies Campaign Source = organic, you will filter in only visits that came from people in Pittsburgh who used the organic search to find you. It’s not good enough that they came from Pittsburgh and used paid search, or followed a link. Both criteria must be met.

Exclude filters, on the other hand, could be looked at as OR filters. If the visit matches any of the exclude filters in a single profile, it is thrown away. So using the same example as above, Visitor City = Pittsburgh, and the Campaign Source = organic, you will exclude anyone who comes from Pittsburgh, no matter what their campaign source. And you will exclude everyone who comes on an organic search, no matter what their city.

There are ways around the AND and OR issues. For example, let’s say that I wanted to include everyone from Pittsburgh OR Raleigh. I cannot do it with two separate city filters — right? Include filters are AND filters, so two separate filters would say: include everyone who comes from both Pittsburgh AND Raleigh, which is a pretty hard act. But you certainly could do include visitor city = (pittsburgh)|(raleigh). By putting both cities in the same filter with an OR pipe, you achieve that goal.

Now that I showed you how to defeat the “ANDness” of Include filters, you will be wanting to know, how to do defeat the “ORnesss” of Exclude filters. Plus, you might want to know how to make include filters be OR filters for different kinds of fields — include if they came from Pittsburgh OR they came organically.

But that is a different topic and a new post. Maybe even tomorrow.

Should WA consulting firms — or for that matter, conversion consulting firms, or management consulting firms — price by the hour or by the deliverable?

That was the question I started to discuss with Frank Demmler, who teaches in the Tepper Business School at CMU and is portfolio manager of Innovation Works. But we were cut off by time, and I never really got to tell him my problem (which is that my company mostly does fixed price work, and then works twice as hard as we should even if the customer doesn’t ask for it, just so that we delight the customer. Another one of my CMU-professor-friends, Economist Marty Gaynor, told me that this is a variation on “winner’s curse.” But I digress.)

I saw Frank yesterday again briefly, while on break from the Google Analytics Implementation seminar I taught across the hall, and then he sent me this advice on fixed price vs hourly:

Regarding pricing, as we discussed briefly, it’s a classic damned if you do, damned if you don’t conundrum. The granularity and visibility of hours is providing ammunition for an almost inevitable battle. A fixed price contract is one in which the client will rarely be satisfied and will claim that the contract included things that it didn’t.

One overriding comment is that you need to be disciplined and fairly aggressive to train your clients in either case. If it’s hourly, don’t allow yourself to get bogged down in the details. If there’s a substantive disagreement about project scope and resultants costs, that’s worth discussing. Otherwise, stand your ground and claim that you are a professional providing professional services and that’s what they cost. If they want to terminate the contract, so be it. Similarly, on flat fee project work, I’ve never seen a project that was correctly spec’ed out at the beginning. You’ll know when things are going off track. Rule of thumb is at the midpoint of the project, you should have a meeting to discuss progress and where changes have occurred that require an amendment to the existing contract and the fees for doing so.

I am hoping to talk to a lot of my friends this coming week about how to run a consulting company (but remember: not about prices themselves. That’s collusion.) So I wanted everyone to have the benefit of Frank’s thoughts.

Also, if you are a WAA member, you should read this article I wrote a week ago on the WAA site about the economics of running a WA consulting firm. I keep trying to pull the analytics for that page (does anybody read it beyond the eight people who starred it and the three who left comments?) but they make my computer crash. Every time.

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.