Archive for the ‘Analytics’ category

Pretty, New Competitor To Google Analytics

April 5, 2008

For those of you who are a little askance with the idea of handing over your information to the Google Borg, there is a new analytics package out there you might like: Clicky.  Having used it for a day or two, it doesn’t offer all that much unique from Analytics which is useful to a uISV, but it has some usability wins (no need to tag URLs on your website, automatically tracks downloads and inbound/outbound links, etc), and there are features to stalk particular folks across your site if you’re into that sort of thing.  Personally I wouldn’t suggest it but if you had a webapp with a privacy policy which allowed it it might be a useful support tool.

One thing folks sometime neglect when making a webapp, and I’m certainly guilty, is making it look gorgeous.  Clicky does not have that problem.  I’ll be honest: I signed up precisely because it had the “new car smell” to it.

Using Google Website Optimizer Safely

September 7, 2007

In my recent post about Paypal’s new icons I mentioned that I was going to use CrazyEgg to check whether the icons were more loved or not than the old ones, and I am, but I decided to go the extra mile and use Google’s Website Optimizer to do a split test.  A split test is when you randomly send half of all prospects to one version of a page and half to another to determine, in a rigorously scientific and statistical manner, which of two alternatives is better.  They’re typically a pain in the hindquarters to accomplish, but this one wasn’t so bad, thanks mainly to a new feature that Website Optimizer includes.

Previously, you had to markup the bejeezus out of your web pages to Optimize them, which harms some user experience (if the bracketed portion of the page loaded slowly, congrats, you lose) and took far too long.  Now, while thats still an option for multivariate tests, Google has a simpler option — make two pages, put two bits of Javascript in the first (front and end of the file) and one in the second, and then put a tracking code on your conversion page, and Google takes care of the rest via transparently redirecting folks who hit the first page into the second with 50% probability.

They recommend that you leave the second page up indefinitely, because folks could conceivably link or bookmark it.  I first thought that was good advice.  Then I realized DANGER WILL ROBINSON having two pages on your site with 95% similar HTML is an excellent way to get smacked down by the duplicate content penalty, and that would hurt me oh-so-much more than getting a modest bump in conversions from rigorous testing.

Happily, there was a simple one line fix to my robots.txt file that I could make to ward off any possibility of that:

#Somewhere above here we have the “User-agent: *” or “User-agent: Googlebot” line
Disallow: /name-of-my-file.htm

After I’m done with the test, I’m going to use .htaccess to 302 redirect the alternate page to the main page rather than leaving the alternate up forever, which will keep any links or bookmarks good without forcing me to keep an outdated page around and consistent with the rest of my site for eternity.  Why do more work?

Anyhow, if you want to see the site, take a gander at my purchasing page which has a fifty-fifty shot of actually showing you the old purchasing page.

Google Analytics Redesign — More Web2.0, Less Useful

May 10, 2007

I have just started using the new Google Analytics interface and, ouch, its a doozy.  Information which was previously available through drilling down is apparently being forced into a new widgetized reporting format, which might be great but seems like it has a learning curve steeper than Mt. Everest.  In the meanwhile, getting to the information which is of paramount interest to me, like “Am I doing better this month than last month on Google, in terms of conversion rates?” seems to be much more difficult than it used to be.  Hopefully in a few more weeks I’ll have mastered this interface and can get back to being productive again…

Insomnia Leads To Download Page Redesign

May 9, 2007

Tell me what you think of it.  I’ve disabled automatic downloading, given the users big bright buttons to find the PC vs Mac trials, and hooked it up to Analytics to see if I’m insane.

Actually, since its almost 5 AM now, I think I am certifiably insane.  Going to grab a few more hours (I slept from 9 to midnight, which is what is killing me).

CrazyEgg vs. Google Analytics

April 20, 2007

(All images in this post are cut off by WordPress.  Click to see the full versions.  I suggest opening them in a new window.)

I have been using CrazyEgg for the past week or so, trying to make some usability changes to increase the conversion of my website.  I was skeptical that it was going to provide better information than Google Analytics, because Google Analytics also has a site overlay feature, and because I had been using Analytics for months and presumably was getting all or nearly all I could out of it.  The test has showed me several places where my website was broken where I thought it was working perfectly, and as a result I’m going to dig out my credit card and sign up for CrazyEgg as soon as I get done writing this post.

Here’s a comparison of two views of my website over the same interval.  The first is Google Analytics, the second is CrazyEgg.  Focus your attention, like I do, on how well this page drives people to the free trial.

Front Page As Seen By Analytics

Here is what I see when I look at this image: the website appears to be functioning well.  The most popular link on the page is the blue Download Free Trial button, which is exactly what I want.  Screenshots also appear to be pretty popular.  One concern is that folks seem to be banging on the Information tab a lot, which will take them directly to this page, so its clear they don’t quite understand the highlighted tab metaphor (not suprising given my audience).  The three links in the first paragraph of text are performing moderately well and primarily directing people to the trial, which is exactly what I want (screenshots and free information are nice to the extent people download as a result of them).

It turns out that these conclusions are faulty.  Lets see this page again, in a CrazyEgg heatmap.

Purchasing Page in CrazyEgg

What do we learn here?  Well, first, no one is banging on Information — thats just a quirk of the Urchin (Analytics) script on some browsers when people double-click to open a link instead of single click.  The download a free trial link in the first paragraph went totally ignored (not one click out of two thousand visitors!) while the screenshots and free bingo cards got much more significant attention than you would have thought from looking at Analytics.  The bright blue Download a Free Trial button performs admirably.  The screenshot in the middle performs extraordinarily well — almost 33% of visitors to the page will at least click it to see what it looks like!  This was catastrophically bad news for me when I heard it, since I know that clicking a screenshot is the quickest way to bounce a prospect, since so many of my visitors have limited web-navigation-fu. 

So, here’s what using CrazyEgg to enhance my borked understanding of the Analytics numbers let me do:

a)  I installed Lightbox, and watched my bounce rate drop considerably.  Not only does it look stylish, the “click anywhere to go back to what you were doing” mechanism works very well for technically unsophisticated customers.  (More evidence of this: take a look at how many people are banging on the New button in my screenshot.  Yeah.  Believe it or not, I’ve gotten emails about that before — “The New button doesn’t work.”.)

b)  I am going to redo the first paragraph of text to deemphasize “download” and emphasize that you can “create free bingo cards” using the trial download.  I may end up burying the free bingo card link in the free resources section, since its far too effective at siphoning people off where it is right now.

c)  You can’t see it from these two photos, but I also did tests of my purchasing page and download page.  Whoa doctor, are there some easy and obvious things to change.

Incidentally, if you’re wondering “Why does Crazyegg report different data than Analytics?”, the answer is that CrazyEgg tracks clicks on a per-coordinate basis and Analytics tracks them on a per-URL basis.  To override that, you have to go to considerable work when coding your webpage.  This means that Analytics treats that blue button and the first paragraph “download the free trial” link as the same place, so it munges their stats together when displaying them, which makes it not-obvious that the Big Blue Button is an awesome success and that the first paragraph text link is a crushing failure.  I had previously gone to a heck of a lot of work when I redid my purchasing page to make sure that every link was hooked up correctly, and I ended up having so much information flowing at me in the statistical summary I could not make heads or tails out of it.  The heatmap, on the other hand, tells the story in a really efficient manner: “Customers want to ‘Pay with their credit card’ and need to be guided on to what to click on to make that happen”.

CrazyEgg Heatmap of Purchasing Page

Given that CrazyEgg is dirt cheap ($9 a month), I’d be … crazy… if I didn’t sign up for it.  All I would need to do is decrease my bounce rates anywhere in the funnel by about 1-5% and it would pay for itself.  I think I’ve already accomplished that several times over.  I’m obviously not going to stop using Analytics, because I do need the big heaping helping of stats (especially referrers and search queries) and the view on how people move through the pages as opposed to what they do at each individual page, but CrazyEgg provides an easily understandable visualization of the things I need to focus on — what my customers are focusing on, naturally.

(Edit: it took some work to make the pictures fit.)

Using Analytics To Improve Your Web Design

November 16, 2006

So if you’re like me, you’ve been obsessively tracking folks through your website using Google Analytics, using the information on what the visitor paths are (home page -> trial explanation -> trial download -> purchase), what the high value paths are (anybody who clicks to read my license terms has a 50-50 shot of buying from me), and whatnot.  If you identify problems doing this (such as “Hmm, nobody who clicks on a screenshot ever comes back”), you’ve fixed them.  But you probably aren’t tracking people leaving your site, for example to go to an offsite payment processor, like Paypal.

Well, you should, and its really easy to accomplish.

You need to manually edit all of your links offsite to include the text


where local path example is a non-existent web page on your site.  Google will report someone who clicks on that link as visiting the page /local/path/example, just as if they had visited a page with that name on your site.  I use /paypal/purchasing.htm/top-corner-button, for example, which tells me that somebody clicked on a link to Paypal from purchashing.htm on the top corner button.  This lets me see what part of purchasing.htm is really motivating people — in my case, its NOT the top-corner button (who knew!  I always expected folks would go for the easiest button to reach). but rather the part later on in the text where I describe Paypal as a safe, secure place to shop online.  I guess I understand why Paypal trumpets that so much in their marketing now!  My takeaway lesson from this is that my customers are a bit hesitant to give over their credit card number to somebody they’ve never met before, which fits my mental profile of them, and that I could probably increase conversions by stressing how safe and secure it is to buy through Paypal earlier on the page.  I’m going to do that and see how it pans out.

Anyhow, tagging your outbound links only takes a few seconds per link, and you can learn valuable stuff about your customers’ behavior.  I recommend that everyone does it, most especially to links that are in your conversion pathway.

Peeking Over Your Customers’ Shoulders

August 5, 2006

I’m going to assume there are multiple possible paths to get your your Buy Now button in your trial software.  Pop quiz: which one gets used most often?

Oh, you weren’t tracking that?  Well, it seems like its fairly useful information to you — after all, from your perspective thats the most important mouse click in your program there is.  And if you know that information, you can test the placement of the link — does a nag screen produce appreciable results?  How about an exit screen?  Do people click the button more when the menu item is highlighted in blue (*cough* yes *cough*)?  Is your idea to put a tip-of-the-day to put another opportunity to buy a sound one (nice idea, Peter Muys)?  If your product is feature limited, what feature is the one that people can’t stand not having?
Yeah yeah yeah, nice to know but its excessively hard to implement, right?  You’re probably thinking you need to have some sort of shopping cart application and pass it a parameter and parse the output and and and…   No.  We’re computer programmers because we are too lazy to do things like that when there are 10 second solutions.

Step #1: Append a dummy parameter to your static HTML page which you send folks too.  e.g. Instead of accessing access  (Did you know that your web server will happily take any number of parameters to a request for a static HTML file and promptly log and ignore them?  Yeah, who knew.)

Step #2:  Using your favorite Analytics software, take a gander at the “dynamic content” for your file of choice.  Under Google Analytics this is under Content Optimization -> Content Performance -> Dynamic Content.  For example, I can see instantly that roughly half of the hits on my purchasing.htm page come from referals from my application.  My application has 3 paths to get to the purchasing screen (nag screens on two disabled features and the Purchase Now button), and if I wanted I could segment on those.  Something to try for v1.04.

P.S. Keep in mind its generally poor web design to include lots of dynamic parameters, because they confuse search engine spiders.  But since GoogleBot isn’t spidering your application, you can go hog wild with referrals from there and it will never be the wiser.