If you’re like me, you’re always worried about your site’s performance. How long does it take for your page to load on the client side? It’s a hard statistic to even gather data on, especially across a large sample of visitors.

Fortunately, Google Analytics recently added Site Speed, their solution for page load time tracking. Their load time tracking only works with IE9 and Google Chrome (browsers that support the HTML5 NavigationTiming interface), and it only samples a small percentage of page loads, but it should typically be enough data to give you great feedback about how long it’s taking your various pages to load.

Learn all about how to set up Site Speed and review Site Speed data on the Google Analytics Documentation site.

For the longest time I thought that Google Analytics was something that you just add to your site and review from time to time. I’ve learned, though, that tweaking Analytics can give some terrific data that wouldn’t be available by using their default setup. This week we implemented some new tweaks that we’ve made to our analytics setup that might be useful for others, so we figured we’d share them.

Custom Variables and Advanced Segments

One of the biggest questions that we weren’t able to answer with Analytics before this week was what the traffic breakdown between members and non-members was. You can look at new visits and assume that they’re non-members, but that’s a pretty big assumption. You can also assume that members only pages are only members, but that still leaves a lot of traffic up to interpretation. To solve the problem of understanding the split between members and non-members we’re employing Custom Variables.

A Custom Variable in Analytics lets you tag either a page view, a session, or a visitor as having a certain attribute. Since we were interested in noting whether or not a user has ever logged in, we went with a visitor variable. Here’s the JavaScript code for setting a custom variable:

1

_gaq.push(['_setCustomVar', slot, name, value, scope]);

Google’s JavaScript can look pretty strange, but you can basically think of the _gaq.push method as pushing commands to Google. It takes an array, and the first element in that array is always the name of the command to send. The following elements in the array are the arguments for that command. The arguments for the _setCustomVar are:

• slot – An integer between 1 and 5. The purpose for the slot number will make more sense when we look at analyzing custom variable data.
• name – A string containing the name for the custom variable
• value – A string containing the value for the custom variable
• scope – An integer for scope for the custom variable.

The scope can have one of the following values:

• 1 – Visitor scope. This variable lasts as long as the visitor’s cookie is present
• 2 – Session scope. This variable lasts as long as the current visit lasts in analytics
• 3 – Page scope. This variable is only in effect for the current pageview

In our case, we used the following code:

1

_gaq.push(['_setCustomVar', 1, 'Member?', 'Yes', 1]);

This code sets a new custom visitor variable named “Member?” with a value of “Yes” into slot 1. In our page we wrapped this code in a conditional so that it only appears in each page if a visitor is logged in.

Initially we also included code that set the variable to “No” if a visitor wasn’t logged in, but realized that if a user logged out the value for the variable would be changed from “Yes” to “No”, so we wouldn’t remember that the user actually was a member. In our case we’d rather remember that a visitor was a member even if they log out, so we got rid of that bit of code that set the variable to “No” when a user wasn’t logged in.

So we’ve got our variable set up. What do we do with it now? That’s where Advanced Segments come into play. To set up an Advanced Segment, log in to Google Analytics and navigate to the profile that you’ve added your custom variable to. In the sidebar, under My Customizations, click on the Advanced Segments link to get to the Manage Advanced Segments page. Once you’re there, click on “Create new custom segment” to build your advanced segment. The tool Google created to design advanced segments is pretty neat – it lets you drag metrics from the list on the left to build the logic for your segment. Here’s what my segment for Members looks like:

Notice how I’m using the “Custom Variable (Key 1)” metric. That maps to the first variable slot that we used when we set the variable. Here’s my segment for non-members, which just checks that slot 1 isn’t set to the “Member?” variable:

Now that we have our segments set up, we can select one or more of them from the segment dropdown on the top right side of almost every page in Analytics to focus on those segments.

Events

One other problem I noticed in Analytics was that we could see when a user comes to one of the pages for our videos, but just visiting the video page doesn’t necessarily tell us if the user watched the video or not. Google Analytics Events allow you to track arbitrary events with JavaScript. Our player includes JavaScript callbacks, so using Events was a perfect fit.

The Google Analytics JavaScript for tracking an event looks like this:

1

_gaq.push(['_trackEvent', category, action, label, value]);

• category – A string representing the category of the event
• action – A string representing the type of interaction that occurred
• label – (Optional) A string representing an additional layer of detail about the action
• value – (Optional) A integer value relating to the action

Here’s the JavaScript code that I added to our player’s callback when playback is started:

1

_gaq.push(['_trackEvent', 'Video', 'Begin', videoName]);

You can review the results of your events in Analytics by going to Content, Event Tracking in the sidebar. The views for Events let you drill down into each event action to see the list of labels and values for that action. In our case that means that we can drill into the “Begin” event to see exactly what videos were watched and how many times they were watched.

Search Analytics

One last feature of our site that I was interested in getting more information about was search. We allow users to search for videos, but we haven’t been tracking those searches at all. Setting up search tracking in Analytics usually doesn’t require any changes to your site’s code. All you have to do is go to the settings for your site’s profile in Analytics, enable Site Search, and enter the query parameters that you use for search terms on your search forms. On our site we always represent a search term with a query parameter named q.

In the Analytics dashboard you can go to Content, Site Search to review your site’s search data. Google includes information on how many searches were performed, what terms were searched for, and what pages users went to as a result of the search.

Finishing Up

Well, now that we’ve tuned our Google Analytics settings we’ve got to wait a bit and then we’ll start making tweaks to our app as a result. I’m hoping that segmenting our members and non-members will tell us more about what topics non-members are interested in and also help us figure out the best ways to get users to our sign up page. The video events and search statistics should give us interesting information about how we’ll we’re helping people find the videos they want to watch. Once we’ve gathered more data and made some changes, we’ll probably circle back and fine tune what data we’re collecting with even more custom variables and events. It really is a never ending cycle.

Do you have any tricks that you’re using with Google Analytics? We’d love to hear about them in the comments!

You can add Annotations in Google Analytics for any major events that might effect conversion rates (or any other metric that’s important to you in Google Analytics).

Annotations are simple notes that you can tie to a specific date. They show up visually on the timeline as you view any graph. Super simple, but very powerful.

Here’s why it’s a great idea: You can look back at your traffic and see how different events affect your conversion rates. Trust me: You’ll forget when and how you changed the site, which means you won’t know what’s positively or negatively affecting your stats. Annotations solve that. Bam! :)

If you haven’t had the opportunity to see Alex at Future of Web Apps before then check out the video below from FOWA Miami, it’s certainly worth 20 minutes of your time.

Alex has worked for some of the biggest online brands and knows a thing or two about creating a successful digital marketing strategy as well as how to use bleeding edge techniques to keep your brand ahead of the competition. He’s also written a number of articles for Think Vitamin on startups, marketing and more.

Early bird passes are available for £295. Visit the “Adventures In” web site for full details.

I’m going to assume that readers are more than familiar with the basics of web analytics, and the point of this piece is not necessarily to educate in this regard but rather to encourage that we do as much as we can to help clarify the situation. Each time I work with a new client I make a specific point of talking them through analytics and explicitly defining the difference between hits, visits (uniques & recurring) and views.

Without fail, each one has mentioned that they had misunderstood the term ‘hits’ and were glad to be informed (although often a little disappointed) that they should ignore it as a representation of a site’s popularity.

A Quick Recap

A a hit is a request received by a web server for a single file. Thus, a page load may result in a number of hits, for example: one for the page itself, another for the stylesheet, another for a JavaScript file and a further hit for each image. See the Wikipedia web analytics article for a more in-depth look at the key definitions.

Why is it important to make the distinction between a hit and a page view? The problem is that the figure represented in the monthly hit count for a site really has no bearing on the number of page views and visits. Of course, there’s a linear relationship between page views and hits, assuming that each page has a similar number of hits per view (which often shouldn’t be the case if client-side caching is on and has been correctly configured).

Counting Hits

If a page has for example 20 external resources to be loaded, one view will of course generate 20 hits. Another page may have only 5 resources and so 5 hits are generated per view. This clearly indicates a huge discrepancy if a sole reliance on hits is used for judging the popularity of the page. The figure could naturally be far more extreme.

Help Your Clients Understand

So I’d like to encourage everyone to, at the very least, insist on making the above distinctions clear with their clients. Hopefully this will help clear up the confusion to some extent.

We replaced the example tutorial video on the homepage with a 50-second overview of the service.

I thought showing people the product (the example video) right away was the right thing to do. Boy was I wrong.

The takeaway here is that potential customers want a quick summary of the product, not an actual demo. The bigger takeaway here though is this: you have to be testing your site to see what works and what doesn’t. Going on a gut feeling won’t work. In fact, this very method that worked better for us may not work for you. You won’t know until you test.

Here are some screenshots from Optimizely showing the results. We had a healthy sample size (1,997) with a very strong statistically confident result of 98.8%.

Now that we know what kind of video works best, we’ll start testing variations of it. For instance, what’s the most effect thumbnail to display on the video?

We use Optimizely for all our A/B testing and would highly recommend it. We secured a deal with them for you guys that gives you 40% off their Silver Plan for life. We’re not receiving any benefit for recommending them, I just think it’s a great product.

Make sure to also read the previous post where I covered how to create and measure campaigns.

We’ve been obsessing over numbers for Think Vitamin Membership, our new video training service for web designers and developers. We just passed 700 paying monthly members, so it appears we’re doing something right. (Thanks to every single one of you who have signed up – we love you!)

It’s nice to see leaders like 37signals demonstrate that this numbers/analytics driven culture is important, with their latest job post: We’re Looking for a Business Analyst. Here’s a snippet from the job post:

“You’ll be responsible for reviewing daily sales, upgrades, downgrades, and cancellations. You’ll spot trends, establish and review analytics, monitor conversions, propose, implement, and measure strategies to increase revenues, grow profits, and improve margins. Using data and numbers, you’ll tell us things we didn’t know about our customers.”

Understanding, knowing and acting upon your key six metrics is vital to the success of your web app. If you’re bootstrapped (like us) you won’t be able to afford hiring someone full time to do this in the beginning. In fact, I’d highly recommend that it should be founders who live and breathe these numbers for the first couple years of the startup, then it can be handed over to a dedicated new hire.

From the announcement …

“Additionally, if you use the FeedBurner Socialize service, and your platform uses PubSubHubbub or you ping us when you post, you can for the first time get stats on how much traffic your feed items are receiving from Twitter, as well as feed reading platforms like Google Reader in one place. Again, all within seconds of posting your content. Ping? Pong! Yep. That fast.”

You can log in here or read more about the launch here.