Ask The SEO

Intro to Google Analytics: Part 3 – Making Progress

Now that you have a firm understanding of the data Google Analytics offers and what you should be doing with it, let’s look at methods for analyses and developing your business over a longer period. Citation

Descriptive, Predictive, and Prescriptive Analyses

Google Analytics is primarily a program for descriptive analytics, but it’s important that you understand the different analyses you’ll encounter across GA and other programs.

Descriptive Analytics: Analysis that tells you “what is” and “what was,” taking the mass of data and presenting it in a digestible format

This is very useful in determining your current standing as an organization and building insight. Most of what you see in Google Analytics will be of a purely descriptive nature. This is the most accurate form of analytics by far, as there is no guesswork involved.

Predictive Analytics: Analysis that tells you what to expect; the search estimates you see from Google Adwords, for example

Other Popular Posts

You won’t see much of these in baseline Google Analytics unless you’re integrating with other systems, but you will see this in other Google and SEO products you are likely using alongside it.

Prescriptive Analytics: A form of analytics that takes the input data and tells you what actions you should take to achieve your goals

This need not be extremely complicated; for example, navigation aids such as Google Maps meet this definition, offering you several ideal routes based on hard data. This isn’t something Google Analytics can do without a lot of customization and tweaking at the moment, but it is worth keeping in mind as you dig deeper into the potential of data-driven growth.

Predictive analytics can be useful in theorizing about directions to take your company for growth, but can’t be wholly relied upon. Prescriptive analytics are best applied to extremely well tuned organizations intimately familiar with their own data, and even then are used primarily for fine-tuning details. Thus, we’re left with proper use of descriptive analytics for reliable long-term growth.

In-depth Look at Split Testing

Once you’ve thoroughly identified your goals and where you stand in reaching them, nothing is more helpful to your progress than split testing. If you make adjustments to your site without split testing, you learn nothing; conversions go up, conversions go down, but do you know why they did? It could have been a simple fluctuation of the market neutralizing a bad decision—or losing you money despite a good one.

Split testing allows you to directly compare two or more approaches and see how they change the results. More important, you’ll be able to see how those approaches differ in more subtle ways, by watching their differences in Google Analytics.

Imagine a few changes to graphics and text that result in more sales from visitors entering via one source and fewer sales from visitors entering via another. Without split testing, you may see an uptick in sales, a drop in sales, or no change at all—without understanding any of the nuance involved. This, of course, means split testing is crucial in the development and application of marketing tailored to different segments.

Fortunately, Google Analytics was built with split testing and data-driven experimentation in mind; you will find the tools necessary to conduct accurate, effective split tests within the basic Google Analytics package—no need for external tools. Look for them under the Behaviors section of Google Analytics, labelled “Content Experiments.”

A Word of CautionCitation

This is the tool you’ll use to set up experiments in GA. First, select the goal you’re testing for, then identify the URLs for the original page and each variation. You’ll receive an experiment code you’ll need to add to the original page. Content Experiments can send the code directly by email to whoever handles your website’s coding if necessary. Once this is done, you review the experiment configuration and begin.

Experiment List

The front page of the Experiments interface, once you’ve set up your first experiment, shows all experiments and some brief data on their status, schedule, etc.

Experiment Reports

Clicking an experiment from the Experiment List brings you to a report displaying color-coded stats from each variation on a single chart. You can view specific metrics, or even compare different metrics within a single report.

Tips for effective experiments

To improve your site, you’ll need to set up the right experiments; if your site is fundamentally underperforming, split testing to see how many adjectives you should use in copy isn’t a worthwhile endeavor.

Keep your goals in mind

Build experiments that directly work towards your goals. If you need more traffic, focus on splits that will reduce your bounce rate. If you need better traffic, try splits that will filter out low-value prospects.

Pay attention to unexpected results

If you get a result you didn’t expect, that can, in many ways, be more useful to improving your understanding of your audience and business than if everything worked as expected. Unexpected results can reveal unexpected insights into your customers.

Try something new from time to time

Don’t get caught in a pattern of similar experiments always changing the same few variables. Try different tones, different calls to action, different graphics, different page layouts. Try targeting a different audience than usual.

Remember, different doesn’t always mean better or worse. Savvy marketers will take two variations succeeding with different groups and work to build separate sales funnels for their different market segments.

Experiment WizardCitation
  • Data Distortion:

    If you split-test more than your traffic can support, you’ll get faulty data. It can be useful to look at demographics in a given split compared to your normal demographics for a page, just to see if you’ve narrowed splits too far and distorted your results.

  • Assumptions

    Data is only useful when you trust it. Many organizations falter by making assumptions, testing those assumptions, and skewing their reception of the results with those assumptions. Come up with a hypothesis and test it, then trust the data, unless you have a concrete reason to believe it’s wrong.

  • Lack of communication

    You want your entire organization on the same page regarding goals and information. At the very least, you’ll want to make certain you don’t have different departments working on cross-purposes; if you are using analytics to improve your web marketing for a particular demographic, make sure to build strategies tailored to that demographic.


Ryan Freeman is president of Strider Inc., founder of Florist 2.0, and publisher of Canadian Florist.

theRKF on EmailtheRKF on LinkedintheRKF on Twitter
theRKF
Publisher at Canadian Florist

Ryan Freeman is the Publisher of Canadian Florist, 5th generation florist, and President of Strider Online Marketing. He has been engaged in web design and online marketing since 1994.


Share:
0

Your Cart

%d bloggers like this: