How to maximize profits by becoming an analysis ninja

Posted by Mark Nunney
Learn SEO
Use data analysis to inform decisions about your website

Web analytics guru, Avinash Kaushik, can teach you how to use data analysis to change the way you make decisions about your website. Ahead of Kaushik's keynote speeches at the Search Engine Strategies (SES) conferences in London and New York, Mark Nunney looks at his vision for Web Analytics 2.0.

Do your website analytics reports …

… fit on one A4 page?

… focus on response for multiple goals?

… include test results?

… include qualitative data like open questions in visitor surveys?

… include data from Twitter, facebook, YouTube and Feedburner?

… include intelligence about your competition?

… include analysis, insight and related actions?

If you’re answering ‘no’ to the above questions, you’re probably still in the world of Web Analytics 1.0, where analytics reports containing realms of unread data. This is a place where ‘clickstream’ data from software like Google Analytics, Yahoo Web Analytics and Omniture offer little insight.

Avinash Kaushik has a vision of Web Analytics 2.0, outlined in his book of the same name. Here is a summary of the six elements of that brave new world …

Clickstream data

Web Analytics 2.0 starts where 1.0 ends, with ‘clickstream’ reports showing visits, visitors, bounce rate, visitor sources and response. Kaushik recommends that you segment your reports, so:

  • segment visits into direct visits (those who know the site) and visits from referring websites and search engines.
  • segment search engine visits into those from paid campaigns and organic (unpaid) results.
  • segment unpaid visits into those using brand-related keywords and those that do not.
  • segment unpaid non-brand keywords into those used in different stages of the buying cycle

Also, give context and meaning to your reports by:

  • comparing time periods, eg year on year.
  • comparing segments against site averages.
  • comparing with industry benchmarks and competitive data from Google Analytics benchmark reports.
  • knowing what’s happened in your company and your marketplace. Eg, did an email campaign cause a reported traffic surge? Or perhaps it was a competitor’s bankruptcy?
  • learn who does what in your company so that you know who can take the actions your insights recommend. So, you might test new shopping cart checkout pages because your fall-out rate is rising.
  • showing segmented reports for important metrics like ‘visits’ alongside response metrics like conversion rates. Kaushik's method is simple, but it’s surprisingly rare to read reports that say, for example, which groups of keywords deliver the most response.

Multiple outcomes

Avoid lengthy reports that rarely lead to insight or action. But do report multiple outcomes (goals) - actions that you want your audience to take.

And you should go beyond the small world reported by your own site’s visitor stats. Your users certainly do - they go to facebook, Twitter and YouTube. And your content should go to facebook, Twitter and YouTube too.

This means, on your site, reporting on bounce rate (as a measure of the quality of a segment’s visitors), e-commerce or lead conversion rates.

And it means, off your site, reporting on RSS feed subscribers, citations on Technorati, Twitter mentions and retweets, submissions to Facebook, Digg and Reddit, video plays on YouTube, site visits and response from mobile phones.

Testing

Next comes testing – learning to fail faster. Use A/B tests (one option against another) for big things like template changes on campaign landing pages and major changes to checkout pages. Use Google Website Optimizer for multivariate testing (testing many different options at once), such as which images and text on action buttons deliver the highest response.

Voice of the customer

The voice of the customer is at the center of Kaushik's vision. He uses smart usability tests and qualitative customer surveys.

The “single biggest surveying mistake” is to ask too many questions and ask questions that won't result in actions (why bother?)

And the “three greatest survey questions ever” are:

  1. What is the purpose of your visit to our website today?
  2. Were you able to complete your task?
  3. If you were not able to complete your task today, why not?

Notice that those three questions are all about your site users getting things done – actions or ‘outcomes’, as Kaushik prefers.

Competitive intelligence

If response doubles but market demand has trebled, are you doing well? If average response for e-commerce sites in the US is 2.2% and you get 2.8%, is that good?

To judge your site’s performance and (more importantly) to know what you might achieve, you need to know how competitors in your market niches are doing.

Again you must look outside the closed world of your own site’s metrics which you can do with Google Analytics benchmark reports and Compete.

Insights

Insights should be accompanied by actions (they are academic otherwise). And your web analytics should be focused on insights that recommend actions.

Avinash takes this to an extreme that I love. Focusing on fewer than 10 metrics (six is best), analytics reports should give only recommended actions and expected outcomes. No clickstream numbers.

Of course, the numbers have been used to reach the recommendations but the report doesn’t need them - the report is about action.