Who Are You and What Are You Doing On My Site?–Web Analytics

Web Analytics Panel: (L-R) Melissa Blaney, David Smith (Moderator), Mike Sweet, Mark Johns, Jake Zarnegar

When I saw the title of this panel, I wondered if it should be called “The Big Brother Panel”.  Web analytics provide lots of data on who is visiting your site, and in turn that allows you to develop strategies for understanding your business.

Melissa Blaney, Manager of Platform Analytics and Communications at the American Chemical Society (ACS), led off with a description of some of the common tools that the ACS.  These include Atypon’s Literatum, which provides COUNTER reports, identity and content reports, and advertising statistics; Google Analytics, which is used to track usage of ACS online journals and community sites, and Omniture SiteCatalyst, which is used for tracking accesses to ACS’s weekly news magazine, Chemical and Engineering News.

Analytics are used by many internal stakeholders in an organization:  Executives, advertising, sales, web strategy and innovation, marketing, editorial , sales analysis and support.  The ACS has been providing COUNTER reports to stakeholders since 2002, and among other uses, one of the factors in determining renewal prices is online usage.  Platform enhancements and web development initiatives are also influenced by the data.  Beyond COUNTER other statistics can measure key performance indicators (KPIs) for the business, such as referrals, searches, geography, unique registrants, etc.  Useful information can be obtained from data on usage in various time periods, such as seasonality.

Here are some useful data that can be obtained from search analytics:

Data on referrals shows how people get to the site and where they come from.  Although Google may drive 90% of the traffic to site, discovery tool use may be more valuable.  In general, browsing seems to be decreasing, which indicates users are finding information less frequently by serendipity.  Much web traffic is international, so geographic data are important.  And world events can influence traffic; for example, when the Olympics were in China, there was a big decrease in traffic, and when storms keep people at home, traffic from that area decreases.  Tracking social media is often a challenge.  You must know your audience, define expectations, and document and categorize what works.

Using analytic data, future developments can be planned.  For example, when the ACS wanted to develop products for mobile platforms, the following data were used to prioritize which platforms to develop for first.

In conclusion, it is important to recognize that web analytics are only one source of data.  Others include focus groups, user testing, customer surveys, and direct feedback from sales teams.

Mike Sweet, CEO of Credo Reference titled his talk “Web Analytics:  Pragmatics Rule–“It’s the People”.  Many analytics tools give you a lot of data, but what does it mean?  You need analysts to interpret the data and then figure out how to implement the results. If you try to focus on all the data that’s available, you won’t achieve anything.

Credo’s primary platform goal is to increase traffic.  They began by choosing a web analytics package that provided a number of techniques:  on-site web analytics, usability testing, focus groups, market scanning, but soon learned that Google Analytics was the best for their purpose.
Choose wisely:  They chose the wrong package initially but found it not useful, so they switched to Google Analytics.  Focusing on collecting lots of data (especially from log files) and generating reports may obscure what is best for the business.  It is best to define what you are trying to do on the site and if you were able to do it.  From this insight, you can make all the platform improvements you need to make.  Prompting users to give you information on why something isn’t helpful is very simple to do and tells you a lot.  Here are Sweet’s lessons learned:

1. Assess where you are on the evolutionary curve.
2. Choose packages and data mining projects carefully.
3. Don’t plan to rely solely on on-site web analytics data.  Mix on-site and off-site data to get a complete picture.
4. Assess your teams’ agility and your platform’s extensibility.  Only gather insights into things you are actually ready to act on.

Conclusions to ponder

1. Don’t bite off more than you can chew.
2. Numbers aren’t customers.  Use a larger approach to improvements.
3. The experimenter’s mindset is a key–get started and have fun!

Mark Johns, Manager, Publication Management Group at HighWire Press noted that robots are widely used to crawl sites and obtain usage data for them.  “Good” robots are good for business, so having your websites exposed to them is extremely critical.    The “not so good” robots are the overzealous up-time tracker that hits the website frequently and causes problems for it, or malicious ones crawling the web.

The web is becoming more personalized and is molding itself to users in real time.  Publishers cannot get direct access to information users because of the institutional purchasing model.  It is time to start thinking about individual users.  One example of this is the BBC, which lets users rearrange their home page to their liking (except for the ads).  We know a lot about subscribers so we can target things to them.  But we also have data about anonymous users, such as their IP address, search terms, geographical location, language settings, content viewed in a session.  Therefore, that metadata can be used to create a user profile.

Semantic User Profiling
Jake Zarnegar, President, Silverchair

The shift to the institutional subscription has created a widespread problem of anonymous users making up the bulk of a site’s users.  In an era when it is possible to track personal topic interests more closely than ever before, many publsihers currently know less about their customers than ever before.  Two ways to overcome this problem are to require users to give you information about themselves or create a semantic user profile invisibly to the user based on their site intelligence.  This can even be done for anonymous users, but this immediately raises questions about privacy.  Silverchair has developed a statement on privacy that they use with their customers:

Silverchair privacy statement

How to build up profiles:

• Have your content semantically tagged.  Semantics provide a normalized, logical metadata layer on content.
• Accumulation: look at user interactions by accumulating the tags of documents they look at.  Build this over time and look for patterns.
• Construct basic semantic profiles of users.  Parse your raw logs into semantic profiles  The rules for doing this are proprietary and vary from organization to another.  (photo of typical report for an anonymous user)
• Create affinities and put profiles together.  Affinities can be to topical interest gorups, ads, products, events, individuals, etc.  User affinities are constantly updating as the site captures more usage.
• Use affinities to create personalized profiles for users, create a marketing campaign, or promote products when people come on to the site.
• Use the resulting semantic profiles to understand your audience as individual information consumers, tackle the anon user problem, and provide more detailed targeting for marketing and advertising efforts.

Don Hawkins
Columnist, Information Today and Conference Circuit Blog Editor