Analytics 2.0
Workshop: Marketing Analytics for Web 2.0
Date: Monday, 4/16/2007
Presenter: Brett Crosby, Senior Manager, Google
This is my second set of notes from O’Reilly’s Web 2.0 Expo. These are from a talk by Brett Crosby, Senior Manager with Google Analytics and a former exec at Urchin (Which Google acquired to develop its analytics product.)
This was an introductory talk on how Google Analytics (GA) is used in the Web 2.0 world. It was primarily an overview, explaining how and why GA was developed for Web1.0. It then shows how GA has had to adapt to the non-page-centric Web 2.0 world and how it can be used to track offline marketing efforts.
I also attended a more detailed talk about multivariate analysis, which will be online soon. Full notes on this talk are after the jump:
Building Web 1.0
In the bad old days, we built sites in the dark. Design was primarily heuristic. Where any empirical data was used, it required usability studies with long cycle times. Analytics tools began to change the game considerably, but were extremely expensive (xK$/mo.) Google acquired Urchin for the express purpose of making analytics cheap for anyone to use. Their aim was to improve the design of the web, both for altruistic reasons and for business reasons (searchability of sites, driving usage of AdSense). Hundreds of thousands of sites now use Google Analytics.
Using Analytics
The basic flow that any site wants a user to follow: Find, navigate, convert. Ideally, one tracks all referral sources, including offline ones. Only with this empirical data can you truly calculate your ROI for any marketing effort.
Features Google Analytics has to help you get there:
- Break apart data by campaign, geography, etc.
- See conversions at every point in the pipeline.
- Visual click and conversion indicators overlayed on site.
Most analytics products do similar things. You embed JS code that talks to a data collection server when pages load, and the analytics are generated.
The shift in Web 2.0 sites: Tracking user actions and interactions as opposed to pageviews. The page is no longer the base unit of content. There are major implication that the industry has not yet adapted to. For example, Yahoo’s comScore Media Metrix numbers are falling behind competitors because of fewer user page refreshes. The metrics industry is soon going to need to change how they measure popularity. GA is addressing the issue by allowing JS snippents to ping the analytics server even with Flash content or AJAX interactions.
Google is also helping companies develop analytics around their offline content, including print, radio, and billboards. This work revolves around creating special landing page URLs that can be tracked. Tracking podcasting marketing works the same way.
[Pondering: Isn't it difficult to get unique IDs for content when you rely on human memory? It seems so. These analytics will naturally be much more coarse. The presenter gave an example with Weatherbill.com, a radio ad with an 888 number with a 3-character code embedded (WB4.) To remain memorable, they were able to split just that last number, splitting across ten geographic areas.]
There are other ways of extending online tracking to real-world interactions, as well. Trashy Paris Hilton/Burger examples, e.g.:
- Restaurant locator
- Print a coupon
- Send to a friend
- …etc.
Analytics also allows for constant, interative testing of marketing efforts. Not only can you test your own website, you can also try out messages that will later be used in more expensive traditional media campaigns.
Related: Let users design your site. GA has embedded tools for seeing which combinations of variations are doing best on your site. [Multivariate testing seems useful here, but he isn't getting into it. See this next blog post for more.]
The speaker also pitched a lot of other Google services as ways to enrich the metadata around a website, helping you get to the top of search results:
- Google Base and Google Catalogs – What parts of Instructables could be tied into this API?
- Google Webmaster Tools – These let you track users throughout your site, but the incidental benefit is that Google will crawl your site completely.
- Google Video/YouTube
- Google News – Add images, trackable URLs, and Keyword to your press releases so that they will show up properly in GN.
- Google Maps and Google Earth – Upload content if you have local information.
Major advice: Spend money on the analysis, not the analytics tool. With good, cheap (or free) tools like GA, there’s no need to spend money on software. Use GA and hire someone in-house to work with the data.
Audience Q/A:
- Will multivariate testing register with crawlers as different versions of a page and thus affect PageRank? It won’t. Crawlers only look at the “default.” You should also set up your testing so that individual users (IPs/sessions) only see one version of the site.
- Vulnerability of GA to cookie blocking, etc.? GA uses a first-party cookie, meaning that the cookie will be issued by the domain you’re visiting.
- Performance concerns: GA is fast, but still delays page load. Place the analytics code as low as possible in the page. See yesterdays talk!
- Is Google using analytics data for its own nefarious purposes? Google is NOT using this data, even internally. Their motivations are twofold: first, to improve the web. And, in self-interest, the product drives usage of adwords and the Google Store. [Note: If this is true, then I think it's remarkable. Analyzing and reporting on that data would allow Google to take on ComScore and MediaMetrix pretty aggressively.]
Related Posts:
- Empirical Design Using Analytics (April, 2007)
- Building the Facebook Platform (November, 2007)
- CSS Cleanup (April, 2008)
- Web 2.0 Notes: Optimizing the Frontend (April, 2008)
- “A Group is its Own Worst Enemy” (November, 2008)
