Categorized | the i.t. files

Google Analytics: the race is on! Here’s my wish list . . .

Google today (literally about two hours ago) released the API for Google Analytics, a Web analytics package featured prominently in our book Shoestring Venture, and two upcoming books in the Shoestring Venture series now in development. Without going into details, this means that anyone who cares can now freely develop applications, add-ons, and plug-ins to the basic Google Analytics service. Why should you care? Google Analytics has the potential to light a fire under the feets of developers and, if Mambo, Joomla!, WordPress, iPhone, Facebook, or MySpace are legitimate precedents, it means that Google Analytics with a publicly-available API will in a few short months be just as good as the most full-throated Web analytics software out there — software that would set you back tens of thousands of dollars. All for the low, low price of free (or darn close to it)!

So, since the flood of Google Analytics add-ons and apps starts today (well, maybe in a few weeks), as a veteran Web analytics and statistics geek, let me propose a few projects for the few, the strong, the programmers to put on their to-do list.

One of the great pleasures of my old age was discovering a real talent for math. After having spent much of my adult life as a literature professor and then as an advertising creative director and artist, I wafted into an MBA with not a little trepidation at the math-laden curriculum. Within a few short weeks, I was a statistics deity and I have gone on to take several graduate level courses in stats and forecasting and data mining.

So, although I’m primarily a whatever-side-brained (I think it’s right) creative-artisty-type, I love statistics from top to bottom. And there’s nothing we stats people like better than crunching numbers from every angle possible.

So while Google Analytics is a perfectly fine application for relatively low-level uses (such as a start-up Web site in which you are not trying to maximize traffic or conversions), it doesn’t provide the insights to really supercharge the Web presence of your site or optimize the user experience.

But the really valuable number-crunching that the beefy Web analytics packages require typically require at least a rudimentary understanding of statistics — which should never be a requirement. Ever.

So, in the interest of really adding some meat to the Google Analytics package, you developers should get working on some meaty little add-ons:

  • Regression analysis: without getting into the arcana of multiple regression analysis, suffice it to say that the most valuable information statistics can give is predictive information. And nothing helps more than being able to isolate the most predictive variables. If I stress keyword A over keyword B, will that increase site visits? If I change the navigation, will that increase sales? What aspect of the site best predicts a high-quality site or a conversion? Now, no user should ever have to know how to do an ANOVA, a t-test, or determine the coefficient of determination (I love doing ANOVAs because, well, I’m a total geek). The system should do it for them. The user should be able to do two things: a.) pick out a result (such as length of visit, number of pages visited, a sale, or filling out the contact form) and ask Google Analytics data to pick out the most predictive variables in order or b.) pick out one or more variables (such as keywords, geography, time-of-day, referrer, size of banner ad) and determine how predictive they are relative to a result (such as a sale, length of visit time, etc.). Users should be able to include dummy variables. Right now, Google Analytics data don’t tell you what to do to get your numbers up. A top-of-the line multiple regression add-on would also have a simple Google-like interface (you know, “I’m feeling lucky”) in which the user simply asks the plug-in to tell the user which variable to focus on to improve the site traffic or quality of visits.
  • Blog statistics and analysis: since so many people use blogging as part of their Web presence or marketing, it’s a shame that analytics programs don’t allow you to chart single users, since they are the warp and woof of blogging success. The unique aspect of optimizing blogs is that you have to separate out gold visitors from lead visitors — the folks who come to your blog day in and day out are far more important than the occasional visitors. Tracking individual bloggers (which is how the big guys like Wall Street Journal and Time do it) gives you the real insights into the effectiveness of your blog and how to increase your audience. Now, there are analytics packages specific to blogging needs, but Google Analytics collects all the same data. Bloggers just need to have that data so they can access the behaviors of their individual readers.
  • Clustering: most of the fancy analytics packages offer some form of clustering (such as HitBox), but you have to take your Google Analytics data into a stats application like JMP or SAS to get any kind of useful data clusters. Understanding how users “group together” in terms of shared characteristics and Web visit results would, like regressions, tell site owners how to make the site and its presence more effective. For audience-based Web sites, such as blogs or wikis, clustering will give valuable insights into how audience characteristics translate into audience quality. Again, the interface has to be easy. You simply ask the application to cluster or group users together in terms of certain results or page views. And, again, there should be the option of identifying the most valuable “group” in a cluster without having to bother with more distant groups.
  • Demographics integration. Right now, virtually every analytics package on the planet can tell you “where” a site visitor is. So you look over your analytics, and you find that 3% of your visitors are from Anoka, Minnesota. Another 2% from Puyallup, Washington and another 2% from Tehachapi, California. Big deal. What do you know about people from Anoka other than that they elected Michelle Bachmann to Congress? How much money do they make? What kind of interests do they have? What is an Anokan or Puyalluper or Tehachapian like or do more or less? That’s where you have to marry the data to some outsourced demographic data, like Claritas. Why not just do it in an application? Why not just ask Google Analytics, through a plug-in, to sort all that geographical information into a demographic approximation (average age, family size, religion, income, rental/own). These are, of course, only guesses, but they’re more accurate than nothing. Or, if you don’t like guesses like this, just simply group users into PRIZM categories (I love PRIZM categories — I never make a marketing decision without them). In other words, oh, Google Analytics Apps developers, I don’t just want to know where my users are, I want to know who they are.
  • Forecasting: It’s simple. SugarCRM does it. ACT! does it. Why doesn’t my Web analytics software do it? An analytics database is, after all, one big honking time series data set. So why can’t I get a forecast? Tell me, O Google Analytics, if all remains the same, how many visitors will I have next month? How much in sales? Without having to export all my analytics data into SAS. It’s not hard, man, just a little bit of exponential smoothing here and there so that I can at least make plans. It doesn’t have to be true, it just has to be statistically valid.
  • Attaching a neural network to Google Analytics: Okay, so long as this is a wish list and considering I’ve already mentioned SAS, how about simply attaching a non-linear data modeling tool to Google Analytics, say, a neural network that is continually “on.” The job of the neural network would be to continually sift through all the complex inputs (geography, banner ad size, keywords, referrers) to sets of outputs (sales, contacts, page visits, repeat page visits) and every day make suggestions for improving the performance of the site. Rather than sifting through site analytics data and “guessing” that you should either change your keyword, knock off a few referrers, or change your banner ad size, the neural network simply gives you a set of instructions on a regular basis (increase your AdSense buy, dump your text ad version 1, etc.). Since neural networks are for all practical purposes “black boxes,” there’s no need for the user to interact with the statistical data. The application simply gives them basic data (numbers and trends) and tells them what to do. On the next visit to the application, more basic data (numbers and trends) and another set of instructions. Since the neural network is always “on,” it’s constantly changing its pathway weights relative to changing inputs and outputs. So, if rectangle ads produced the best results in March and skyscrapers outperform them in April, then it tells you. Now, if this neural network is really worth its salt, it would be aggregating other Web data — including stats from other sites — into its weighting algorithms, so that it’s monitoring universal trends as well as local ones. Add to that demographic data as inputs and . . . HOLY SHOOT! What a great program! Okay, I admit it, I spend hours at SAS playing with neural networks, but you have to admit, what I just described above REALLY ROCKS!
    Be Sociable, Share!
  • 2 Responses to “Google Analytics: the race is on! Here’s my wish list . . .”

    1. I recently happened upon a SAS like software called Project -R, featured in a NY Times column. Then main buzz about it was that, it has the power of SAS, was free and has lots of plug ins developed by enterprising users.

      Your post about extending the insights we get from clickstream data from google ( and anywhere else for that matter) is particularly insightful. That is why i think the API released by Google is really a big deal.

    2. kaa says:

      I am a newbie in the data analytics field. And I am as amazed as you, in the fact that no software does these basic analytics modelings. a big opportunity indeed


    Leave a Reply

    Shoestring Book Reviews

    Shoestring Venture Reviews
    Richard Hooker on Jim Blasingame

    Shoestring Fans and Followers



    Business Book: How to Start a Business

    Shoestring Book

    Shoestring Venture in iTunes Store

    Shoestring Venture - Steve Monas & Richard Hooker

    Shoestring Kindle Version # 1 for e-Commerce, # 1 for Small Business, # 1 for Startup 99 cents

    Business Book – Shoestring Venture: The Startup Bible

    Shoestring Book Reviews

    Shoestring Venture Reviews

    Invesp landing page optimization
    Powered By Invesp
    Wikio - Top Blogs - Business