Tuesday, September 26, 2006

Web Analytics Wednesday - Québec

Une première rencontre de la communauté Analytique Web du Québec aura lieu le 11 octobre 2006, de 18h à 21h au restaurant "Le Cartet",106 rue McGill, à Montréal (Google Map).

L'idée est de démarrer la tradition de rencontres conviviales favorisant les discussions et les échanges au sujet de tout ce qui touche les statistiques Web, l'analyse du comportement des usagers Web et l'analytique Web en général. Professionels, utilisateurs, représentants de solutions ou de services se regroupent dans un endroit (habituellement une section reservé dans un restaurant) afin d'échanger et d'établir de nouveaux liens.

L'évènement est appuyé par la Web Analytics Association (http://webanalyticsassociation.org/)

RSVP en contactant Stéphane Hamel.

Wednesday, September 20, 2006

Web Analytics implementation challenge

The problem presented here should be generic enough for any solution provider. I will try to summarize the answers and provide them as a real business case for the Web Analytics Association and the University of British Columbia Award of Achievement in Web Analytics. Of course, I will also provide info on how I will end up implementing this!

Don't hesitate to ask for more info, this will just make the case better!

Here's the case:
While working on a web analytics implementation, we ran into quite an interesting challenge.

What would be your recommendation regarding this scenario if we want to track visitors across 3 (in fact, 4) different domains:
  • SiteA.com (A) is the parent company, SiteB.com (B) and SiteC.com (C) are two different brand sites. SiteD.com (D) is a transactional web site. (B) and (C) share design similarities but are two different brand names. (D) serves the transactions for both (A), (B) and (C) with the appropriate brand facing.
  • SiteA.com (A) refers traffic to SiteB.com (B) and SiteC.com (C). Those later two refers traffic to SiteD.com (D) for transactions.
  • But sometimes, traffic goes straight from (A) to (D)
  • So: A -> (BC) -> D and A -> D
The dilemma is this:
  • We need tro track (B)+(D) or (C)+(D), that is, track the transactions with there respective brand sites. We do it on (D) by checking if the traffic is coming from (B) or (C) and assigning the right web analytics account id.
  • (A) receives about 30 times more pages views a month than (B)+(C)+(D) altogether
  • There is only about 10% of the pages on (A) leading to (B) or (C)
  • (A) is actually conducting an evaluation for implementing a web analytics solution that might, or might not, be the same as (B), (C) and (D).
  • (B) receives about 80% of the traffic referral from (A), the remaining 20% referrals goes to (C)
The questions are:
  • Should we use a distinct account id for (A)? In this case, what are the compromises? For example, is there a way to track unique visitors across domains and/or conduct fall-out/funnels reports across different sites?
  • Should we use (B) site id for site (A)? thus, merge (A)+(B)+(D)? What are the consequences? For example, this would significantly impact the ratio of visits to conversions!
Updates will be posted below:
  • 2006-11-15: The joys of cross-domain tracking, on Lies, Damned Lies...
    Honestly, I think the proposed solution is a bit too complex and error prone. I would rather go for a non-intrusive solution such as a "friendly 3rd party cookie" outlined above instead of playing around with URL rewriting and introducing scripting code throughout the site.

WebAnalyticsQuébec LinkedIn Group Logo

I have requested the creation of a LinkedIn group for easier profesional networking among Web Analytics professionals in Québec.

One of the requirement is to provide a logo for our group and I was wondering what would be cool enough, and significant at the same time. Then I tought of the Web 2.0 Logo Creator by Alex P, which is realy intended as a parody of all the Web 2.0 company names and logos coming out everyday. Then, why not add some colors, which needed obviously to be in the same "Web 2.0'ish" trend: "Web 2.0 Colour Palette".

Finally, I played with different ideas, which should be obvious enough and offer a bit of cool Web 2.0 style:
  • http://msig.info/web2v2/[c=BF3119]WebAnalytics[/c][c=3772A7]Québec[/c].png
    Generated Image
  • http://msig.info/web2v2/[c=BF3119]WA[/c][c=3772A7]Qc[/c].png
    Generated Image
  • http://msig.info/web2v2/[c=BF3119]WA[/c][c=3772A7]Q[/c][c=36393D]![/c].png
    Generated Image
The red comes from the WebAnalyticsAssociation logo while the blue is borrowed from the Québec government web site, which is the "official color" of Québec.

Of course, if we decide to use such a logo, the "Beta" note will be removed! :)

Your logo sugestions are welcome!


I have found two other services to generate cool logos!
  • LogoMaker.com offers advanced editing and easy logo design. Check the logo I have created in a few minutes:
  • Cooltext.com is simpler, but the Pixel Badge style is cool:

Thursday, September 14, 2006

Premier WAW à Québec!

Bien que modeste et limitée, nous pouvons estimer qu'une première rencontre du type "Web Analytics Wednesday" a eu lieu hier avec un magnifique souper au Château Frontenac de Québec. Nous avons eu l'occasion de rencontre M.Scott Wells de Omniture, ainsi que Simon Rivard de Wiser Interactive, qui offre des services professionels en analytique Web. S'était joint à notre petit groupe madame Deschènes, conseillère en recherche marketing chez Desjardins Groupe d'Assurances Générales.

La prochaine rencontre aura lieu le 11 octobre, 18h, à Montréal (lieu exact à déterminer).

Tuesday, September 12, 2006

Web Analytics Solution Profiler - WASP!


I was thinking about various ways of using web analytics and/or commonly faced problems when implementing or using web analytics solutions. At least two ideas emerged:
  1. Checking site tagging integrity is a common problem to all vendors, consultants and analysts.
  2. Providing realistic information about 3rd party vs 1st party acceptance is often raised, but actual studies are very limited (one of the best reference found is from Benjamin Edelman, "Cookies Detected by Anti-Spyware Programs: The Current Status", if you find others, let me know!)


With the WASP (Web Analytics Solution Profiler) project, I will address the first concern. After some investigation, I found those solutions:
  • WAVcheck, from Webbanalys, extends on the same idea with an executable version which can detect up to 27 different vendors.
  • Rahul Revo posted on his blog a request for a Greasemonkey extension that would detect Google Analytics. Mohnsish Rao proposed a simple solution.
  • Mike Keyes, on his blog "On the trail", created a simple bookmarklet (a one-liner JavaScript you can put in your favorites) that will detect a bunch of different vendors. Cool, simple and is sufficient in most cases.
  • Some vendors usually provides their own debugging aids, usually in the form of a somewhat limited bookmarklet that will display the parameters being passed to their data warehouse.
  • Some have proposed developing Unix-based scripts with grep and wget or perl but it looks to me like a pretty complex endeavors that have its own limitations.
  • Or you could get help from your vendor or ask for a independent consultants to help you out. Maxamine is one of them.
  • Other "complementary" solutions that might help: Watchfire WebQA is particularly good at crawling a site and looking for specific code.
  • AlertSite recently posted a Firefox extension named DejaClick. Not only is this extension wonderful as a very powerful macro recorder, but it is also one of the best packaged extension I have seen so far.

Proof of concept

I created my own little proof of concept using Firefox and the Greasemonkey extension. Basically, while you browse, it will show which vendor is being used. The next is to display which parameters are being set. I added more intelligent detection to highlight default values and those that are overwritten by the tagging.

The solution

WASP combines all the above ideas into something much better and powerful:
  • Come as an handy Firefox Extension.
  • Detect a large list of web analytics and ad network solutions (over 50) while you browse.
  • Detects the exact handshake between your browser and the vendor's data warehouse by providing multiple levels of details:
    • HTTP Headers: helps understand if the query is getting trough
    • QueryString: shows exactly what is being sent
    • Cookies: indicates their values and the way they are being served (1st party vs 3rd party)
  • Analyze the tagging and offer best practices advises:
    • highlight tags default values and overwritten ones
    • common mistakes includes using the wrong tag variables
    • missing page tags
    • JavaScript errors
    • Using old versions of the vendor scripts
    • Mixing pages with 1st and 3rd party cookies
    • etc...
  • Crawl a whole site to detect missing/incorrect tags.
Your comments and suggestions are welcome!

Tuesday, September 5, 2006

Five “Ecosystem” Challenges for Web Analytics Vendors

Another excellent article from our friend Avinash Kaushik, "Five Ecosystem Challenges for Web Analytics Vendors".
  1. Privacy: basically, 3rd party cookie is dead, 1st party is fine, but more work toward privacy is required. Can't agree more!
  1. Demand for multiple solution options: hosted vs in-house, the never ending debate. Here I would tend to disagree with Avinash. Bringing web analytics hosting in-house might seem an interesting venue for cost saving, but budgetary expenses being what they are, it's not obvious to compare a capital investment (hardware and software) and ongoing service-based yearly expenses. Furthermore, there is a strong push toward outsourcing whatever is perceived as a commodity and use it as a service, the Microsoft Live and Google Apps are two indications of a strong move toward this direction. In the case of web analytics, hosting and data crunching is the commodity, timely presentation of quality results is the real value. If web analytics vendors starts Applying Web 2.0 concepts, we might see interesting possibilities emerge (want your dashboard in Google Personalized Home? build your own interface using YUI (Yahoo! UI Library), mash up your data with other Web 2.0 services? Fine!
  1. Web Analytics is not enough: If there is an opportunity for web analytics vendors to extend their offering, I think it will be by offering new complimentary services. We have already seen most of the web analytics vendors offering at least some degree of SEO and PPC tools. I think we need to look for other services such as polls, surveys, competitive /market analysis, etc. Ease of merging with internal business data will also become crucial, when we say "web analytics is not enough", I would push for seamless and easier "analytics" in general. The platform for web analytics could become a data repository for the analysis and visualization of any type of data, not just web related.
  1. Reliance on page centricity: this item is getting a lot of attention. Rich Flash interfaces and even more, Web 2.0 types of applications require a different unit of measure. The WAA definition of a "page view" is of interest ("an analyst-definable unit of content requested by a visitor"), but to me, what seem even more interesting is the concept of a "Meaningful Interactive Event".
  1. The 10/90 rule: Again, I agree with Avinash. The tool is part of the equation, but it's not and end in itself. At the end of the day, you still need a human being able to understand, analyze and do recommandation on actionable data. In some cases, the lower-entry level solutions are what makes the most sense, but I think there is plenty of room for the "bigger" players, and I hope it's more than 5%!
Great post Avinash, it's always a pleasure to read you!