Web Intelligence

As I said in an earlier comment, I didn’t get to see many of the sessions at eMetrics DC due to a raft of WAA stuff and great interactions with the people at the show outside the sessions.  But I have seen a lot of commentary, notably from Gary, Judah, and Eric, and related, from Christopher, on the overall message.

I have to say I agree (or is it have agreed?) – web analytics is headed for the BI shop.  In what form, we can only speculate.  But I have a few ideas, and a great resource that could be quite helpful depending on where you want to go with your analytical career.

The Google Analytics API, for one thing, is going to be huge from a BI perspective.  Just exactly what you have access to and in what format will be an issue for some BI folks, who tend to want “all of it”.

If BI really wants all the data, WebTrends was talking about cleaving the reporting from the processing – just like a traditional BI scenario, where the analytics app sits on top of any warehouse.  But I think in general most BI folks are over-thinking this issue and in time, they are going to be more satisfied with the “right” data, as opposed to “all”.

Here is what I mean.  I think you are going to see this split, where the higher level web analysts are going to head into BI, because they already know what they know, and what data is important.  BI will suck the “right” data out of the web anlaytics area and leave the rest of it.  Meanwhile, there will still be web analysts that will use “all the data”, but that will be pretty much relegated to what a lot of folks do now – straight traffic reporting, basic analysis, optimizing the web site at the campaign and page level, all that good stuff.  Important work, but for many business models, not mission critical.

Said another way, we’re going to get a split along Visitor versus Customer.  Web analytics will still be dealing with the transactional, backword-looking web stuff at the Tactical / Visitor level, and BI will handle data integration and all the forward looking stuff – Predictive analysis.  This means BI will be dealing with the C-Level and Strategic issues for the web.

Now, it certainly doesn’t have to turn out like this, and many shops using the advanced or Enterprise level web analytics platforms that run on a real data warehouse no doubt are already there – BI has come to the web analytics shop.  I suspect a lot of the pure online commerce folks are in this bucket.

I’m sure many BI folks think of web analytics apps as “bridge technologies” that will have limited use once BI gets their hands on the raw web data.  I doubt this is really true in most cases; there will still be a need for “reporting” at the site level for optmization, and the BI folks I think dramtically underestimate the overhead of dealing with this kind of reporting in a data warehouse. 

So, there is going to be a split, with commodity Visitor Reporting down the web analytics path and high value Customer Analysis / Insight moving into BI.  This means a potential player split as well.

Some players, on both the Tech and Marketing side, will stay with the commodity Visitor reporting side of the business.  Others on both sides will choose to move up into the BI world.

On the Tech side, a move up into BI means data warehousing, ETL, and that whole area, where a web analyst would be extremely valuable in terms of understanding and configuring the combined solution.  Pure analysts will start to grapple with (aggregated or event) web data in SPSS / SAS or similar apps.

On the Marketing side, it’s a move away from the Tactics of Advertising to the Strategy of Marketing, where you’re not just looking at the media but also customer service, product configuration / packaging, pricing, fulfillment – all the customer touchpoints.  This movement is well underway at companies that were born customer centric: catalogs, TV shopping, other Relationship Marketers.

To me, one of the great paradoxes of online is everybody talks about customer centricity but few actually measure anything at the customer level;  the non-stop blah blah blah is all about Visitors and Campaigns instead of Customers.

I think this paradox will start resolving itself during this next business cycle, since there is never more focus on customer analysis than there is in an economic downturn.  As I said in my presentation at eMetrics, it’s fairly easy to get into the prediction game with some simple analysis that is really Strategic in nature. 

Imagine the power of predicting your business will be doing better or worse 3 months from now.  What are we doing right or wrong today that will help or hurt us in 3 months

That’s what the BI / Customer angle brings to the party, for alas, web analytics / marketing has for the most part stubbornly refused to get into the Customer measurement game.  The general model and approach I presented at eMetrics, which I referred to as the Past / Last model (Frequency / Recency) for that audience – is on this blog, for Campaigns, Visitors, and Customers –  if you are interested.

So, what to do about all this if you are a web analyst?

If you see yourself going the tech / warehouse route, and especially if you already have the WAA / UBC Award of Achievement in Web Analytics, I would consider going for the Certificate in Web Intelligence from UCI.  I heard a fabulous story at eMetrics from Shaina Boone at Critical Mass about some very intense warehouse stuff they are doing, and it’s working out quite well for the clients.  She’s in or completing the UCI Certificate program right now, and said this UCI material was a tremendous help with that task.

On the Marketing / Business side, are you one of those folks who can consistently pick the winners before the A/B testing even starts?  Or have you provided a solution that outperformed all of the machine testing just by looking at the page? 

If so, you have developed deep insight and empathy for the customer, are a truly customer-centric thinker.  This kind of person can excel in execution on the BI side, Turning the Data into Profits.  Not by creating Campaigns, but designing entire Marketing Programs that may involve other areas like Customer Service. 

If you’re looking for education in this area, check out some of the courses offered by the Direct Marketing Association or read one of the great books out there on this topic.  The WAA is planning on expanding the Certificate program to cover these areas as well.

In the end, the difference between Visitor and Customer analysis is a matter of perspective.  Clearly a Visitor and a Customer can be the same entity; the difference is how you look at the data.

Visitor analysis tends to be very “one-off” transactional and is backward looking; Customer analysis tends to be about behavior over time and is forward looking / Predictive

For this reason, Visitor analysis tends to optimize towards performance “now” while Customer analysis optimizes for the future and is willing to take a hit in the present for higher ROI down the road. 

Finally, Visitor analysis is very Marketing-centric while Customer analysis is often cross-functional, often dealing  with Content, Customer Service and Product issues as well as Marketing.

So, where are you headed?  Any thoughts on the above?


7 thoughts on “Web Intelligence

  1. Hi Jim

    This is the Uber-Post for many of us, those tired of analyzing “traffic”. It will be quite fascinating to see where the WA field is going. I, for one, have started to eat the BI pie, and find it pretty tasty!

  2. I think what many people – especially those who don’t go to eMetrics – are missing is the Predictive power of the web. There is so much value yet to be unlocked, and for many companies, this value can be much greater than they currently receive from the web.

    The classic example of this at DC was the NY Times using web activity to PREDICT how many copies of the newspaper to print. Print too many, there is waste cost. Print too few, there is opportunity cost, both current (revenue) and future (customer dissonance). Get it right day after day, and you’re talking about some real money.

    That’s BI in the web analytics shop.

  3. Thanks for sharing your thoughts, Jim. In my simplistic view of the marketing world, there are three components: 1) what goes IN; 2) what goes ON; and 3) what HAPPENED.

    The first is campaign planning and design (including testing), the second is campaign execution, and the third is campaign (and more broadly, marketing) reporting.

    What I hear you saying is that web analytics is moving more into the mainstream of #3 — marketing reporting.

    To me, marketers are missing a huge opportunity to incorporate web analytics into #1 — namely using web behavior data to drive propensity and cadence models.

    Why does this void persist? Your comment contrasting “high level web analysts” with others gets at one aspect of the problem — skills. But on the other side of the coin, in my experience, I continue to see many firms whose database marketing teams look down on web data. I think they simply don’t understand how to incorporate the data (and with possibility that the web analysts in those firms aren’t “high level”, then the gap doesn’t get filled.

    Thanks for letting me have my $.02

  4. Ron, I absolutely agree with the idea that web data could be used to drive all kinds of offline modeling. If it’s not happening, seems to me more of a CMO problem than an analytical one. Of course, if the CMO isn’t particularly data-centric to begin with, well, that’s a whole other problem…

    One of the reasons offline folks tend to look down on web data may simply be technical; if they don’t have access to or can’t get the data into their warehouses they can’t trust it.

    Another reason had to do with the perceived “quality” of web data. If you look at web data at a granular level, it is indeed quite “dirty” when compared with offline data. This is because offline data is typically made up of mostly significant events, whereas online data often contains a lot of insignificant events that have to be filtered out.

    Some of these events are only insignificant if you consider the offline equivalent of “viewed envelope”, “opened letter”, “read letter”, “grabbed pen”, “started filling out BRC”, “abandoned BRC”, “threw out mail package” as insignificant. Most offline analysts probably don’t yet understand how powerful access to this kind of data can be, and how online copy / offer / script testing can drive offline direct mail success.

    The eMetrics Summit is chock full of presentations demonstrating the use of online data to predict and model offline events. Some of the more striking examples I can remember are Ford using configurator data to forecast the demand for auto colors and options and manufacturing to that data, Scandanavian Airlines using onsite search data for load balancing and scheduling particular destinations, and the NY Times using visit patterns to forecast how many papers to print.

    These examples may lean more towards operations, but if the online-offline behavioral link is strong enough to be used in operations when “real money” is at stake, these same links could clearly be used for Marketing purposes – if the CMO wants to use the data, that is.

    Pretty soon, there will be no excuse as the WA and BI platforms merge.

  5. >> the non-stop blah blah blah is all about Visitors and Campaigns instead of Customers

    Agreed. Having enjoyed explosive growth, online marketers have remained focused on acquisition type analytics mostly. I am not even sure that “visitors” have taken place among the top two slots for attention. It seems to me that we have talked more about 1. campaigns and 2. content, and then visitors were only third place.

    Another obstacle/misconception that I have heard from direct marketers about web data is that the – quantity – of it causes sheer performance problems. It would definitely be advised to only look at meaningful business events in order to reduce the data volume. But what is meaningful and what is not depends on the question. Therefore, I think the best will be to have a web data mart at the ready where one can grab just the meaningful subset of data for each question. But I admit I am biased towards that kind of solution.

    Thank you for the post.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.