facebook sdk

google tag manager

Monday, February 13, 2006

Research: Competing on Analytics

The fine folks at Babson College have a research report out entitled "Competing on Analytics". You can get a copy via the Harvard Business School, or via the authors' blog.

The report does not focus on web analytics specifically, but it does provide a few good reference points that apply in our industry. They describe 5 stages of "analytical competition" a company falls in, the highest being "analytical competitors". These are the companies who "have embarked upon analytical competition as a primary dimension of strategy". They are clearly using analytics in all areas of their business, and analytics are driven from the top down as part of the culture of managing their business.

There is an interesting section on "predictive modeling and optimization techniques" that discusses the value of analytics when testing ideas and concepts (think multivariant testing). They note that "Capitol One, for example, conducts more than 30,000 experiments a year with different credit card interest rates, incentives, direct mail packaging and other parameters...".

The authors make one argument that isn't completely clear to me. They suggest that "if analytics are to be a company's basis for competition, and if they are to be broadly adopted across the firm, it makes more sense to manage them at an enterprise level." I understand the need to keep costs as low as possible by re-using technology where possible (hardware and software standards for example). That can only be done effectively through an enterprise approach. However, I don't believe analytics must be managed at the enterprise level to make this happen. Perhaps I'm not fully grasping their argument, but I don't think centralized IT organizations can really manage the rapid development and innovation required to help internal organizations succeed.

Any comments on this subject?

Filed in:


  1. Thanks for pointing us to this article.

    We felt that there were quite a few things that had an "ivory tower" smell to them. Much of the commentary was in broad, sweeping terms and didn't recognize the complexities and nuance involved with making analytics work in a organization. I'm guessing these realities are more obvious to analytics practitioners vs. business professors. We put together a response on our blog: http://www.juiceanalytics.com/weblog/?p=97

    I had the same reaction to the point about analytics managed at an enterprise level. While I'm not certain fully decentralized analytics is the answer, there is much value in the agility and proximity to the issues at hand.

  2. Thanks for the article and the comments, on both blogs.

    I agree with what the both of you say: it's not as easy as it sounds, when reading through the HBR article but you have to put things into perspective. At least, on my European side.
    The article remains a high level HBR comment that I treat as a macro vision that gives me an idea to where we might be heading.
    It reminded me a lot of a book I read about a year ago called Strategy Maps (Kaplan & Norton - HBS Press) which sets a framework of thought for this kind of high level metrics based management.

    The thing is, we're not there yet and the question is off course, would we want to actually go that far. I agree with the datawarehouse stance not being flexible enough and taking too much time to correctly set-up. With the new release of WebTrends (due in March), I've been harassing them to know whether they would be integrating some kind of datawarehouse (again) as their choice will impact my strategy with my clients. If not, other flexible products such as BO's Xcelsius might due part of the job for now.

    But I also find, within the field of Web Analytics, for some of my clients, descrepencies in objectives: while one department holds as objective a maximum amount of subscribers, the other needs to sell online and thus wants to lower the barrier to the subscription process. And we are talking about a major Telco company. Thus, a higher level view and the alignment of objectives would be very much welcomed. Until which extent is still not clear to me.

    I also see Web Analytics as, what WebTrends calls the "hub of marketing". When I first heard this in Austria last year during the partner conference, I wasn't convinced but now I view it more as an experimentation field, as described in the January Webcast from WebTrends' Marketing VP, Brent Hieggelke. For me, Web Analytics can play the role of catalysor (do you say that in English? sorry, I'm French speaking) for the other departments, showing ROI and expanding it to other processes or departments (we mainly focus on Marcom for now).

    It was also interesting to see the parallel between the webcast and the article with the statement that CMOs are experincing a "confidence gap". I also experience that they remain in place for a little less than 2 years and then get fired because they don't have the figures to show for.
    Some funny Bavarian guy from Solvay Business School here in Belgium stated that to him, CMO are currently merely "Brand Stewards" and not takken seriously by the CEO. They should wake up to reality but unfortunately, here in Belgian they remain "older" and disillusioned by the first days of the Internet. As they don't understand the technology, they stick to what they know i.e. the "art" of marketing.

    I therefore remain patiently waiting for the right client and scan the market for the right company in order to build upon. Might have found what I was looking for actually ;-)
    Only time will tell whether the dashboards will indeed make their intoduction into the boardrooms and whether decisions will actually be based, at least partially, upon them. In the mean time, my company's strategy focusses on "best of breed" and on building long term relationships with our clients.

  3. The article may have an "ivory tower smell" to it, as is often true with academic content, but if you know anything about Babson, you know that most of the senior profs there are also practicing consultants. They don't operate in a "publish or perish" environment, which is why it's such a great biz school (MBA '83).

    I know this is an unpopular position because I have made this same point several times in Eric's web analytics discussion group, but over time, you will see all the silo-d analytics areas rolled up into one "Research & Analysis" function reporting directly to the CFO or CEO. This *must* happen as analytics in general become the basis for corporate decision making; if it has not happened yet, that means the company does not yet take analytics seriously; the decisions being made using analytics are not "life or death".

    Unfortunately, this analytics "rollup" typically does not happen because of "vision" in the company, but happens quite traumatically due to analytical "failure". The typical scenario is some kind of senior level meeting where there is the opportunity to compare / contrast the analytics coming out of several different silos, for example:

    The VP marketing presents analytics based on the company having 10,000 customers. The VP product area presents analytics based on the company having 11,500 customers. The VP customer service presents analytics based on the company having 9,500 customers.

    The CEO / CFO simply asks: How many customers do we have? And nobody can answer the question. Each VP defends their number as the "right" one, and cannot describe / defend the differences in customer counts.

    Pressing the issue, the CEO / CFO finds that several VP's have "reasons" to count customers the way they do - it skews the metrics / KPI's in their favor. It is found that the VP Product KPI's benefit from having higher customer counts; that the VP customer service metrics benefit from having lower customer counts. So the analyst in each silo, under (probably subtle, unspoken) pressure from the VP, has skewed the customer counts to make the metrics look better.

    Right there, right then, comes the "throbbing forehead vein" meltdown - if the company *truly* relies on analytics for decision-making. If corporate life & death decisions are being made based on the analytics, the above situation is outrageous, deadly.

    The CEO / CFO then declares all the analytics areas will be rolled up into one, reporting directly to either the CFO or the CEO. There will be one definition of a customer, and all reports *must* tie to each other. Any discrepancies must be explained.

    It's SarBox outside Finance. It will happen; I have seen it happen. Frankly, there is nothing that will *make it happen* faster than a bonus system tied to KPI's. When your ass is really on the line for the KPI's, and some other VP is producing data that infers your data is suspect, you will *want* it to happen.

    Typically, the highest ROI projects in terms of cost savings / improvements in service / CRM come from solving cross-silo problems. This work is almost impossible to do if you don't have "one voice" metrics. If you want to avoid the above meltdown and really get some work done, have the vision to roll up the analytics before the meltdown happens.

    With a unified analytical group, the company will make improvements at a much faster rate. It's just amazing how fast the different silo analysts learn from each other and begin to piece together the "real" picture of what is going on in a company *when they are not under pressure from a silo head*.

    All of a sudden, problems that seemed silo-specific become cross-functional process problems that end up actually getting fixed, and this is the real benefit of the analytical rollup.

    This is an "analytical culture" issue and I suspect most companies are simply too early in the analytical lifecycle to understand where all this is headed. When you look at companies that truly live and die by the analytics, you will find most have already done the rollup. Alternatively, there will be a "code of conduct" that has all analysts on the same page, or hybrid reporting structures - both messy alternatives.

    The analytical rollup rocks. Try it, you'll like it.


  4. Thanks Jim. As we continue to innovate quickly in the industry, I'm increasingly optimistic that we are building the right tools to deliver accurate and actionable analytics data into the enterprise.


  5. Jim,

    You describe centralizing analytics as "SarBox outside Finance". That's a apt description--given the cost/benefit ratio of SarbOx.

    Depending on the nature of your business, centralized analytics may be a good decision or it may tax resources and slow your ability to innovate.




webtrends reinvigorate analytics