August was a busy month for our SAS Text Analytics team, but I was very pleased to get the invite to attend this year’s CIO 100 Conference hosted by Maryfran Johnson, Editor-in-Chief of CIO Magazine.  As Master of Ceremonies, she inspired lively discussions and shared insights on IT leadership, technology and business trends. But what I found particularly valuable was the chance to interact with the IT community.

As an analytics practitioner, I tend to meet primarily with business users. So, listening and speaking with CIOs and other IT professionals gave me new perspectives and insights into both the challenges they face as well as the emerging opportunities in front of them in this new "big data" world.

One of the key topics of the conference, and one that SAS is committed to, is how to enable our clients to leverage advanced analytics as a competitive differentiator.  There was one standout session that really got me thinking and motivated to blog about the ever-widening gap that continues to persist between IT and business.

In the session titled Who’s Driving Your Business Analytics? Bob Melk, Senior Vice President and Group Publisher of IDG Enterprise presented the results of a recent survey of IT leaders.  Here was the question asked:

Does it matter who’s in charge of your business analytics projects?

“Yes” was the resounding answer.  And the respondents clearly communicated that both IT and business should have equal ownership over business analytics projects. The unfortunate reality though is that fewer than 40 percent of organizations share this business critical responsibility.

Highlighted was the fact that IT and business users still struggle when it comes to driving business analytics capabilities throughout their organizations.  IT seems to be of the opinion that business users aren’t giving them the insight they need to collect and transform data for analysis.  They feel that business users seem to “hold back” knowledge about business context needed to make these projects successful.

From my (analytics practitioner) point of view, this couldn’t be further from the truth.

Analytics professionals as interpreters
I’ve spoken with hundreds of business users over my career and the problem is really about effective communication.  Business users and IT don’t share the same language and are biased by the lens that they look through to describe what they need.  Often business users know what the end goal is, but lack some combination of three things:

• The ability to translate needs into what kind of data is needed.
• The knowledge of what data preparation needs to be done for each kind of analysis.
• The advantages or disadvantages of various analytical methods to ensure best results.

For example:  Your marketing department wants to run a campaign to sell your latest widget. They ask the obvious business question: “Who should we target for this product?”  It’s really that simple in their mind.

What isn’t quite so simple is getting the necessary data in the right format to enable them to execute such a campaign with the best chance of maximizing their return on investment.

I contend that there are three parties that need to make business analytics projects successful: business, IT, and an analytics practitioner.  This analytical intermediary is needed to listen and translate between business and IT to facilitate the analytical decision-making process.

Questions to be considered jointly by business, IT, and the analytics professionals
In this example, in order to produce an effective campaign, it’s necessary to identify the kinds of data available to model which prospects are likely to respond.  The findings could include things like data from previous marketing campaigns, past buyers and non-buyers, prospect list demographics, survey comments, and other social media like customer reviews, etc.

Assuming that this kind of data is available, there are still lots of questions one could ask:

• Is the data in the right format for our needs?
• Is more data necessary to support our approach?
• How much better would our results be if we could get additional data?
• And what would the cost be of getting it?
• What kind of analysis should we perform to decide whom to target?

Let’s explore some of these questions.

To determine a customer's likelihood to purchase your new widget, you could assume your customer’s previous purchases should be a good indicator of any future purchases.  So, are those past purchase records sitting in the database? And, are they formatted as individual transactions?

For an analytical model, the total amount of previous purchases would likely need to be calculated from the individual transactions.  And, looking at the projected revenue of a proposed campaign, the amounts should all be in the same currency.  So multiple currencies need to be converted to a standard currency for the analysis.

As a business user or IT professional, you may not be accustomed to thinking like this. But as an analytics person, you know that there are many data manipulations that need to take place before you can begin the analysis to develop insights.  These data transformations probably will not be stored in the data warehouse, but more likely they will reside in an analytical data mart used for a specific type of analysis.

Still more questions
So, with the data I have, what kind of underlying analytics could be performed with the available data? Most often, analysts might create a predictive model to determine the likelihood that prospects will purchase the new product.  The question we always ask ourselves is: What kind of model should we use here?  The data sometimes determines our strategy, which in turn, determines the best model to use.  Should our model use regression, decision trees, neural networks, others?  Ah, so many decisions! And still more questions…

If the data about each prospect contains a lot of missing values, then a regression analysis might not be the best method, as regression analysis is not tolerant to missing values.  So, should missing values be imputed in order to use the regression technique?  Or should a different method be used… like decision trees that are much more tolerant to missing values?  How many of the records have missing values, 10 percent, 20 percent, 50 percent?  How will that affect our analysis?

What about decision trees?  While easy to understand, they often don’t provide the predictive capabilities of other methods.  Why does that matter?  Because the biggest question is, what are we giving up in terms of return-on-investment for the campaign using each of these methods?

You see where this is going: Someone (hint: the analyst!) needs to evaluate the data to see which analytical method would fare the best.

Analysts are paid to know the pros and cons of each approach inside and out. It’s what we do!

So you see how one simple question from a business user: “Who should we target?” quickly presents both analyst, and subsequently IT, with many, many questions to both contemplate and answer. Who better than analysts to interpret marketing’s needs and translate them into IT requirements?

Where to find analytical intermediaries
It’s hard to say where they might be in your organization.  These people don’t consistently reside in either IT or in the business units.  They are scarce resources, even in large organizations, and their unique talent needs to be shared across departments within the enterprise.  I believe they should reside in Business Analytic Center of Excellence (BA CoE), an organization staffed with members from IT, the business, and analytical professionals, which are a joint resource accessible to support the entire organization.

The BA CoE enables organizations to be more proactive, agile players, as described by Thornton May, an IT Futurist, in discussing how organizations need to transform their business models from backward-looking operations.

The good news is that many organizations are establishing BA CoE’s to support their business strategies, and compete on analytics.  These groups, when structured properly, can also act as a change agent to boost the organization’s maturity to an analytically-driven level and provide structure for organizations to hire, mentor, and grow their analytical talent.  An example is the bank that built a SAS BA CoE to help people in different functional or geographic areas shatter boundaries that had prevented collaboration. As more people across the organization deployed fact-based decisions, this organization built the foundation for optimization and innovation.

Today, competing on analytics has become a fortunate and profitable reality for organizations that bring together business, IT and analytics talent.  The SAS BA CoE organizes and engages analytical talent – assigning problems to those with just the right skills. This not only expands analytical bandwidth but as importantly helps nurture and retain what I believe is one of an organization’s most valuable assets – the Analyst.

But sue me, I’m biased!