How can we make sure that we’re marketing the right products to the right customers? How can we reduce our default rate? If we reduced the price of the product, would it increase demand? Those are valid questions for any company, but they are difficult to answer without delving deeply into your stores of historical data and finding effective ways to slice, dice and analyze that data.
For years, large enterprises have turned to sophisticated data
mining and predictive analytics tools to do just that. With these tools,
companies can use their own data to build a predictive model that fits their
business and feed the system a variety of sophisticated “what-if” scenarios to
determine the best way to grow their business.
“It’s a great tool if you’re looking to improve things like
default rates or credit scoring, or to do more opportunistic marketing,” said
John Elder, president of Elder Research Inc., a data mining consultancy in
Charlottesville, Va. “It’s very useful, for example, to be able to pull up
information about a customer and other products they may be interested in when
a customer calls the company. If they have already done the analysis
beforehand, they can use the information as a sales opportunity.”
So what’s the problem? If data mining is such a valuable tool, why
don’t more midmarket companies use them?
It’s not that they don’t want to, said James Kobielus, a senior
analyst at Forrester Research of Cambridge Mass. It’s that these tools tend to
be very complex, requiring a knowledgeable, specialized IT presence within the
company—something many smaller companies simply can’t afford.
But data mining vendors have gotten the message, and in the past
several years, many of them are working hard to make their products more
user-friendly. What’s more, other data mining vendors have surfaced that focus
more specifically on the midmarket segment.
“All of these vendors have made their tools more user-friendly,
and they don’t all require a PhD in statistics to use,” Kobielus said. “They
can all pull historical data from customers’ existing databases, and the Excel
plug-in is also pretty much universal. Usability has definitely improved over
the past five years.”
Microsoft has been at the forefront of the trend, offering
plug-ins to both Microsoft Office/Excel and SQL Server 2005. Using the
Microsoft SQL Server 2005 Data Mining Add-In for Office 2007, for example, a
business analyst can retrieve structured data sets from SQL Server, build an
analytical data mart, and then apply a variety of statistical algorithms to the
data to cluster the data into specific groupings. From there, the analyst can
look for specific buying patterns among specific segments of customers, and
then create interactive visualizations of the data, such as histograms and
maps. The plug-ins also allow analysts t perform a variety of what-if analyses.
Other easier-to-use data mining tools come from companies like
Angoss Software Corp., InforSense and Unica Corp.
With these tools, business analysts can answer a lot of important
questions, most in the form of “what if?”
“If we change this underlying independent variable in this way,
how would it affect these dependent variables? Or if we lowered the price by
10% for this customer segment, how much would it increase sales?” Kobielus
explained.
The next set of data mining tools is more comprehensive, but more
difficult to use without external help. They include stalwarts like SAS
Institute, IBM, Micro Strategy, SPSS, Fair Isaac, Teradata, IXM, ThinkAnalytics
and Portrait Software.
In most cases, simpler data mining tools such as those from
Microsoft, Angoss and others will do the trick. But in cases where you need to
delve more deeply into the data, you might need the deeper and broader algorithms
of these more sophisticated tools. If you want to run the same model with
different statistical algorithms and compare the results side by side, the
high-end tools are for you, Kobielus said.
But if you go that route, don’t expect to be able to do it all
yourself. It can make sense to hire a consultant, either short- or longer-term,
to help set up scenarios and teach internal analysts how to get the most out of
the investment.
“It’s an extremely good use of money to have a consultant come in
for a day or two and look at your problems and make recommendations, even if
you don’t use them for a long-term engagement,” Elder said. “Consultants know
what has worked for similar companies with similar challenges, and that can be
invaluable.
For companies that want more ongoing help, Elder Consulting and
others like it also provide training and ongoing support for data mining
activities. They also provide full-service outsourcing, which can be useful for
companies with little internal expertise or IT staff.
“Finding the right partner
is key for midmarket companies,” Kobielus said. “Make sure the vendor or
consultant you pick can help you build the analytical data mart, and train your
people to do it themselves.”
