80 5.5 Wave 2 analytical CRM
5.5
throw away my hard-earned money), and my online purchase behavior does
not include the purchase of complex financial products through this medium.
It is clear that none of the accepted segmenting techniques (e.g., cluster analy-
sis, regression analysis, K-NN/CBR) was brought to bear when creating target
lists. Why should Hedge Funds World care? It wants tactical sales maximiza-
tion, not relationships.
It should be noted that analytical CRM is not focused exclusively on
marketing campaign management. Reporting and personalization of Web
content also were introduced. In the case of reporting, the statistical work-
benches that had been so prevalent when CRM was merely data mining were
applied to CRM data. These reports were used either as diagnosis tools or as a
means to report performance to management. Areas of focus varied as greatly
as the applications they supported. Click-stream analysis, shopping cart aban-
donment, cross-selling, and upselling (as per the Anthony Robinson example
at the beginning of the chapter) were all common reporting areas. Reporting
in customer interaction management applications, for instance, uses either
homemade or embedded tools to report agent utilization levels, categorization
accuracy, time to respond, and the like. These reports were at the end of the
chain. They did not, in Wave 1, typically trigger an additional system
response.
Personalization, likewise, was introduced and refined during the first
wave of analytical CRM. The focus initially was on data filtering (e.g., I have a
predilection for the "Oddly Enough" news category and, therefore, elect to
have those news articles forwarded to my portal) and on profiling (e.g., my
portal has recorded my postal code and, therefore, displays my local weather
forecast accordingly). Once again, while these were important first steps
toward analytical CRM, they were not advanced in terms of the analytical
applications embedded within them. That would begin in Wave 2.
Wave 2 analytical CRM
As brick-and-mortar companies increased their use of and presence on the
Internet during the past several years, they have brought with them their
knowledge of data mining and customer profiling. Not surprisingly, there-
fore, the demand on
analytical
CRM vendors has become greater. The use of
the "analytical" qualifier is a bit amorphous as analytics have become more
embedded in all CRM applications. In fact, with time, it is clear that analytical
CRM will become less a category of CRM software and more a set of embed-
ded processes. This will be examined subsequently (when we discuss Wave 3).
5.5 Wave 2 analytical CRM 81
What enables this second wave to take place is that customer data, once
stored in multiple locations and systems, is in the beginning stages of amalga-
mation. Just as bar codes facilitated the first wave of data mining, the addition
of real-time customer interactions via the Web is fueling the second. While
the terms that govern this new data repository vary from customer data
warehouse to CRM data marts, leading CRM vendors have recognized that
domination of the CRM marketplace is predicated by control over customer
data. Companies such as Siebel, Kana, and Oracle want to "own" the master
customer files in order to control this strategically precious resource. It should
be noted, however, that it is unclear whether a select few CRM vendors will
come to dominate this area or if customer data warehouses will be revolution-
ized by incumbent data mining solutions such as IBM Data Miner, SAS, or
SPSS. What is certain is that such customer data files allow analytical conclu-
sions to be reached "just in time" with operational processes.
So what is this customer data warehouse? Simply, the customer data ware-
house is an area where existing customer transactions, customer service
incidents, demographics, and macroanalysis can be stored, accessed, and coor-
dinated to ensure relevant, timely, and appropriate customer communications
that in some way enhance the customer experience. While much of the infor-
mation stored within it will likely come from internal sources, it is also likely
that this information will be combined with external information as well (e.g.,
credit reports, credit card data).
Let's not take this customer data warehouse as a replacement for existing
customer data. As data mining specialists and their applications have evolved,
they have become adept at pulling information from disparate systems. Not
only will this trend continue, but since Internet media channels are being
added alongside traditional ones, integration will continue to be pursued. As
we outlined in Chapter 3, there are several ways that integration can be under-
taken. Middleware integration, which is currently the cleanest and least
labor-intensive way to connect disparate applications and legacy systems, will
clearly play a more prominent role during the next several years. It is likely,
however, that as standards emerge for the customer data warehouse, CRM
applications of all sorts will rush to integrate as tightly as possible. This ware-
house will, in fact, become a backbone of sorts.
The most direct effect that this customer data warehouse has on customer
interactions with companies is that it enables more sophisticated techniques
to be employed for analyzing customer data, interactions, and behavior. It is
possible to combine information about Web site interactions with financial
I Chapter 5
82 5.5 Wave 2 analytical CRM
demographics and geographic information. That said, we should not mistake
complex analysis for good analysis. Take the following example:
A company notices that there is a high drop-off rate in shopping carts
and decides to launch a "point of sale" promotion to encourage
shopping-cart conversion. The company is considering several girls as
a part of this promotion: a minidisk player, frequent flier miles, and
free shipping. It sets a trial of each of the three gifts with a limited set
of customers. The company finds that overall the minidisk player
converts an additional 2.5 percent of customers; the frequent flier
miles, an additional 3.5 percent; and the free shipping, an additional
3.0 percent.
Based on this information, the vice-president of marketing is pre-
pared to offer frequent flier mileage as the incentive, when his trusted
data miner, Vanessa McShelton, bursts into the room. She points out
that even though frequent flier mileage was preferred overall, by seg-
menting the customer base, she determined that younger customers
prefer the minidisk player and that free shipping is by far the most
popular choice for people with families. "By targeting the incentive,
we should be able to boost our overall conversion rate to 4.3 percent,
a full 1.3 percent higher than if we simply offer the frequent flier
miles."
While it is a bit flippant, this example does highlight two important
myths:
Myth 1: Analysis has to be complex.
What Ms. McShelton did was a
very simple set of analyses by predetermined demographic segments.
Although care must be taken to determine how these segments are
defined, it is not an algorithm-intensive process. What it does require is
a good bit of common sense and some business knowledge. None-
theless, through very simple analysis Ms. McShelton was able to iden-
tify important information that would earn the company additional
revenue.
Myth 2: Analysis and reporting are the same thing.
Based on a series of
analyses, Ms. McShelton recommended a targeted marketing cam-
paign. In essence, she is recommending that the analysis she com-
pleted be captured and then used on a pro forma basis to make targeted
offers to inbound potential customers. What differentiates this from
the simple reporting that companies were conducting in Wave I is that
analysis is used in an ongoing basis to continue to create value for the
5.5 Wave 2 analytical CRM 83
company and for customers. By contrast, most reports are used for data
aggregation and performance reporting.
It should also be noted that this proactive offering is not possible without
a customer data warehouse to identify the demographics of inbound custom-
ers. In fact, this points to the type of "one-to-one" marketing that data
warehousing makes possible. Rather than focusing exclusively on product
upgrading and cross-selling (although let's maintain no illusions as to whether
these forms of "marketing" will continue; the compunction to ask, "Do you
want fries with that?" is just too strong), incentives that are specific to the indi-
vidual can be offered.
The customer data warehouse has a number of other important effects. As
more complex analysis becomes available, marketing campaigns become both
more targeted and more complex. Referring back to our Hedge Funds World
example, it is my great hope that that company might become more sophis-
ticated. Hedge Funds World might combine response rates in future solicita-
tions, incorporate third-party data to more directly target the market of
potential candidates (and leave the rest of us poor saps alone), and even
involve multiple communication channels within the same marketing cam-
paign. Along with this enhanced targeting, the ability to track individual
solicitations is also becoming increasingly apparent. The ability for firms to
identify information gatekeepers is critical to understanding and manipulat-
ing viral marketing.
The benefits from the customer data warehouse are not limited to out-
bound marketing. Customer service applications are better able to serve their
most valued customers, because they can easily identify them; sales force auto-
mation programs are able to incorporate both real-time prospect data and
aggregate level analysis from this target group; and all customer-facing staff
have a complete 360-degree view of the customer.
It should be noted that as analytics advance, however, there will be pock-
ets that develop a specialized focus. Search engines have, for example, evolved
from simple word matching at their core to allowing the user to use natural
language to ask his or her question (e.g., Google or Ask Jeeves). Neither the
user nor the implementer need understand the specifics of the artificial intelli-
gence that makes the search engine work. Similarly, text categorization
for inbound electronic customer service interaction has developed to the
point where a select few vendors in this area use artificial intelligence to train
and categorize inbound messages with a focus on reducing administrative
burden.
I Chapter 5
84 5.5 Wave 2 analytical CRM
In addition to the advanced analytics mentioned previously, it should be
noted that there will be an increasing need for specific analytical applications
that drive parts of the overall CRM effort. There are, however, a couple of
important distinctions between the two. Primarily, embedded artificial intel-
ligence will need to function in an almost invisible fashion. The advanced
heuristics and techniques are designed to ease the burden on the implement-
ing companies; thus, we will see continued development of applications that
are powerful but have easy-to-use training tools and user interfaces. This will
be doubly important, because statistics and data mining professionals are
likely to be in high demand in the coming years. It takes very sophisticated
knowledge to generate useful analysis.
To be truly adept at data mining, a professional needs skills in a number
of areas. Not only does he or she need to be familiar with statistical techniques
and tools, but the data miner must be able to apply industry logic, business
acumen, and a healthy dose of common sense to the analysis as well. Inciden-
tally, since this array of knowledge is difficult to maintain in one individual, it
is extremely valuable if this person is a skilled communicator. While the other
skills might be easy enough to come by, it can often be a challenge to find pro-
fessionals who meet the common-sense criterion; therefore, it is easy enough
to recognize that those with this confluence of skills will be in great demand as
the second wave continues forward.
It should also be noted that during this second wave, personalization is
evolving from data filters and profiling to include preference matching as well.
For preference matching, an individual's preferences are recorded. These
collections of preferences can then be compared with other individuals' pref-
erences. Suggestions and recommendations can than be put forth.
For a great example of this personalization technique, go to Amazon.com.
Amazon gives book and record purchasers the ability to enter their collection
of books and/or CDs and then rate them. As a result, the company suggests
titles that others who rated your book or CD favored as well. In essence it gives
Amazon customers the ability to talk with a friend about books or music.
In addition, Amazon provides individuals with the ability to publish lists
of favorites. As users peruse books and CDs, they can see the top-ten lists of
others who included the title. I realize that choosing Amazon as an example
is somewhat risky. The company has been criticized for never running a
profit; however, I believe that this company consistently leads the way when
it comes to customer interaction management.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.