Click for a Printer-friendly Version - Adobe PDF
Managing the Marketing/Analytical Partnership
By Cynthia Baughan Wheaton
Principal, Wheaton Group
Original Version of an article that appeared in the
April 1999 issue of Catalog Age
All too often, the relationship between analytical people and marketing people
is a difficult one. The marketing people do not understand the
terminology, or buzz-words, of the analysts. Often, marketing does not
understand what is possible through the use of analytical techniques, and they
do not know how to ask. On the other hand, the analysts do not clearly
understand the desired objectives of the project or the relevance to the bottom
line. At the same time, analysts believe that marketers do not understand
the power of what they offer.
In 1977, I was hired as a Systems Engineer at IBM. At that time, they
decided to take marketing people and train them in systems issues, while taking
systems people and training them in marketing. My management felt that
was important because the systems people they had were not always adept in
communicating with the end-user or client. At the same time, marketing
personnel could not "talk" systems issues. Just prior to a
month-long Assembler coding class, I decided that marketing was my true
calling. I was fortunate enough to join Sara Lee Direct in its infancy
and learn the world of database marketing from the ground up.
Since IBM, I have used my training to work with service bureaus and to
understand data issues involved in starting new catalog businesses or running
existing ones. Especially when starting a new business, teamwork with
analysts is critical to the creation of a well-constructed test - minimizing
test time, maximizing test effectiveness and evaluating test results.
The lessons I learned at IBM have served me well. The truth is that
marketing and systems/analytical people need each other, but do not spend
enough time communicating with each other.
One of the strengths of database marketing is that the numbers can "talk" to
you about your business, if you know how to massage them. Data is
important to the success of your catalog business. Understanding data
enhances your marketing ability and adds efficiency to your processes.
Most marketing people in our industry are at least somewhat adept at evaluating
standard data. However, there are processes and abilities that require
sophisticated technical training. For that, we need specialists. In
today's world of evolving technology, the role of analytical resources will
continue to grow in importance, and more often mean the difference between
failure and success.
Data analysis is an iterative process. The data points to
opportunities. For example, it may direct you to a change in positioning,
or help generate a spin-off catalog. Once results are analyzed, further
study of the data will help refine next steps.
Maximize the Relationship
Let's look at what you
need to know to make the relationship between marketing and analytical
personnel both productive and profitable:
Hold regular meetings to review marketing results and analytical
findings. Include analytical personnel. Let them be part of the
process and contribute with their special abilities and experience. Time
spent reviewing past and future promotions is critical to understanding as they
dig through the data. Without this, the most important questions may
never be asked.
Allow dialog between groups.
Use management by walking around. Get to know each other. Much is
discovered within casual conversation.
Set clear objectives for the business, the "event," or analysis.
Develop mutual understanding and respect for the processes each group
uses. Acknowledge that you need each other and can benefit from a
partnership. Be willing to acknowledge another's experience has validity
and work together.
Make every effort to speak the same language. Highly trained analytical
personnel can be intimidating to marketing/management. For one thing,
they often speak a different language. There should be at least one
person who has some knowledge of both languages who is willing to be a
translator. Everyone should attempt to speak the same language.
(See "Basic Terminology Explained.")
Know that using analytical tools is a process rather than a quick fix.
Being a data detective is not a one-time thing. Rather, in an environment
conducive to good research, it is ongoing.
Seeking the "Right" Analytical Help
could benefit financially from improved, or at least expanded, analytical
capability. The number of people who can do this work is relatively
large. But, if you want someone with industry experience,
the pool of talent is very small. This low availability means
that it is even more critical for each organization to have someone
on staff who understands both marketing and analytical issues, and
can "translate" for everyone else.
The perfect analytical person would be:
Knowledgeable of processes and techniques, but also of the data, from both
internal and external sources.
Able to focus on a project and complete it on time without going off on a
tangent. There are individuals who are absolutely brilliant, but become
distracted by low-priority patterns within the data and are unable to set them
aside to concentrate on data that can translate to the bottom line.
Interested in digging into the data, in order to understand why things
happen. This adds richness to the marketing process. And, when
there is a meaningful pattern within the data, that person is more likely to
address it in an appropriate - even optimal - way.
Respectful of other people and their contributions.
Experienced in different analytical processes, and able to communicate the
strengths and weaknesses of each, offering recommendations as to the best
methodology for a given situation.
The perfect marketing person would:
Respect other people and their contributions.
Not be intimidated by buzz-words or terminology.
Seek education on terminology and key issues, in order to increase
effectiveness in evaluating research results and implementing change into
Clearly communicate objectives and needs.
Understand the customer based on past behavior and research results.
Put the business in context within its market.
Develop the strategic plan.
Be available to answer questions, review progress as appropriate, and
Keep everyone focused on the desired end, and set deadlines with input from
Internal versus External Resources
If you are in
the market to hire analytical expertise, how do you decide between
hiring full-time employees and using external consulting resources?
There are advantages and disadvantages to each. The important
thing is to honestly look at your organization - its strengths and
weaknesses, as well as its "personality" in light of the following
Internal Analytical Resources
A full-time hire can be cost effective if the person is used steadily. It
can be a tremendous asset to have someone with a true data orientation think
about your business all of the time.
For example, one clever analyst was intrigued by an unusually high number of
back-orders. He uncovered a programming glitch that, under certain
circumstances, identified back-order situations when, in reality, the
merchandise was in stock. This had cost his company a significant amount
of money in cancelled orders from customers unwilling to wait for the
merchandise to be shipped. Fortunately, there was an inexpensive fix for
this "false back-order" problem, which had immediate revenue impact.
There will be day-to-day opportunities to fully integrate with marketing.
Not only will there be communication during scheduled meetings, but at the
water cooler or in impromptu lunches or office drop-in conversations.
If you have a small organization, you may have no one in-house to train,
develop, or evaluate that analytical person effectively.
There is a small pool of experienced talent from which to recruit.
External Analytical Resources/Consultants
There is a wider availability of talented/experienced personnel.
Outside resources usually bring the broader perspective gained from working
with multiple clients. They are further along the learning curve of
A long-term relationship can supersede personnel turnovers in-house. An
outside firm can actually become the repository for the analytical (and
marketing) history of an organization.
In a retainer, or ongoing, relationship, there are individuals thinking about
your business on a continuing basis. If they see opportunities
appropriate to your business, they will tell you.
Expertise is available for occasional projects without having to hire
Are more likely to be up-to-date on new techniques and processes.
The absolute cost can be high. However, if the research results are
implemented effectively, the investment can be paid back. With retainer
relationships, the real cost can be similar to in-house employees. One
large financial institution estimates that it costs $250,000 per year (in a
major metropolitan area) to pay salary, benefits, training, recruitment, data
processing support, licensing fees, overhead, and equipment for experienced
By definition, outside resources are not as accessible as someone in the same
building. However, some large clients do pay consultants to work
full-time on site.
Business issues do not always have statistical
solutions, just as statistical techniques do not always result in
good business decisions. And, there are always political issues
that require slow, evolutionary changes as credibility is built.
However, a solid partnership between marketing and analytical groups
will provide the give and take necessary to guide a strong catalog
Basic Terminology Explained For Marketers
RFM Analysis: Selecting
panels of buyers or prospects for promotions based on permutations
of historical criteria such as recency, frequency, monetary value,
and product type. Generally, the RFM criteria, as well as
their associated categorical breaks, are intuitively determined,
based on historical performance. By definition, each panel,
or "cell," is homogeneous in composition (e.g., everyone within
a particular cell will have 0 to 6 months Recency, two Lifetime
Orders, and $50 to 100 Average Order Size). Some companies
have cells that number into the thousands, causing sample-size problems.
Multiple Regression: A statistical technique that: 1) interrogates
multiple potential predictors (i.e., "independent variables"), 2) finds the
subset that best predicts future behavior (i.e., the "dependent variable"), and
3) weights them (i.e., "assigns multiplicative coefficients") in such a way
that a file of customers or prospects can be sorted in terms of most to least
desirable predicted behavior.
Regression assigns a unique score (i.e., predicted behavior) to every
individual. But, very different individuals can receive the same scores
(e.g., a person with 36 month Recency and three Lifetime Orders might have the
same score as someone else with 3 month Recency and one Lifetime Order).
Tree Analysis (e.g., CHAID): A statistical method of dividing customers
into homogeneous groups by purchase history and/or demographics. The
resulting groups can be rank-ordered by some performance measure such as
response or sales. They are always applied to situations where there is a
dependent variable and a number of independent variables. They can
provide insight into customer behavior and result in the identification of
Statistics defines the best variables to include for grouping. Tree
analysis creates cells just as does RFM, but is less dependent on human
intuition. Generally, human intuition determines the variable "breaks"
during the data preparation stage (e.g., Average Order Size = $0 to 25, $25.01
to 50, $50.01 to 75, etc.). However, Tree Analysis uses statistics rather
than human intuition in determining how the variable categories should be
Tree Analysis often is used as an intermediate step in regression, to find
"interactions" in the data (e.g., individuals who are both older as well as
affluent might be particularly interested in buying a Cadillac).
Multiple Discriminant Analysis: A classical statistical technique to
classify observations and assign them to distinct, mutually exclusive groups.
These groups are reviewed and descriptive names applied to each (e.g., "soccer
Neural Networks: Originally developed to mimic the human brain process,
these methodologies "learn" from data via various mechanisms. They make
no assumptions about the distributions of predictor variables or targets.
They can model highly non-linear relationships and capture interactions that
are difficult to see. Often used for pattern recognition. Because
they are highly sensitive, any problems with the data are exacerbated.
However, if the problem is defined correctly, with a strong research design,
these types of methodologies can be quite effective.
SAS & SPSS: Popular
statistical packages of application software that are used to do
regression, create clusters, perform tree as well as discriminant
analysis, manipulate data, interrogate data sets, and more.
Statistical Significance: The determination of whether or not two results
are different enough to be considered "real." For mailers, most often
used to determine if the response differences between two test panels, or test
and control panels, can be used to determine a "winner." Also used to
evaluate research results. This is determined by statistical
techniques that incorporate: (See illustration)
Bell Curve: A visual depiction of a normal distribution of
observations. It is, by definition, always symmetrical.
Confidence: (a.k.a. Confidence Level): Degree of certainty in the
accuracy of a test result; often expressed as a Range (a.k.a. Confidence
Interval) around the test result. Example: Test result of 1.0% @
26,790 quantity. Have 90% Confidence that "true" universe response is
within the 0.9 to 1.1% (Range). The theory behind this example is
that if an experiment is repeated 100 times, we should expect the result to
fall within the Range 90 times.
Precision: One-half of the Range. If you want the results to e +/-
10%, 10% refers to the precision. Note: A Confidence Level of - say
- 90% does not necessarily have Precision of +/- 10%.
Tails: Extreme areas of the distribution, both to the left and right of
the Mean (a.k.a. Average). Tails typically lie plus/minus two standard
deviations beyond the Mean of the distribution.
Outliers: Atypical observations. Approximately 95% of the
observations should fall within two standard deviations of the mean.
Extreme observations can fall into either the left or the right tail of the
Name Selection Techniques
Nth Name Selection:
Generally used for test panel record selection. Every "nth"
record is chosen. If you are selecting - say - 10,000 records
out of 1 million, every 100th record is chosen. (In this case,
"n" = 100.). This technique can cause problems if: 1)
there is an intrinsic pattern within the data set (e.g., you want
a 50% sample, and the data set is ordered male, female, male, female),
or 2) for list order fulfillment (e.g., ten orders of a 10,000 Nth
could result in the same 10,000 names being selected ten times).
Nthing With A Seed: Often used in list order fulfillment to ensure
against the problems inherent in "normal" Nth'ing. For example, if ten
companies ordered list tests of 10,000 from a universe of 100,000, this process
would prevent all of the companies from receiving the same records. For
each company, one record from the first 10 (i.e., the universe divided by the
desired panel size) would randomly be selected, and then every 10th record
subsequent to it.
True Random Sampling: Each record is selected randomly and has an equal
chance of selection. For instance, to select 10,000 out of a 100,000
base, the first name chosen may be #682, then, # 99,204, then, #88, and so
on. This is relatively expensive from a data processing standpoint.
Stratified Sampling: A process to improve the "representativeness" of
test panels compared with "normal" or "common" sampling. This is done by
dividing the universe into subgroups based on a factor(s) that correlates with
what you are trying to measure, and then sampling within these subgroups.
For example: to measure next 12-month sales: 1) rank the universe
based on previous 12-month sales, 2) divide into equal 25 panels, and 3)
Nth-select within each panel. Often used when dealing with extreme price
Test Panel Design
Holdout Panels: Known customers
who receive no database driven promotions. They can be influenced
only by other, non-database driven, media.
Longitudinal Test: A long-term test using an A/B split(s) to measure
cumulative behavior over time. Multiple panels can be used. Often,
6 to 12 months in duration. For example, Panel A receives multiple
database promotions; and Panel B no database promotions. Useful because
database promotions often change behavior slowly, which cannot be measured with
a single test. Especially useful in open loop situations such as retail
where the target audience is receiving stimuli outside of database promotions.
Mean: The average result.
In the Bell Curve illustration, 1.0%.
Median: If all of the observations in an analysis were listed from high
to low, the median is the one in the middle, regardless of value. For
instance, if you took a sample of 99 observations, the median would be the
actual value of the 50th observation.
Mode: The observation value that happens most frequently.
Cynthia Baughan Wheaton is a Principal at Wheaton Group, which specializes in
direct marketing consulting and data mining, data quality assessment and
assurance, and the delivery of cost-effective data warehouses and marts.
Cynthia can be reached at 919-969-9218, or email@example.com