ACM 476 DATA MINING PROJECT
Cross selling is a strategy that pushes new products to current customers based on
their past purchases. It is designed to widen the customer's loyalty on the firm and
decrease the probability of the customer switching to a competitor.There are some
elements that might influence the definition might include: the size of business,the
industry sector it operates within and the financial motivations of those required to
define the term.The obvious metric that most people think of when evaluating
cross-sale promotion effectiveness is conversion rate.Conversion rate identifies the
percentage of people who, when buying the “original” product, choose to also
purchase the “promoted” product.
Making changes that raise conversion rate increase sales of the promoted
product.However, this conversion ratio is more a reflection of the relevance of the
promoted product to the original product than it is a measure of profitability.The
core goal is to maximize profitability, while providing additional value to customers
(who are better off or more satisfied with the original purchase when they also
purchase the promoted item). Introducing a cross-sale promotion can increase,
decrease, or have no effect on the rate of purchase (conversion percentage) of the
original product.Business owners need to measure the original-product conversion
rate for their customers who were shown the cross-selling promotion versus those
that were not. Second, you’ll want to know what the ideal products to promote are.
Typically, more than one cross-sale promotion is presented to a customer at a time.
For this example, assume 3 promotions are displayed.
Let assume that a data-mining exercise has identified 5 products that could be
promoted as cross-selling items for this “original” product. Which three of the five
can be selected? The three that is expected to be the most profitable should be
selected “Most profitable” can be calculated as (profit per promoted item) x
(conversion % – of the promoted item in the context of the original item).
If there is no conversion percentage data, it can be gathered (through testing) or
predicted (through modeling). There are many aproaches for predicting the degree
of similarity, or implied relevance, of one product to another – but all of them are
too detailed to cover in a blog article. Some companies also report average order
value (AOV), but that’s not necessarily an indicator of profitability. It may be a
component of profitability, but not necessarily.There is one of real life business
example which is in fast food industry.Customers are often invited to try new
products or established complimentary items.For example, when an individual
orders a hamburger at a local fast food restaurant, the server will often ask the
customer if her or she would like a side item to go with the hamburger. If the
restaurant is offering a new dessert, the server may also suggest to the customer
that the new item may be a desirable compliment to the hamburger. By employing
this simple approach, the server may entice the customer into making another
purchase above and beyond the one originally intended.
There is one more example about cross-selling,and it's in retaling sector. When
purchasing a washing machine, the salesperson may extend some type of special
offer to the customer in order to entice the buyer into also purchasing a dryer.
Managers who are responsible with their own product's marketing should not think
that we are irritating them with too many sales pitches to buy more , but they
consider the some ideas to improve opportunities for cross-selling.
One of them is that let nature take its course.Many cross-selling opportunities arise
naturally. If you are selling tennis racquets, for example, you can also offer a bag,
balls, lessons and accessories. To gain the extra sale, you might simply have to
mention that the other products or services are available.
The other idea is "Stay Relevant".If you overload customers with too many unrelated
cross-selling suggestions, you may blow it. Offering socks with shoes is certainly a
good fit. But if your attempts to cross-sell are not closely related to the original
purchase, they are far less likely to succeed.
Moreover, "Post expert recommandations" is the another idea.One way to facilitate
cross-selling and up-selling success is to state specific recommendations from
professionals, experts or other customers. This could be a chef’s recommendation
on a menu, a doctor’s recommendation on a mailer, or lists of related items that
other customers have purchased on a website. When you buy a book at
Amazon.com, for example, the site automatically lists other books purchased by
people who bought the same book you just ordered.The fourth idea is "Train
employees in cross-selling techniques". The approach must be built around serving
the customer, not just selling more stuff. For example, you might describe how the
additional products or services would complement the original purchase and
further solve the customer’s problem.
The other important idea that time is important.ss-selling and up-selling can occur
at different times, depending on the products and services you are selling. In some
cases, the best time is while a customer is trying something out. If they are looking
at a low priced digital camera, for example, but seem disappointed in a lack of
features or performance, they may really want a higher priced model. Or you could
suggest a belt to go with a pair of pants while the customer is trying them on. Other
items are more appropriately offered once the initial buying decision has been
made, such as an extended warranty.
Furthermore, "leverage the cross-selling potential of your website" is an other
crucial idea for cross-selling. Position cross-sell and up-sell items throughout your
site in places where they can help educate shoppers on the depth and variety of
what your business offers. Try mixing and matching different items to see what
works best.One of the important idea in order to improve the opportunity for cross-
selling is "offer a range of prices". If you suggest three items to complement a
product, try to offer a mix of price points. The lowest cost items are most likely to be
picked up as impulse buys. But other items that meet the customer’s needs can also
sell at higher levels.
The last important idea is to try product or service bundles. Bundling has long been
used as a way to entice shoppers to buy not just a single item, but an entire group
of items that go together. Offering a price break on package deals will help close
Moreover,there is one thing important which is need to be together with these
eight ideas is the recognizing the customers well. I found some short discourse
about cross-selling and up-selling which were written by Okan Özdemir, Agents
Sales Group Manager of Allianz and he has a motto about it like "Cross sale
increases the loyalty of the customer ".
"Customers should be described well for cross-selling"
If one insurance agent keeps many policies which are belong to one insuree,the
insurance agent becomes a consultant for this insuree so that he may trend to have
the familiar agent make his own transactions. Also,we are strongly recommending
that the insurance agents should consider the customers' datas,purchasing
behaviours and overall tendency in order to increase the agent's selling
potential.Moreover,CRM system must work actively in the background so that these
goals which were mentioned can be fulfilled.On the other hand, there is one
negative fact about the people who do not cross sell or up sell. The one reason
which they don't use these techniques because of. The point is the "FEAR !". Fear of
losing is the number one reason salespeople don't cross sell or up sell more
often.Here are the some big reasons why sales professionals continue to resist
cross-selling and up-selling ;
"Will I risk the order I just received if I ask for more ?"
"I'm not going to confuse this account by pitching a bunch of new products."
"I'll ask for a referral later-once we are in a better position with this account."
"This isn't my area of expertise."
"10 Cross-selling Techniques That Work"
• Be Relevant
Offer suggestions which are related to the original sale, but do not attempt to sell
your customer the entire store. If a customer buys some peanut butter, show them
the jelly rather than the green beans. Your customer does not want to be inundated
with too many choices when they have selected an item.
• Maximize positioning on your site
Place your cross-sell items where they will be seen near related items on your site.
Use a suggestion tool which is offered with a lot of eCommerce software. Let your
customers see the accessories that you have available, but again, keep the selection
limited to a few.
• Bundle your products
Offer discounts for buying items in packages. Combo meals have worked extremely
well for the fast food industry. Your customers thrive on the thought of making
deals, so they will spend ten extra dollars just to save five on a pair of products. That
behavior increases your bottom line.
• Offer several price points
When you are giving your customers a choice, present products from three different
price points. You want them to know about the latest and greatest items, but you
also want to give that customer some practical choices about what to buy. Do not
be completely focused on cross-selling or up-selling the most expensive items.
• Be observant
This suggestion applies more to the brick and mortar cross-sellers. Watch your
customers and understand the signals that they give you. If you can tell that
someone is not happy with the features of your skis, give them an option of buying
the skis that are higher on the price chain. Timing is everything, and the astute
salesman can tell when their customer is making their decision.
• Offer free shipping
Free shipping is a great way to get your customers excited about an order. Set a
price threshold that your customers must pass to qualify for that bonus, and they
will do quite a lot to reach it. If a customer is a few dollars away from the mark, they
will add another item to reach your arbitrarily determined price point.
• Create some urgency
Use the idea that your items are available at this price for only a limited time. Give
your customers some urgency by telling them that there are only a few of their
favorite items left on the shelves. Keep the deals short so you can use phrases like
‘ends Thursday’ in your advertising.
• Show your top rated suggestions
Set up a system of relationship selling. Let your customers know what other
customers have bought and what products they believe are the best. Give your
customer the ability to compare feature to feature on each of the products that you
have, rather than striking all over the board. Your customers love to receive
• Keep ad-ons on check out
Customers do not want to repeat the process of checking out when they have an
add-on item. They do not want their impulse buy to throw them to the back of the
line. They do not want to re-input their credit card information just because they
wanted to buy a pair of socks with the shoes that they were purchasing.
• Make the experience personal
Your customers are the most important individuals in the world. They are the ones
who keep the doors open, and they want you to cater to their needs. Focus the
attention on them when you are cross-selling by using phrases like, ‘you might also
like’ or ‘do you want’ rather than canned corporate phrases that start with ‘we
Data Mining – Cross Selling
"What is Data Mining ?"
Data mining is the process of extracting patterns from data.Data mining is seen as
an increasingly important tool by modern business to transform data into business
intelligence giving an informational advantage. It is currently used in a wide range
of profilling practises, such as marketing, surveillance, fraud detection, and scientific
"Data Mining in Business"
Modern organizations are overwhelmed with data, producing mountains of textual
documents, spreadsheets,and web pages. Every activity of the organization yields
more documents with each passing day.But a surprisingly large volume of data
remains relatively inaccessible. Any organization is likely to have data that remain
tantalizingly out of reach, either through a loss of institutional memory or through
needle-in-a-haystack search complexity.The idea that information resources should
be counted as an asset, and managed to yield maximum advantage emerges in
cycles every few years. The concept has most recently manifested itself as part of
the “knowledge management” approach. However the idea is packaged, there is a
growing acknowledgement that information resources are important, and that they
should be consciously cultivated.
"Data Mining Models for Cross Selling"
If your company offers multiple products or services, you can use data mining
models to find prospects for cross selling:
• Based on the products customers have bought in the past, data mining
models can help you determine which products they are most likely to buy
• In addition to the information that you have about your relationships with
customers, you can leverage demographics to see which customer segments
are most likely to add more products or services to the bundles they currently
• With the results of these data mining models, your company can allocate
resources toward your more valuable customers and prospects.
If you record information about your customers in a database, you have the
information necessary to deploy many data mining models. At The Modeling
Agency, we use the latest models to provide our clients with insights to improve
their businesses. We also offer a range of seminars so that you can learn to
implement the same data mining methods in-house.
Success Story:Dexia Bank
Dexia Bank is part of the Dexia Group, which operates in Belgium, Luxembourg,
France, Germany, Austria, Spain, the United Kingdom, Portugal, Switzerland,
Sweden, Italy and the United States. Dexia Group’s three main activities cover three
areas: the funding of public utilities and financial services to the public sector and
major corporate customers, general commercial banking services, and private
banking and administrative and financial wealth management. In a recent data
mining project, Dexia Bank set out to determine what products its customers are
likely to purchase.
The objective of Dexia Bank’s first large-scale data mining project, which ran
between March and September 2001, was to select 5 target products from the
bank’s range of around 100 bank products and search for the most appropriate
target public for each of these 5 products among the bank’s existing customers.
Thus the project had to produce a list of products and a score per customer to rank
the likelihood of each specific product being bought by that target public.The
corporate data warehouse provided the information in three categories: personal
data, whether each customer had already bought any of those specific products,
and business indicators for each product, such as characteristic values (e.g. for a
mortgage loan, this would include the term, monthly repayment, etc).
While these data from the data warehouse were verified and selected, they were still
raw data and not mineable, i.e. Not readily interpretable by business people.So why
did Dexia Bank choose SAS for its data mining? “The choice was not a foregone
conclusion,” said Geert Van den Berge, Data Miner and project leader at Dexia Bank.
“Dexia Bank had already built up a certain amount of experience with SAS by using
it for other purposes, so that was definitely an advantage,” he conceded. “We were
also aware that SAS Enterprise Miner was being used very successfully in other
banks and that they were very satisfied with it, which was another strong argument,
but we still wanted to make our choice based on the results of exhaustive
One of our top criteria was the potential for communicating results to all the parties
concerned in a uniform way, so that we can be certain that everyone is working with
one version of the truth. A second criterion was assessment, i.e. the ability to
compare the performance of different models in a single graphic. By assessing the
results gained from each stage of the process, you can determine how to model
new questions raised by the previous results, and thus return to the exploration
phase for additional refinement of the data. Moreover, the SEMMA data mining
methodology used by SAS is the most natural approach, and the SAS data mining
solution addresses the entire data mining process."
SEMMA data mining methodology
SAS Enterprise Miner's "Sample,Explore, Modify, Model, Assess" (SEMMA) approach
provides users with a logical, organized framework for conducting data mining.
Beginning with a statistically representative sample of the data, this methodology
makes it easy to apply exploratory statistical and visualization techniques, select and
transform the most significant predictive variables, model the variables to predict
outcomes, and confirm a model's accuracy.
"However, we reversed the first two stages," said Geert Van den Berge. "We started
by exploring the population first, then did the sampling afterwards, since the
better insight into the population studied." Geert Van den Berge also underlined the
power of SAS data mining tools and the teamwork necessary to conduct such a
project successfully. "SAS offered us so many powerful functionalities that non-
specialist users were able to draw false conclusions.
Although the point-and-click graphical user interface of SAS Enterprise Miner is very
userfriendly, a certain amount of statistical and data mining knowledge remains
essential. There is also a definite need for teamwork throughout the entire project -
with marketing people for the business knowledge, with people who are very
familiar with the contents of
the corporate data warehouse, and with hardware specialists, because a project like
ours needs large amounts of disk space and processing power."
The sample needed to construct predictive models comprised 370,000 people, 75%
of whom were used for modeling, and 25% for a business test. Why this split? "The
business people insideDexia Bank were a little skeptical about our approach, so they
wanted to verify our results via a business test," replied Geert Van den Berge.
Association rules were used to select the 5 target products, i.e. you could say that a
customer already using product X, Y and Z was likely to be interested in a specific
target product. Mineable data were obtained by calculating useful business
indicators per customer and per product. Assessment, scoring, and finally
extrapolation followed the construction of predictive models to the 2.8 million user
group through iteration.
Apart from the need for close cooperation between all the parties involved, Geert
Van den Berge concludes that other lessons were learned with this first data mining
project at Dexia Bank. “You must have a precise, carefullyevaluated definition of the
expected deliverables, good data preparation is essential to enable experimentation
in SAS Enterprise Miner, the hardware configuration must be fine-tuned, and the
results must be interpreted in a uniform way by all the parties involved. "A
successful data mining project such as this is an excellent way to realize a high
return on our data warehouse investment," concluded Geert Van den Berge.
Using first a sample, followed by the complete corporate data warehouse
containing monthly refreshed data on 2.8 million customers, Dexia Bank set up a
data mining project with SAS Enterprise Miner. Raw data were transformed into
mineable data, a scoring model was developed and, finally, a customer’s degree of
maturity to buy a target product was determined.