Business Intelligence Successes at A Catalog and Online Retailer
Business Intelligence Successes at A Catalog and Online Retailer
Business Intelligence Successes at A Catalog and Online Retailer
American Air Liquide currently uses the ant strategy As companies expand their web of customers, BI is
to manage its operations. It produces medical gases at used to further mine the customer relationships. BI
about a hundred locations in the U.S. and delivered applies intelligent agent model in consolidating,
them to buyers. Because the daily fluctuation in analyzing, and providing access to vast amount of
electricity rates causes the fluctuation in gas data for business decision making. Major tools of BI
production cost at each location, Air Liquide include on-line analytical processing (OLAP) and
developed an AI based model that mimics ants- data mining (DM).
finding-the-best-routes for foods for its delivery
trucks. Every night, the software gathered forecasts
220
An interesting data mining application in retailing offered elsewhere. Fingerhut also had sophisticated
business is market basket analysis or recommender data mining capabilities. They were able to
systems using an item-to-item collaborative filtering efficiently analyze and extract information about
technique [10]. This technique determines which buying habits and demographics that allowed it to
products customers are likely to purchase together in pinpoint products that may interest consumers [15].
a shopping cart (a “market basket”). Once the retailer
knows that customers who buy one product are likely 3.3 BI Success
to buy others, it can recommend other products. [1]
and [2] noted the long tail phenomenon of this Fingerhut’s data mining capability helped the
recommender system in selling obscure products. The company track customers’ buying behaviors [13]. As
long tail is a colloquial name given to a product soon as intelligent agents could identify significant
distribution curve at the long tail end because the new patterns in demographics and cluster behavior,
demand for the products is low. This technique has Fingerhut sales agents could reach customers with the
helped Amazon and Netflix satisfy customers’ right offerings at the right times. After the orders
demand for obscure products that traditional stores were entered into the system, the recommender
would not stock. system searched for buying behavior of other
customers who ordered the same products, and
suggested Fingerhut to ship specialized catalogues to
3. Case Study: Fingerhut Inc. the customers. If a customer bought cookware,
Fingerhut would first follow up with specialized
In 1988, Fingerhut was the second largest catalog cooking and kitchen catalogues. Then telemarketers
mail order company in the U.S., after JCPenney. would call to follow up with other products,
Revenues reached 1.8 billion. By 1999, it entered e- especially to sell products that were in the “long-tail”
commerce and was purchased by Federated - products of low demand. The recommendation
Department Stores. After the losses of the Christmas system worked well with a good rate of repeat
sales in 1999, Federated Department store closed purchases. Through data mining sequence analysis,
down Fingerhut and liquidated all Fingerhut assets. Fingerhut found that customers who recently changed
During the summer of 2002, the remaining of their residence were likely to triple their purchasing
Fingerhut was repurchased by former Fingerhut in the 12 weeks after their move, with a peak in
executive Ted Deikel and Petters Group. By buying in the first four weeks. Their selections often
November 2002, Fingerhut was resurrected by followed a pattern— new furniture,
November 2002. In 2008, total sales reached $500 telecommunications equipment, and decorations, but
million. seldom jewelry or home electronics. The company
used this discovery to tailor a new "movers’
3.1 Methodology catalogue" to entice such customers. Thanks to its
data mining capability, its customized credit line to
This study employed several data collection methods, customers, and the effectiveness of its tele-marketers,
which include personal interviews, invited guest Fingerhut put more effort into selling furnishings and
speakers, attending press conference, and information appliances to customers who moved. Because of the
gathering from local media coverage. The interviews data mining success, by 1998 its warehouses were
with managers included two former Fingerhut stocked with an abundance of appliances and
managers in IT and Marketing department, an on furnishings [15].
campus discussion with three Petters Group’s
executives, and an interview with president of Petters The data mining capability could tell Fingerhut what
Group. customers would buy, how much they would spend,
who would be likely to pay their bills, and who
would not. For many sub-prime customers who could
3.2 Fingerhut’s Competitive Strategies not manage their spending and credit by themselves,
Fingerhut credit agents gave individual advice to
Fingerhut had two main competitive strategies. The make sure that they would not fail. This win-win
first involved locking in customers while the second strategy required a lot of dedication and effort by
involved its unique data mining capabilities. It Fingerhut.
targeted and locked in customers in the “sub-prime”
market. These customers were willing to pay a Thanks to the accuracy of behavior predictions,
premium for the extra buying power that was not Fingerhut customized how much to sell to individuals
based on the ability to pay on a one-on-one basis, and
221
therefore gained completed trust from their management wanted to emulate their successes and
customers. Whenever a customer wanted to buy more remake Fingerhut into an e-commerce player. In 1998,
than the limit allowed by Fingerhut, the credit agent it hired William Lansing to launch Fingerhut e-
first advised them to delay the purchase. After commerce business.
payment status was improved, the agent
recommended the customer to make purchases. “We In 1999, when Federated Department Stores wanted a
wouldn’t let them buy if we knew they could not pay quick entrance into e-commerce, it acquired
the bill,” said a former telemarketing manager. Fingerhut for $1.7 billion. Lansing formed an e-
commerce group, ramped up hiring, and increased
Another application of BI was the Mail Stream investment in new e-commerce startups. By 1999,
Optimization to optimize the number of catalogues Fingerhut received order fulfillment contracts with 22
mailed to individual customers. In order to reduce the companies including eToys and Walmart.com.
cost of printing and mailing catalogs, Fingerhut must
determine the right catalogues to be mailed to 3.5 The Failure of Fingerhut
individuals and the appropriate mailing time periods.
Previously, Fingerhut built an application called a With increased number of orders over the Christmas
Promotional Scoring System (PSS). PSS scored 1999 season, Fingerhut order systems encountered
Fingerhut's entire customer list with customer massive failure. eToys, the company that contracted
segmentation models for each catalogue, and based Fingerhut to fill and deliver orders, suffered numerous
on these scores, Fingerhut would create a mailing list order problems and terminated its contract with
for each catalogues. Because of similarities between Fingerhut. Other companies that hired Fingerhut to fill
catalogues, customers often received redundant orders also terminated their contracts.
catalogues to the point that the most profitable
customers received most of the 100 plus available Fingerhut revenues dipped significantly in 2000 and
catalogues. However, removing a catalogue might 2001 which caused Federated to suffer a $795 million
cause the loss of up to 80% of potential sales loss. By 2000, Fingerhut’s order fulfillment contracts
generated from that catalogue. Fingerhut needed to dropped from 22 to eight. Federated suffered heavy
analyze all the permutations of mailing over 100 losses and closed the entire Fingerhut business in the
catalogues and the impacts on profits from each spring of 2002.
permutation to determine the point where the savings
from not mailing redundant catalogues would not 3.6 Fingerhut resurrected
harm profits. With 1,400 pieces of data for each of
the 70 million customers, this was a time consuming When Fingerhut closed its doors in 2002, there was
task (22 days) for the existing Promotional Scoring again a dearth in the niche of the sub-prime market.
System (PSS). With the Mail Stream Optimization, This is due in part to the fact that K-generalists had
the processing time was reduced from to 10 hours no interest in such a market. As most customers in
and thus the decision to reduce number of catalogue the sub-prime market could not easily switch to other
mailed to customers could be done more frequently. retailers, there were many loyal customers in need of
As a result, Fingerhut was able to reduce advertising products. Capitalizing on this opportunity, FAC
expense by 6 percent [4]. Acquisition Inc. of Petters Group and Ted Deikel
completed the purchase of what remained of
Fingerhut’s BI was so successful that about 80 Fingerhut from Federated in the summer of 2002.
percent of the company’s sales came from repeat Fingerhut Direct Marketing Inc. was launched as a
buyers and viral marketing. “We’ve never done privately held company. The new catalog for the
business without some form of database marketing. fall/winter 2002 season was mailed out on November
It’s the heart and soul of this company, and we 2, 2002 to its former customers in the sub-prime
wouldn’t be in business without it” said Andy market. By December, Fingerhut’s online retail web
Johnson, senior vice president of market development site was launched, just in time to take orders for the
at Fingerhut in 1998 [13]. holiday shopping season.
3.4 Entry to Online Business and Federated Learning from the previous failure, the new
Acquisition Fingerhut did not return to many of the old markets it
entered in 1999. Taking a more conservative
Seeing the rapid revenue growth and soaring stock approach, they now only enter selected markets
prices of many other online retailers such as which they feel can be served well by their core
Amazon.com, buy.com, and price.com, Fingerhut competencies. Today Fingerhut is gradually
222
transforming itself into a more eco-oriented, self- 5. Internet Retailer, Fingerhut plans to expand e-
organizing, and self-learning company and relying on commerce in 2008, May 21, 2008,
Business Intelligence capabilities. For customers, its http://www.internetretailer.com/dailyNews.asp
direct mailing catalog sales and sales calls on ?id=26457 (accessed September 3, 2008).
paydays generate frequently generate buying swam 6. Internet Retailer, Fingerhut gets its hands on a
from its low income customers. For employees, 2.9% hike in Q4 web sales, January 21,
intelligent agents detect and communicate to other 2009, http://www.internetretailer.com/dailyNe
associates what products would sell well to each ws.asp?id=29140 (accessed March 9, 2009).
customer, credit limit, and whom to avoid. This 7. Kalakota, R., Robinson, M., e-Business 2.0:
swarm process is very effective because technology, Roadmap for Success, Boston, MA: Addison-
fashion, and taste change rapidly. In the area of Wesley, 2002.
customer service, each service agent can use his/her 8. Kelly, Kevin, Out of Control: The new
own judgment to change credit and payment Biology of Machines, Social Systems, and the
schedules if a customer cannot make their payment Economic World, Reading, MA: Addison-
rather than closing their account and turning it over to Wesley, 1994.
a collection agency. 9. Laudon, Ken, and Traver, Carol, e-commerce:
business, technology, society, Upper Saddle
As a results, Fingerhut revenues continues to grow River, NJ: Prentice-Hall, 2009, 566.
double digits in both online sales and total revenues,
10. Linden, G., Smith, B., York, J., Amazon.com
exceeding the growth of online retailing industry in
recommendations: item-to-item collaborative
each categories for the period of 2005 to 2008 [6, 7,
filtering, IEEE Internet Computing, Jan-Feb,
9].
2003, 76-80.
11. Miller, Peter, Swarm Theory, National
4. Conclusion Geographic,
http://www7.nationalgeographic.com/ngm/070
This study has validated existing ecological and 7/feature5/ July 2007 (Accessed August 20,
business intelligent models. Today online retailers 2007).
today can rely on intelligent agents in self-learning,
12. Moore, J. F., Predator and prey: a new
self-organizing, and adaptive given rapid changes in
ecology of competition, Harvard Business
the environment.
Review, No. 3, 1993, p.76.
After the failure, Fingerhut was resurrected and
13. Pearson, David, Marketing for Survival, CIO
magazine, April 15, 2003.
transformed into a humble company. It now focuses
only on businesses within its core competency that 14. Phan, Dien, D; E-business success at Intel: an
exploit business intelligence. There are plenty of organizational ecology and resource
opportunities to exploit the network relationships of dependence perspective, Industrial
its business ecosystem. Opportunities also exist in the Management & Data Systems, Vol. 102, No. 4,
use of Web content and blogs to generate swarm 2002, pp. 211-217.
behaviors for special discounted promotion products. 15. Phan, Dien, D.; Chen, Jim, and Ahmad, Sohel,
Lessons learned from the failure of a major
retailer, Information Systems Management,
5. References Spring 2005.
16. Slickdeals.com,
1. Anderson, Chris, The Long Tail, Wire, 10,
http://forums.slickdeals.net/showthread.php?t=
2004, 171-177.
1064040&page=11 (accessed December 22,
2. Brynjolfsson, E., Hu, Y., Smith, M., From 2008).
Niches to Riches: Anatomy of the Long Tail,
17. Tarasewich Peter, and Patrick R. McMullen,
MIT Sloan Management Review, Vol. 47, No.
Swarm Intelligence: Power in Numbers,
4, Summer 2006, 67-71.
Communications of the ACM, August 2002,
3. Byrne, John, Management by Web, Business Volume 45 Issue 8, pp. 63-67.
Week, August 28, 2000, pp. 84-96.
18. Uptal, D., and Soltysinski, K, Coveted or
4. Decision Intelligence, Fingerhut direct Overlooked? The Psychology of Biddings for
marketing case study, 2000, http://www.dii- Comparable Listings in Digital Auctions,
online.com/case-studies/fingerhut-direct- Marketing Letters, 2001.
marketing-case-study.html (accessed March,
16, 2009).
223