IMPROVING THE CUSTOMER INTELLIGENCE WITH

CUSTOMER ENTERPRISE CUSTOMER MODEL

Domenico Consoli, Claudia Diamantini and Domenico Potena

Dipartimento di Ingegneria Informatica, Gestionale e dell’Automazione,

Universit´a Politecnica delle Marche, Via B. Bianche, 60131, Ancona, Italy

Keywords:

External Information, Customer Opinions, Customer-Centred Enterprise Information System, Customer Re-

lationship Management, Web 2.0, Text Mining.

Abstract:

Customer can be proﬁtably considered as an enterprise strategic asset. So, it is very important to deﬁne a

bi-directional communication channel between the customer and the enterprise itself. In this work we propose

a model named Customer enterprise Customer (CeC), that continuously hears customer opinions about enter-

prise products/services and behaves accordingly. CeC exploits the inherent nature of the growing so-called

web 2.0, where users spontaneously join and spend time in sharing their reviews. The CeC model ﬁrst collects

and analyses customer opinions, then reports dysfunctions about the product/service to competent ofﬁces for

giving feedback to customers, either by making the necessary improvements or by answering customers. The

proposed model crosses all internal business functions, from design to production, and it is placed on the top

of a customer-centred Enterprise Information System. In this work an overview of CeC model is given.

1 INTRODUCTION

Customer opinions constitute a gold mine for generat-

ing high value added information for making strategic

decisions (Hagel and Rayport, 1999; Fabris, 2007).

The gathering of opinions is classically performed

by ad-hoc interviews (often by phone or by email),

that are based on pre-structured questionnaires. This

approach, besides to be expensive, is limited. In

fact, a customer expresses her opinions by using nat-

ural language better than by answering to a structured

question (Bolasco et al., 2005). In order to overcome

this limit, an enterprise may take advantages of the

huge amount of information available over the Inter-

net in form of forum, blog, wiki, and other web 2.0

tools.

It is, in fact, evident as the coming of web 2.0

promoted the birth of a sharing business philosophy

and stimulated conversations and exchange among

people (Wenger and Snyder, 2000). Customers ex-

ploit web 2.0 tools for both expressing their opinions

about a product and suggesting solutions for improv-

ing it. The enterprises often encourage exchange of

opinions, by making available virtual communities,

e.g. Italian Nikon’s Camera forum, where people re-

view Nikon products (http://www.nital.it/forum/), the

blog on Benetton products (http://benettontalk.com),

and so on. Furthermore there are various web sites

that collect and make free available customer reviews

(Cho et al., 2002): epinions.com, cnet.com, com-

plaints.com, planetfeedback.com, ecomplaints.com,

ciao.it, dooyoo.it.

To capitalize customers opinions is very important

for an enterprise, both for the improvement of prod-

ucts and for the reinforcementof the customer loyalty.

The customer will be motivated to be loyal if the en-

terprise shows a strong attention to her needs and her

identity.

It is estimated that for each customer which makes

a complaint, there are 4 to 26 customers do not claim

even if they are dissatisﬁed. Each dissatisﬁed cus-

tomer, on average, expresses her mood to 10 people.

So, in the best case, behind each complaint there are

40 potential skeptical customers. Nevertheless, it is

interesting to note that if a customer who complained

sees the enterprise accepting the review and solving

the related problem, then her propensity to purchase

from the same company increases of 3 times, while

for the customer who has not claimed, the propensity

to purchase increases by 6 times (Lombardi, 2006).

In the last years the enterprise informative sys-

tem have been designed in a customer-centric way to

gather more accurate and detailed information about

customers. Enterprise information System (EIS) feels

323

Consoli D., Diamantini C. and Potena D. (2008).

IMPROVING THE CUSTOMER INTELLIGENCE WITH CUSTOMER ENTERPRISE CUSTOMER MODEL.

In Proceedings of the Tenth International Conference on Enterprise Information Systems, pages 323-326

DOI: 10.5220/0001733503230326

 SciTePress

Figure 1: The Customer-enterprise-Customer model.

the necessity to have a model that gives, in real-time,

the situation of the market and, in particular, that

makes a continuous monitoring of the customer opin-

ions.

In this paper a new model named Customer en-

terprise Customer (CeC) is proposed, that is aimed to

continuously hears customer opinions about the prod-

uct/service and behavesaccordingly. To this end, CeC

ﬁrst collects and analyses customer opinions freely

available over Web 2.0 tools, then reports dysfunc-

tions about the product/service to competent ofﬁces

for the necessary improvements or for answering cus-

tomers.

Differing from classical operational CRM, CeC

model is not limited to Customer Area (e.g. mar-

keting, sales and post-sales), but it is introduced as a

frameworkthat crosses all internal business functions,

from marketing to design and production. So, we

placed our proposal on the top of the Enterprise Infor-

mation System. Implementing a CRM model the en-

terprise is able to answer a customer only if her ques-

tion is already in the CRM knowledge base. On the

other hand, CeC addresses the customer opinions to

competent ofﬁces analysing from time to time the un-

structured texts containing opinions themselves. This,

of course, introduces various issues to be consid-

ered, mainly regarding Natural Language Processing

and Mining. In this work we brieﬂy discuss these

and other technical issues, concentrating on giving an

overview of the proposed model.

The CeC model forms part of enterprise models

introducing the concept of prosumer (Tofﬂer, 1990),

where it is emphasized the relationship between pro-

ducer and customer. In this vision, the customer plays

an active role in the enterprise, in particular he/she

participates as a co-producer or as a consultant. Ex-

amples of co-producer are customers buying furniture

at the IKEA and then ﬁnishing the productive process

by assembling it at home. An example of consultant

prosumers is given in the Internet site of the “Fiat

500” (500 Wants You - www.ﬁat500.com), where each

visitor expressed her/his creative contribution to the

design of the new car.

The following Section is devoted to detail the CeC

model, brieﬂy discussing its main parts. Section 3

ends the work.

2 THE CeC MODEL

In Fig. 1, we show the Customer - enterprise - Cus-

tomer model. The model is divided into three phases:

sensing, mapping and actuation. Sensing consists of

crawler agents that scan the web 2.0 tools (e.g. web

site, forum, blog and so on) for ﬁnding and gath-

ering opinions. Then, since customer expresses her

opinions in unstructured text, we have to prepare and

process opinions by text mining algorithms in order

to extract useful complaints. In the mapping phase,

complaints are mapped and routed to speciﬁc compe-

tence centers (CCs), that can be departments, inter-

nal experts, external consultants, and others groups of

people that are competent in the problem expressed

in the complaint. The mapping is described by the

Complaint to Competence Matrix (CCM), e.g.:

• Complaints: unfocused photos, distorted images

→ CC: Optic Department;

• Complaints: LCD monitor does not shown photos

→ CC: Electronic Department.

Finally, the goal of the actuation phase is to react to

stimuli, either by making changes to the product or by

answering to customers (messaging). In this second

case, the enterprise becomes, in a peer-to-peer web

2.0 vision, an actor as a customer.

ICEIS 2008 - International Conference on Enterprise Information Systems

324

2.1 Sensing

The sensing module is composed by a set of crawler

agents specialized in different protocols (http, https,

pop3, imap4, nntp), that are responsible to inspect

and retrieve information, respectively, from web sites,

blogs, chats, e-mails, newsgroups and so on. Each

agent can be conﬁgured with policies to extract only

texts (not structured, semi-structured or structured)

with advanced techniques of NLP (Natural Language

Processing). The various crawler agents are usually

coordinated by a crawler manager which deﬁnes poli-

cies shared by all agents (Boldi et al., 2004). Crawlers

could work in parallel exploiting the intrinsic nature

of the network.

The customer reviews may be referred to positive

or negative opinions. The CeC model has its principal

focus on negative opinions. So, in order to reduce the

searching space, intelligent agents can be exploited

(Chan, 2008; Jansen et al., 2006). Such agents im-

plement appropriate strategies and heuristics for ana-

lyzing only documents containing negative opinions.

In order to distinguish between positive and negative

opinions we can exploit techniques from the affective

computing ﬁeld (Grefenstette et al., 2004), for the in-

dividuation of emotions into a text.

In our model, collected reviews populate the

Opinion Warehouse, that forms the corpus to be anal-

ysed by text mining algorithms in the next mapping

phase.

Some examples of customer reviews are:

Review n.1: The edges of the photograph

were faded;

Review n.2: The ﬁgures in the foreground

aren’t sharp.

2.2 Mapping of Complaints to

Competence Centers

The main part of this phase is the Text Mining. The

goal of the text mining part is the application of al-

gorithms for pre-processing and for classiﬁcation of

customer opinions. Since opinions are written in Nat-

ural Language, we need speciﬁc pre-processing tech-

niques: elimination of stop-words (articles, conjunc-

tions, prepositions), division of the phrases into single

words, identiﬁcation of different parts of speech POS

(nouns, verbs, adjectives), lemmatization, and numer-

ical representationof text data (Berry and Castellanos,

2007). In the pre-processing phase it is important the

use of the lexical ontology (e.g. WordNet). The ontol-

ogy containsa set of explicit assumptions that concern

the meaning of the words.

Table 1: An example of Complaint Competence Matrix.

Complaints

distorted unfocused LCD

images photos

Competence Mechanic Dep. 0.15 0.35 0.00

Centers Optic Dep. 0.75 0.05 0.00

Electronic Dep. 0.10 0.60 1.00

Classiﬁcation model allows us to associate an

opinion with one or more complaints on the basis of

measures of semantic similarity. For instance, the

opinions of the previous example can be classiﬁed

as ”The photos are unfocused”. In order to design

the classiﬁcation model, information about both prod-

ucts/services and the internal organization of the en-

terprise has to be taken into account. To this end can

be exploited a Business Ontology formed by two sub-

ontologies: Product Ontology and Enterprise Ontol-

ogy. The former deﬁnes in detail all products, their

components, how they are produced and the related

competence centers. The latter is a conceptualiza-

tion of the whole enterprise: functions, business pro-

cesses, competence centers, and so on.

The output of the data mining process are com-

plaints, that are used in input to the Complaint Cen-

ter Matrix (CCM), a model associating a complaint to

one or more Competence Centers (CCs) (see example

in tab. 1). In order to build the CCM, we can exploit

information contained in the Business Ontology. The

values of CCM elements represent the weight of a De-

partment/CC in facing the complaint inherent prob-

lem.

At this point, a message containing the compliant

and related reviews is automatically routed to indi-

viduated CCs. In the case of enterprise with a lot of

CCs, we can improve the effectiveness of the routing

by sending messages only to most competent centers,

that is CCs with CCM values over a given threshold.

In the example of Fig. 1, we see that the Optic De-

partment is the most competent center for the problem

regarding the “distorted images”.

2.3 Actuation

The actuation phase is the less automatic one, cause

the decision making process is mainly performed by

people. In this phase CCs exploit the complaint mes-

sages for the improvement of products/services or for

answering customers. In both cases, the enterprise

give a (indirect or direct) feedback to the customer.

As a matter of fact, the improvementof a product is an

indirect message, communicating that the enterprise

acknowledged the customer complaints. On the other

hand, the enterprise may direct answer customer over

IMPROVING THE CUSTOMER INTELLIGENCE WITH CUSTOMER ENTERPRISE CUSTOMER MODEL

325

the same channel used by the customer for express-

ing her opinions. For instance, if opinions was found

over a public forum, the enterprise participates in the

forum itself as any other user.

The knowledge of the web sites dealing with en-

terprise products and services, can be also exploited

for improving the effectiveness of communications

towards the market. As a matter of fact, the message

is addressed to interested people.

Furthermore, we introduce a feedback line from

CCs and CCM, that allows managers to dynamically

adjust the weights of the CCM, improving the effec-

tiveness of the mapping and the routing (dotted line in

Fig. 1).

3 CONCLUSIONS

In this work we propose the CeC model, a customer-

centred enterprise information system model, aimed

to exploit customer opinions for enriching the enter-

prise internal knowledge.

The main characteristics of this model is that the

enterprise just limits itself to observe the web and,

in particular, communities of customers discussing

about enterprise products or services. Our model ex-

ploits the inherent nature of the growing so-called

web 2.0 tools, where users spontaneously join and

spend time in sharing their reviews. The CeC model is

aimed to ﬁnd, collect and analyse opinions, to react to

stimuli and to send feedback to customers. So, while

the enterprise plays a passive role in the discussions,

it becomes participative when it send return messages

to market, either by making changes to products (in-

direct messaging) or by direct answering customers

by using the same web 2.0 tool.

The core of CeC is formed by the Sensing and

the Mapping phases, where customer opinions are se-

lected and analysed by tools derived from NLP and

Text Mining areas. In these phases, after an appro-

priate pre-processing, unstructured opinions are clas-

siﬁed as complaint classes, and then routed to centers

most competent to respond to the complaint.

The CeC model is part of a more wide projet,

that, at this moment, is at an embryonic state. We

implemented only some crawler agents and designed

the opinions warehouse for collect customer opinions

about photo cameras and holiday villages domains. In

next works we want to follow two main directions. In

the former, we will use the collected warehouse for

mapping phase, studying Text Mining algorithms and

techniques most suitable for the speciﬁc problem. To

this end, it will be needed to also design the business

ontology.

In the latter direction, we will investigate tech-

niques for design intelligent crawlers that are able

to distinguish both useless from informative sources,

and negative from positive customers reviews.

REFERENCES

Berry, M. W. and Castellanos, M., editors (2007). Survey

of Text Mining II: Clustering, Classiﬁcation, and Re-

trieval. Springer.

Bolasco, S., Canzonetti, A., Capo, F. M., della Ratta-

Rinaldi, F., and Singh, B. K. (2005). Understanding

text mining: A pragmatic approach. In Sirmakessis,

S., editor, Knowledge Mining, volume 185 of Stud-

ies in Fuzziness and Soft Computing, pages 31–50.

Springer Verlag.

Boldi, P., Codenotti, B., Santini, M., and Vigna, S. (2004).

Ubicrawler: a scalable fully distributed web crawler.

Softw. Pract. Exper., 34(8):711–726.

Chan, C.-C. H. (2008). Intelligent spider for information

retrieval to support mining-based price prediction for

online auctioning. Expert Syst. Appl., 34(1):347–356.

Cho, Y., I.Im, Hiltz, R., and Fjermestad, J. (2002). An anal-

ysis of online customer complaints: implications for

web complaint management. In Proceedings of the

35th Annual Hawaii International Conference on Sys-

tem Sciences, pages 2308–2317. IEEE.

Fabris, G. (2007). Sua maest´a il consumatore, nuovo re dei

mercati. In L’impresa Marzo 2007. Il Sole 24 ore.

Grefenstette, G., Qu, Y., Shanahan, J. G., and Evans, D. A.

(2004). Coupling niche browsers and affect analy-

sis for an opinion mining application. In In Pro-

ceedings of RIAO-04, 7th Int. Conf. on Recherche

d’Information Assist´ee par Ordinateur, pages 186–

194.

Hagel, J. I. and Rayport, J. F. (1999). Creating value

in the network economy, chapter The coming battle

for customer information, pages 159–171. The Har-

vard Business Review Book Series. Harvard Business

School Press, Boston, MA, USA.

Jansen, B. J., Mullen, T., Spink, A., and Pedersen, J.

(2006). Automated gathering of web information: An

in-depth examination of agents interacting with search

engines. ACM Trans. Interet Technol., 6(4):442–464.

Lombardi, G. (2006). Capire i bisogni dei clienti. Corriere

delle Opere n. 191.

Tofﬂer, A. (1990). Powershift: Knowledge, Wealth and Vio-

lence at the Edge of the 21st Century. Bantam Books.

Wenger, E. and Snyder, W. M. (2000). Communities of

practice: The organizational frontier. Harvard Busi-

ness Review, pages 139–145.

ICEIS 2008 - International Conference on Enterprise Information Systems

326