Expert selection of marketing databases Hilary Drenth, Gwyneth Tseng & Anne Department of Library &

Morris Studies, Loughborough University of Technology Information

The selection of online sources to answer marketing queries was investigated by interviewing fifteen professional intermediary searchers. Results indicate an overall performance ranking for the hosts which were most used by these intermediaries. Selection of databases was influenced by cost, type and level of coverage, and the geographical focus of a database. The features and services offered by a host had little influence on database selection. Frequently-used databases were assessed for particular types of marketing information. The results were used to build a prototype expert system which recommends databases for marketing queries. The approach seems feasible, but more research into intermediaries’ expertise is needed.

a

Introduction

Regular online searchers are aware that state-of-the-art online retrieval systems are seldom easy to use effectively. Concerns about the complexity of technical protocols, search languages, indexing systems, charging mechanisms and so forth are very widely discussed. New systems proliferate, while existing ones change and develop to maintain their competitive edge. All this makes life complicated for the professional search intermediary, and bewilderingly confusing for the untrained or infrequent searcher. In response, some sectors of the online industry and of the research community have applied artificial intelligence (AI) techniques to making online systems more accessible to

end-users.

One technique, popularly known as ’expert systems’, attempts to replicate the problem-solving behaviour of a human expert, in this case the professional intermediary. Such a system uses a ’knowledge base’ embodying the strategies and facts that an expert uses to solve problems. Choosing a suitable database for specific queries is one of the major problems for end-users with wide-ranging information n~eds~t92~° The number of databases grows yearly; keeping track of available sources and evaluating them is a major task, even for information professionals. Search outcomes can be highly dependent on the suitability of sources used, and their skilful selection requires detailed knowledge and sound judgement.

24

Downloaded from bir.sagepub.com at SAGE Publications on December 5, 2012

25

ABOUT THE AUTHORS

Hilary

Drenth is

prototype

research assistant on Loughborough expert system for business database selection. a

University of Technology’s project

to

develop

a

Gwyneth Tseng is a lecturer in the Department of Library and Information Studies at Loughborough University of Technology, teaching courses in information retrieval and information technology. Anne Morris is a lecturer in the Department of Library and Information Studies at Loughborough University of Technology, teaching courses in expert systems, human factors and information technology.

A study at Loughborough University in 1987 examined the feasibility of an expert sys-

approach to selecting online databases(3,4). Various criteria used by intermediaries selecting sources of company information were identified and incorporated into a prototype. Favourable reactions to the prototype suggested that this approach might indeed be one way to tackle the database selection problem. This article describes the findings of a subsequent studyC5) that focused on the selection of databases providing marketing information. Experts were interviewed to establish which hosts and databases they used most, why, and for what purposes, and to identify the criteria that were used to choose between sources. The study was carried out in summer 1989 with 15 intermediaries, from both public and in-house information services in the UK, acting as experts. Intermediaries were identified through the Marketing and Media section of U1ho’s llho in the UK lnfonnatem

for

°

tion World(6), UK Online Search Services(7), and word-of-mouth contacts, Six of the inhouse intermediaries who participated in the study worked in advertising agencies.

Usage of hosts for marketing queries First of

all, intermediaries were asked about the hosts that they used as sources of marketing information. Reuter ’~‘e~tli~e, PROFILE, Data-Star, Dialog and PFDS were used most (see Table 1 below). Coverage, cost, and the ability to ’pay as you go’ were the most important factors in determining whether they subscribed to a service. Personal recommendations and previous experience were also important. Large bibliographic services were not generally used by advertising agencies; the only agency that did subscribe to them provided an information service to clients. Agencies tended to use Reuter Textline and PROFILE as broad sources of information, and MAID and Target Group Index (TGI) as sources of market research and audience research respectively. Several intermediaries mentioned that Textline’s summary, historical coverage was complemented by PROFILE’s full text, more up-to-date coverage. In order to determine which aspects of using a service influenced the choice of marketing databases, intermediaries were asked to rate the hosts they used. Hosts were rated for: * * *

the range of features offered of use

ease

processing speed

Downloaded from bir.sagepub.com at SAGE Publications on December 5, 2012

26 *

the frequency of loading updates

*

the administration of update information

*

the quality of written documentation the quality of online documentation the quality of help desk services ease of calculating search costs.

*

* *

The scores from rating hosts for each of these criteria were combined to give a ’score of scores’ for h®sts(5>9 as shown in Table 1.

Table 1.

Usage and scores for hosts

Dialog attracted high scores for all criteria, apart from its help desk service. Loading a large number of databases perhaps militates against an in-depth help service, although Data-Star received its highest rating for this. Dialog was the only service to score highly for ease-of-use and for the quality of written documentation. Like PROFILE and PFDS, Data-Star scored well overall. Its lowest score was for of calculating search costs; indeed, its complex price structure discouraged one intermediary from using the service if there was a suitable alternative. Since the study was carried out, Data-Star have improved their online reporting of costs. PR~F°II,E’s lowest scores were for the range of features offered, documentation and help desk service. The overall score for PFDS was also affected by low scores for documentation. In fact, scores for the quality of documentation, both written and online, were generally ease

Downloaded from bir.sagepub.com at SAGE Publications on December 5, 2012

27 much lower than for other criteria, suggesting that users’ serious consideration by most hosts.

requirements in this area need

Though widely used, Reuter Textline did not score highly. Several intermediaries felt that the host’s apparently simple command language reduced the level of precision possible. Like PROFILE, searching on Reuter Textline was thought to require more, rather than less, skill than searching with fuller command languages. ’

Other services such as Kompass Online and Telesystemes Questel were only used at few sites, making it difficult to draw any general conclusions about them. Both MAID users rated the service very highly; its range of sources, overall quality, and high up-front charge made it their automatic first choice as a source of market research. As TGI is a numeric host, it was not rated for its range of features for searching. It was considered difficult to use, but scored well for most criteria, especially for its help desk service. a

Overall, it appeared that the features and services offered by a host had little influence on the selection of databases. As one intermediary put it: &dquo;You can get round features.&dquo; Familiarity with a host was important, however. Selection of online databases for marketing

queries

Information about selecting online databases to

answer

a

marketing queries was obtained

by: * *

*

general questions about retrieving marketing information discussing specific marketing queries rating databases against various criteria.

The main types of

marketing

query handled

by intermediaries were

about

specific

markets, and about companies - as key players, potential clients, or potential targets for

acquisition. c

Factors

selection

One factor influencing the choice of a database,

regardless of the type of query, was cost. services were affected as much as those working in public ones. Cost considerations restricted the sources available and prevented ’looksees’. Connected with this, some databases were characterised as being ’good value’ not necessarily the cheapest, but likely to provide the information required, and thus save valuable time. Intermediaries

working in private

®

Another general factor influencing selection was the type of information required by the client raising the query. Intermediaries reported that some clients &dquo;like raw data&dquo;, &dquo;like reams&dquo;, G‘will read cv~ryth~rlg99, while other clients wanted summaries. The client’s level of interest, budget, or importance to the information service influenced the completeness of a search.

queries about specific markets, the key factors apart from cost were the market geographical area, and the type of information required - overviews, detailed items or market research. The key factor in selecting sources of company information was the type of information required: directory, financial (with or without full accounts and details of shareholders), news of activities (including acquisitions), and brokers’ research. The press was the main source of information about companies’ activities. The For

area,

Downloaded from bir.sagepub.com at SAGE Publications on December 5, 2012

publications used depended on the sector in which the company whether summaries or articles were required. To

sum

up, the main factors

affecting

was

involved, and

the selection of databases for

marketing

queries were: *

cost

*

type of coverage required (eg market research v news)

*

level of coverage required (eg summary v full text) geographical focus.

*

The earlier study at Loughborough into company information(3,4) identified the following factors: *

currency of material

*

types of source covered quality of abstracts/text

* *

intermediary’s familiarity with the database

*

search fields available

*

cost.

In the recent study, the types of source covered - articles, reports, abstracts, news items - were clearly an influence on database selection, and depended on the client’s requirements. Although not stated, familiarity with a particular database was also an influence, as demonstrated by the use of a few, favourite sources (see below). Cost too, was an important factor. The currency of material was particularly noticed by intermediaries in advertising and the media who used press sources heavily. It was not mentioned as an influence, however. The tendency was to run searches on both PROFILE and Reuter Textline and so ensure that no items were missed. The quality of the information retrieved was only mentioned with regard to a handful of sources, and was generally assumed to be acceptable. Search fields were mentioned for their ~sef~xl~ess in certain searches, but were rarely cited as the reason for choosing a database. Database used

The

majority of intermediaries used a core of six or seven online marketing queries. The sources used ®st frcqucntly were: * * * *

* *

* *

*

Predicasts (PROMT or the PTS group) Infomat ICC Key Note Market Analysis UKNEWS file group on PROFILE UK & Western Europe section of Reuter Textline Financial Times (FT))

McCarthy Press Cuttings Service (MCC) Marketing & Media section of Reuter Textline ADS file group

on

PROFILE

Downloaded from bir.sagepub.com at SAGE Publications on December 5, 2012

sources

to

answer

29 *

* *

Retailing section of Reuter Textline ICC British Companies database Kompass UK Companies File.

Predicasts files and Infomat were heavily used for both consumer and industrial market information. Intermediaries described Predicasts as suitable for queries with an American focus and Infomat as suitable for queries with a British or European focus, but in practice intermediaries invariably used both. They were both judged good sources of statistics and summary information; Predicasts was considered better than Infomat for very specific product information as the index was more detailed. Infomat was very good for information on new products. Financial Times Business Reports (FTBR) were also viewed by some intermediaries as a good source of information about industrial and service markets. A comment from one intermediary suggests that the reliance on Predicasts files and Infomat as sources of industrial market information is due to the paucity of this type of information online. Industry-specific databases such as PIRA and RAPRA were only mentioned by one intermediary. Possibly their names disguise the range of products and information that they cover. Many intermediaries used trade associations as starting points for searches in this area, or used printed indexes.

ICC Key Note Market Analysis was the favoured, broad source of consumer market research, because of its strong UK bias and coverage of unique sources. It was also used to obtain background information on companies. Euromonitor Market Direction was used by some intermediaries looking for European coverage. For those intermediaries access to MAID, it was the automatic choice for information on consumer markets because of its comprehensive coverage and the high up-front charge for the service. Some intermediaries always used an index such as the Marketing Surveys Index (MSI) to identify market research. Press sources such as uknews, the FT and MCC on PROFILE, and Reuter Textline’s UK & Western Europe section were used as broad sources of background information on markets, industries and companies. P~C)FIL,E’s ads file, and the Marketing & Media and Retailing sections of Reuter Textline were particularly used by intermediaries working in advertising. The uknews file was very highly used as a general source of information, and as an index to the trade press. The ~°T and MCC were also regular sources of information, the FT because of its quality, and MCC for its unique sources. Reuter Textline was particularly valued for its extensive coverage.

who had

Table 2 below shows how intermediaries assessed frequently-used sources of market information.

Downloaded from bir.sagepub.com at SAGE Publications on December 5, 2012

30

Table 2.

Top sources of market information

The ICC British

Companies database with its access to Financial Datasheets was the of British company financial information, though Jordanwatch was also used in conjunction with Infocheck. The reasons given for using the ICC British Companies database were that it provided full information, was easy to access, had wide coverage, was well-presented (on Dialog) and was a source of shareholder listings. Infocheck was considered to be a unique source of evaluative information. Predictably, Kompass UK Companies File was the first choice for identifying UK producers, though it was considered difficult to search. Dun & Bradstreet’s Key British Enterprises was used by intermediaries with access to PFDS to identify key players, as it allowed sorts on sales figures. These intermediaries used Europe’s Largest Companies when looking for European market leaders. D&B International Dun’s Market Identifiers, Kompass European Companies File and Hoppenstedt files were also used as sources of information on European companies, though Hoppenstedt files were considered relatively expensive. preferred

source

The prototype databases adviser

The results of the study were used to design a prototype expert system that recommends databases for answering marketing queries. The prototype was implemented using the LEONARDO expert system shell (Creative Logic Ltd). It handles the following types of query: information about specific markets and industries, product news, background information on companies, and information about product suppliers.

The prototype recommends databases on hosts that the user can access. It can also recommend suitable databases on other hosts if required. This is useful both for training, and for current awareness. Recommendations are based on the criteria used by intermediaries for different types of query, including subject coverage, the type and level of coverage, geographical coverage, and, for directory information, company size. Depending on the number of databases selected by applying the various criteria, the recommendations are refined using the cost of databases (if this is important to the

Downloaded from bir.sagepub.com at SAGE Publications on December 5, 2012

31

user), the ’good value’ rating (if given), the user’s preferred hosts (if any), and the hosts’ ease of use. The prototype’s recommendations appear sensible, but the current refinement process needs to be further developed. 0

Conclusions

Information to support marketing activities draws on a very wide range of sources - online, in-house, and external organisations - and covers many different types of information. Consequently, expertise in this area is not easily acquired, and tends to focus on a specific aspect or specific sources. An expert database selection aid has potential for helping intermediaries to widen their information horizons while staying within budget, as well as in assisting end-users in identifying online sources either for their own use or for referral to a professional searcher. Further work is needed to explore intermediaries’

expertise in depth, in the following

areas: ~

the criteria that experts use to distinguish between databases at example, the descriptors available

*

the relationship between database selection and appropriate search strategies

*

the types of query not amenable to an online solution; an expert database selection aid must incorporate this type of knowledge, if only to save the user time.

a

detailed level - for

Supporting studies examining information use in various business sectors are also required in order to determine the scope of an expert system for business database selection. Work at Loughborough continues with the development of a prototype expert system that assists selection of databases providing company information. The study is funded by the British Library. Further information: e Morris, Department of Library & Information Studies, Loughborough University, Leicestershire LE11 3TU. Tel. 0509 223073. ®

1.

References

Norton, R.A. and Westwater, J. 1986. Starting end-users.

Aslib Proceedings,

38(11/12):381-388. 2.

1989. Profiling the end-user: a study of the reference needs of end-users on Telebase System, Inc.’s Easynet. In Proceedings of

Dyckman, L.M. and O’Connor, B.T.

the 10th National Online Meeting, New York,

May 9-11, 1989, 143-152.

3.

An Expert System for Online Business Database Selection. Newham, G.H.W. Master’s dissertation, Loughborough University of Technology, 1987.

4.

Newham, G., Tseng, G. and Morris A. 1988. Choosing a business database: the expert approach. Business Information Review 5 (1): 27-41.

Downloaded from bir.sagepub.com at SAGE Publications on December 5, 2012

32 5.

Pugh, H.J. An Expert System for Selecting Online Marketing Databases. dissertation, Loughborough University of Technology, 1989.

6. Who’s Who in the

Master’s

UK Information World. London, TFPL, 1988.

7. UK Online Search Services.

London, Aslib, 1987.

The authors would like to thank all those intermediaries who very kindly gave their time to participate in this study.

Downloaded from bir.sagepub.com at SAGE Publications on December 5, 2012