Report Writing Assistant
Professor John M. McCannFuqua School of Business
Duke University
To improve the role of systems in marketing, the system is extended so that it is involved in all aspects of the marketing process of analysis, planning, execution, and control. This approach is inspired by microprocessors. Just as the microprocessor is having a major impact on marketing data and thus impacting the practice of marketing, it is also having an impact on the types of systems which can be built to support marketing managers. In particular, knowledge-based systems can be added to the more traditional information processing systems. Advanced word processors or desktop publishing software can be integrated with these systems to enhance the report writing aspects of marketing.
This section describes the report writing activities of a hypothetical marketing manager and discusses the role of knowledge-based systems in this process. In order to develop such a system, a "science of report writing" is required. An initial attempt at such a science is presented.
The Report Writing Assistant is an aspect of Marketing Gate which would support the development of a marketing report. The Report Writing Assistant is designed to augment the normal analysis and writing activities of managers by relieving them of some of the analytical and composition tasks. This section explains how a fictitious marketing manager would use the computer to generate a report with the Report Writing Assistant.
The following is a description of how the manager uses the MMIS in the writing process:
I use my MMIS to understand our various markets. We have developed a library of capsules in our Metaphor system which allows me to get different views of the market. When I get a view on the screen, I make observations about it and usually draw a conclusion. I loop through these capsules until I have made enough observations and drawn enough conclusions to tell the brand's story.
At this point, the brand manager wishes that there was some way to make the process more efficient. After all, he is using a library of standard capsules to produce standard data views, and there are only a limited number of different observations and conclusions which can be made from any one data view. One begins to wonder if the same logic which led to the creation of the capsules could be used to further the analysis.
The manager goes on to describe his use of the MMIS:
I usually know what I want to say in my report ... I have a general overview of the message I want to communicate. My problem is operationalizing this general overview. For instance, I may know that I want to make a statement about the category's life cycle position in a local market. The first problem involves getting the necessary data or graph on the screen to allow me to make the proper life cycle classification. Then I must examine the data and make the classification; sometimes this is easy, sometimes it is hard.
After performing these steps, our manager has determined the life cycle position and assembled the supporting evidence. Additional analyses may be performed similar to the life cycle analysis; statements may be made about the brand's history of share changes and the degree of competitive pressure being brought to bear on the brand. Each of these analyses would involve running the appropriate capsules and making correct observations and conclusions. Then the manager is ready for the next phase of the work.
Now I must get these conclusion and the supporting evidence into a report. This sounds like a simple process, but it is time consuming because of the sheer magnitude of our category and the number of markets. And, we have to do this over and over again.
This diagram illustrates the process our manager has been following. He started at the top, thinking about the purpose of the report, perhaps seeing a congratulation from his manager. Then he proceeds in a clockwise manner: thinking about the presentation, selecting the analyses to perform, selecting the appropriate computer system, identifying the data, doing the computer work, and writing the report.
This is an "analysis followed by writing" process which people have utilized for centuries; it is taught in most writing courses in most universities. But, there might be a better way to approach this problem given the nature of marketing in the consumer packaged goods industry.
Our manager begins to think further about a different approach to this analysis and report process.
Isn't there some way that we could reverse the process? Let me think about the report and have the system do the necessary analysis to make the points I want made. Let me tell it to write a paragraph about my category's life cycle position. The system would know how to write this paragraph: what data to use, how to get the data out of the database, what points should be made, what kind of exhibit best illustrates the point, and what conclusion can be drawn. Not only would it know how to do it, it would do it for me whenever I told it to.
Our manager is dreaming of a system which would allow him to reverse the direction of processing in the above diagram. Let him start at the top, but proceed counter-clockwise. Let him select the pages of the report and have the system write them and perhaps prepare the report and the presentation. The manager thinks about the goals and the presentation, and the computer writes the report. This computer could be thought of as a Report Writing Assistant.
In the previous section, we used an interview with a hypothetical marketing manager to identify the marketing analysis and reporting process, and to describe a way to improve the process. This section describes an architecture for such a Report Writing Assistant.
An assistant is someone who works along with the manager, using the same tools and doing as instructed. The manager can off-load some of the tasks to the assistant, or perform them himself. The primary tool for the analysis phase is the MMIS; the manager uses the MMIS to obtain the appropriate data views which allow the observations to be made and conclusions to be drawn. Therefore, the Report Writing Assistant should use this same tool, i.e., use the MMIS just as the manager would use it. This means that the Report Writing Assistant must interface with the MMIS. In a Metaphor world, the Report Writing Assistant would need to know how to execute and interpret capsules; in an Express world it would work with reports; in an Acustar world, with Acu-Execs.
Since each of these items (capsules, reports, and Acu-Exec) are procedures for obtaining a particular view of the data, we will adopt a general term of "DATA VIEWER" to denote the MMIS procedures for doing standard analyses. A manager would have a library of DATA VIEWERS available for doing analysis. For most applications, the MMIS is nothing more than the collection of DATA VIEWERS.
Since managers use analysis and marketing knowledge to apply and interpret each DATA VIEWER, the Report Writing Assistant must contain similar knowledge. This knowledge will be contained in a Marketing Management Knowledge System (MMKS), which is linked to the MMIS via the computer's operating system. Just as the manager interacts with a program like the MMIS using the operating system, so does the MMKS. This allows the MMIS to remain unaltered in the sense that both the manager and the MMKS can issue commands to it and receive its output. In the default case, the operating system gets the input from the manager via the keyboard and/or mouse and sends its output to the screen and/or the printer. When the Report Writing Assistant is substituting for the manager, the MMIS gets its commands from and sends its output to the MMKS. The operating system makes it possible for the same program (the MMIS) to be driven by two "masters": the manager and the MMKS.
Just as the MMIS contains standard procedures called DATA VIEWERS, the MMKS contains standard procedures. One such standard procedure is an ANALYZER.
The process of running and interpreting DATA VIEWERS is usually termed "analysis", so a system which does this work could logically be called an "ANALYZER". An ANALYZER is a knowledge-based system for doing analysis. For every data viewer there are one or more ANALYZERS. An ANALYZER does an analysis using the DATA VIEWER. An ANALYZER makes a "statement" based upon its analysis. Doing analysis and making statements involves a series of linked tasks.
An ANALYZER would need to know how to perform each of these tasks.
These tasks can be illustrated with an example. Consider a Sales Development Report, a popular DATA VIEWER for locating good and poor performing entities (accounts or markets). One such Sales Development Report displays each account in a market in terms of three measures.
A manager knows that a good opportunity would involve a situation in which his brand is under-represented in a large account. This manager would run the SDI report and then look for an account with a large %ACV and a small SDI. Having found such an account, perhaps Safeway, the manager could then make the following statement:
Observations:
Conclusion:
An ANALYZER makes statements about DATA VIEWERS. Since these statements must be communicated, they must be placed in context and the supporting evidence included. The term "paragraph" will be used to denote the output of an ANALYZER. A paragraph contains four elements: context, data, observations, and conclusions. A sample paragraph is shown below, along with a description of each of the elements in the paragraph's format.
The following is an actual structured paragraph that was produced by one analyzer.
A report would be composed of a number of paragraphs.
A Report Writing Assistant performs the same tasks as a marketing manager and uses the same basic tool: the MMIS. The Report Writing Assistant is a part of a MMKS which interacts with the MMIS via the computer's operating system, thus allowing the MMIS to remain unmodified. The MMKS contains ANALYZERS which initiate and interpret DATA VIEWERS; for every DATA VIEWER there is one or more ANALYZERS. The ANALYZER writes statements, which are composed of observations and conclusions. These statements are part of paragraphs, which contain the statements along with the appropriate MMIS output. A report is composed of a series of paragraphs.
Before an ANALYZER can do its work, something must initiate it -- must cause it to run the DATA VIEWER, make the statement, and write the paragraph. The ANALYZER could be initiated in several ways.
Our initial implementation of the Marketing Gate system uses the third approach. Using this approach, the Report Writing Assistant will prepare the report at the manager's direction. The manager views a series of candidate paragraphs that the system knows how to write, and selects paragraphs to be written. The system writes the paragraphs by applying the appropriate ANALYZERS, which in turn apply the appropriate DATAVIEWERS. In this way, the manager thinks about and interacts with paragraphs, rather than with data and analysis methods. This concept is analogous to the way managers interact directly with the MMIS; they select and run DATAVIEWERS which write tables of data and/or draw graphs.
The next section leads the reader through an example session with the prototype of Marketing Gate.