Business objectives are the origin of every data mining solution. This may seem obvious for how can there be a solution without an objective? Yet this statement defines the field of data mining; everything we do in data mining is informed by, and oriented towards, an objective in the business or domain in which we are operating. For this reason, defining the business objectives for a data mining project is the first key step from which everything else follows.
It is possible to describe data mining or other analytical activities without reference to business context but to do so is to omit a crucial component. It's because business knowledge is central to every step of the data mining process. There are two reasons for this:
Data necessarily provides a narrow or limited view of the world—the real world is always much richer than the data we collect about it—business knowledge is therefore always required in order to interpret this data and relate it back to reality
Data mining and all forms of data analysis perform a function such as perception; therefore, like perception, they are determined by knowledge and directed towards goals
Defining the business objectives of a data mining project can be broken down into four steps:
Understanding the overall goals of the business
Understanding the objectives of your client
Connecting these objectives to analytical results
Linking these results to data mining goals
The first step in defining the business objective is to understand the overall goals of the business or the domain in which data mining is undertaken. These vary considerably depending on the business context, but even supposedly simple goals such as commercial ones can have surprising subtleties, for example:
Commercial goals: In a commercial situation, it might be expected that the primary goal will always be to increase profit or revenue. However, most companies are considerably more sophisticated than this and focus on customer relationships, quality, and the market position of their products; they often include broader concerns such as ethical or social motives.
Service goals: For many organizations and government organizations in particular, business objectives focus on the service they provide, although they may also have revenue goals.
Scientific goals: Scientific organizations, and scientific departments within commercial or government organizations, often formulate their goals in terms of the development of knowledge.
Therefore, at the start of every data mining project, make sure that you understand the nature of the business and its goals.
Within a given business, different individuals or departments have their own objectives; these may be stated in writing or may be implicit in a job title or job description. Often these objectives are embodied in one or more KPIs (key performance indicators) that are used to measure success in or progress towards these objectives. When formulating business objectives for a data mining project, it is helpful to relate these to the objectives and KPIs of the client who commissions the analysis; this allows a direct understanding of the benefits that the project will bring in relation to the objectives of the client. For example, in a project for a client with customer management objectives, the business objectives may be expressed in terms of KPIs such as churn rate or cost of acquisition per customer.
Once you have understood the overall goals of the organization and the objectives of your client, the next step is to select analytical results or targets that will have benefits from these objectives or KPIs.
This requires a combination of business understanding and analytical knowledge, and these results can be very varied; however, for data mining, they always fall into two categories:
Insight or new knowledge: An analysis may reveal a new fact, for example, a relationship between customer attrition and the length of service or products held, or a complex relationship such as combinations of factors influencing customer retention. The key property of insight is that it is knowledge in the head; insight is usually delivered as a presentation or a report providing information, which will then be used by the business.
Predictive models: Data mining also has the option to deliver results in the form of predictive models, that is, knowledge in an artificial form. Different kinds of predictive models (such as classification, regression, clustering, or association models) vary in how they are used, but they all have one thing in common: they can improve the information available about a specific example (such as a customer). They do this by adding information that has been derived by generalizing over a range of examples—this is the function of data mining algorithms.
In order to select analytical results that will contribute to a specific objective or KPI, consider how newly acquired knowledge or predictions of models will be used to further these objectives; the use of analytical results should always be at the front of a data miner's thoughts. One guideline is that management or strategic objectives are often served by insight, whereas operational decision making is often aided by predictive models. However, predictive models of more readable kinds such as decision trees and rule sets (and also analyses of the behavior of less-readable models) can be interpreted to deliver insight for management or strategic purposes.
When defining business objectives for a data mining project, the data miner must simultaneously consider the likely data mining goals that these objectives will generate, such as segmenting the customer base in a particular way or predicting the likelihood of a specific customer behavior. This is necessary because the likely consequent data mining goals, considered in the light of what is technically possible, may lead to adjustments or refinements of the business objectives. The specification of data mining goals will be described in more detail in the Translating your Business Objective into a Data Mining Objective by Dean Abbott essay, later in this Appendix.