Practitioners of a craft have often resisted or denigrated attempts to introduce technical advances into their craft. It will be understandable, then, if the reader is skeptical when I, a practitioner of data mining, raise some red flags about data mining "automation." I am not trying to fight automation. I recognize that the benefits of technological change go primarily to those who embrace this change: individuals, companies and societies. Rather, I want to make sure that there is a clear understanding of data mining automation, so that it can be used to its best effect.

Automated data mining and modeling software gives marketing managers a tool to perform analyses that otherwise would need to be handled by a highly trained researcher. This is accomplished by establishing a predetermined analysis methodology. An algorithm is developed that attempts to mirror the step-by- step decision-making process that a trained modeler would follow. At each step in the process, preset criteria are used to select analysis options. Because experts have programmed these criteria, the results should be on par with an expert analysis.

But just because such a tool is available doesn’t mean that the marketer doesn’t need the researcher. Michael Berry and Gordon Linoff in their book Mastering Data Mining compare automated data mining to the process of taking a picture with an automatic camera. The camera can relieve the photographer from having to set the shutter speed, aperture and other settings every time a picture is taken. This makes the process easier for expert photographers and makes better photography accessible to people who are not experts. But this is still automating only a small part of the process of producing a photograph. Choosing the subject, perspective and lighting, getting to the right place at the right time, printing and mounting, and many other aspects are all important in producing a good photograph.

In data mining and modeling there are also many parts of the process that cannot be automated, including choosing a methodology to match a business problem, selecting a data set, quality checking and preparing the data for analysis, choosing among the available options within the analysis process, and interpreting and presenting the results.

I would suggest that an analysis could be automated when:

  • The data being used is from a familiar source,
  • The analysis has been used before in the same context,
  • The variables included have been used before in the same type of analysis, and
  • The results of the analysis will be interpreted and used in an established manner.

Within these guidelines, automation can provide huge advantages in time and cost. One clear example is real-time data mining and decision support for Web marketing applications. Rather than establishing fixed business rules, data can be analyzed in real time to evaluate and optimize custom content delivery systems.
The applications do not have to be this high tech to be valuable. Direct marketing campaigns may need to be checked repeatedly for performance against a number of targeting dimensions, such as demographics and model scores. Automating a model that identifies high-performing segments would allow the marketer to quickly evaluate and modify marketing strategies. It also allows the technical analyst to focus on more complicated and challenging projects. The advantages of this automation are even greater if the data mining software is integrated with the campaign management and tracking tool.

On the other hand, when outside of these guidelines, the elements of data mining that cannot be automated come into play. These elements present a very real danger that automated data mining tools could end up producing poor results. (By "poor result" I mean anything from missing some useful information all the way to drawing conclusions that are incorrect and damaging to marketing results.) Benjamin Disraeli spoke of "Lies, damn lies and statistics." The implication is that in the hands of a skilled analyst, data can be manipulated to be intentionally misleading. It is equally true that in the hands of an unskilled analyst, data can be analyzed in ways that are unintentionally misleading.

How much expertise is required varies on a case-by-case basis. Anybody can put a bandage on a cut finger. A family practitioner or an emergency room physician can set a broken arm. For more complicated injuries, an orthopedic surgeon may be required. For data mining and modeling, marketers have to decide how much they are comfortable doing themselves and when they need outside expertise. Clearly, advances in technology have expanded the areas of analysis that can be automated and handled by marketing managers. But be careful – what you don’t know can hurt you.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access