Data Mining Businesses can receive many benefits from data mining. Which benefits they receive, however, can also depend on the way in which their data mining is undertaken. Predictive analytics are used to understand customer behavior, and businesses use the behavior of the customer in the past to attempt to determine what the customer will do in the future (Cabena, et al., 1997). While it is not an exact science, many companies believe they can use it in order to decide which products will sell most often to which customers (Nisbet, Elder, & Miner, 2009). Association discovery is another type of data mining, and is more involved with the products that are sold and how they match up to specific types of customers, as opposed to specific customers by name or other determination (Nisbet, Elder, & Miner, 2009). In other words, predictive analytics look at what customer A will buy again, but association discovery looks at what customer A will buy based on his belonging to a particular group of customers who also buy a particular product.

Web mining is used to discover information about customers on the web (Hastie, Tibshirani, & Friedman, 2001). Club cards that are used to access a company website, for example, can help a customer be tracked and his or her spending and buying habits can be discovered. Customers may not always realize that this is going to take place, but it is the only way that companies can provide "for you" deals for specific customers who are part of their rewards structure. They have to know what the customers are routinely buying in order to know what to suggest to them. Clustering is yet another way of data mining, and it looks for customers who are related in their buying habits (Nisbet, Elder, & Miner, 2009). If a particular kind of customer buys certain things very often, then it stands to reason that other customers who also appear to be the same kind would buy those same things. Of course, like any kind of data mining it is not an exact science. There are many things that can affect...

While some can be trusted, others make frequent errors in the way they calculate customer information and what they predict based on that collected information (Cabena, et al., 1997). It is possible, with some study, to predict the kinds of errors these algorithms will probably produce. For each algorithm there are different errors, but they all come down to one thing: correlation is not causation. In other words, it is really not possible to predict why a customer will buy something in every instance. Without understanding the why of the issue, the purchasing of anything becomes a guessing game (Cabena, et al., 1997). A consumer may buy something only once, or because it was on sale, or because a family member was in town and likes that particular product. That does not mean that the consumer will buy that product ever again, and does not tell the company why that product was purchased.
Most of the errors that are made in data mining come from not understanding the issue of why something was purchased (Nisbet, Elder, & Miner, 2009). Other errors can also come from using the wrong kind of analysis to get information. Depending on what kind of information a company wants to collect, a different type of data mining operation may be a better fit for that company. Before data mining is undertaken, a company must spend time understanding its various options and determining what it most wants to know (Cabena, et al., 1997; Hastie, Tibshirani, & Friedman, 2001). Once that has been done, then the company can move toward actually collecting the data and using it in order to give customers of that company a better experience.

One of the biggest issues with data mining is privacy (Nisbet, Elder, & Miner, 2009). When personal data is being collected for mining purposes, there is always the chance that customers will not take kindly to that. There are three main concerns that…