Most (all?) businesses have a customer base that consists of a list of accounts, and a group of transactions associated with each of those accounts. The account level and the transaction level are two quite different things, and in order to analyze the customer base properly, there needs to be the capability to analyze the data at both levels distinctly and simultaneously. To illustrate, every transaction takes on the attributes of the account, because each completely belongs to the account. But the account takes on the attributes of the transactions as a group.

For example, a transaction has the attribute of belonging to a person who lives in NYC, even if the transaction was executed in Denver. The account then acquires the attribute of having a transaction that was executed in Denver. With our tools, you can analyze based on people who live in NYC and perform transactions in Denver, because of multilevel capability. The behavior can get interesting, because the other transactions associated with this person now acquire the attribute of belonging to a person who executes transactions in Denver, regardless of where the other transactions took place. If a tool cannot handle both multilevel analysis and iterative analysis (which is a different subject), you cannot investigate this type of behavior.

Long form census data provides an excellent vehicle for us to exhibit our tools' capability for customer account data. The reason is that both census and account data are multilevel data sources, both are voluminous, both have a variety of attributes and volumes at upper and lower levels. While the long form data seems to have a reputation for being difficult to work with because of it's size, it is quite small relative to the telephone account data that our tools have historically analyzed.