Correlating Web-User Data
Marketing on the Internet requires so much data for operations that we sometimes forget the same data can be used for research to gain insight into prospects and customers.
Everything served to a visitor -- from the first page through marketing, sales, and product fulfillment -- generates data about the customer. Web marketers can tap into this "free" source of profile data for just the cost of converting existing data into a format that can be used by a data-analysis program.
In general, two types of data are available about Web visitors:
- Information provided by the individual
- Information collected by observing behavior
Both types of data have their benefits and limitations, but when they are combined, our view of Web visitors becomes much clearer.
The data we typically receive from visitors includes registration and subscription data (for site and newsletter content), and delivery data (for product shipment). Those sites using personalization can collect data about users' interests, preferences, and decision criteria.
One problem with information gathered directly from individuals is that the data can be inaccurate. Marketing researchers have known for years that
survey responses are frequently biased; respondents sometimes attempt to please the researcher or make themselves look good.
As for Web forms, it's common for people to provide random or inaccurate data. If, however, it's essential for them to receive the benefits of the site, they will provide accurate information. For instance, it's hard to imagine a visitor at a music site saying he likes rock music when he actually prefers jazz.
At the same time, observed data has accuracy problems, too. Just because someone spends time on certain pages or clicks links in a newsletter doesn't mean she has a significant interest in those topics.
However, by combining data from observation with data from individuals, it's possible to identify marketing activities that are likely to lead to a sale.
Frequently, when the topic of analyzing Web data comes up, many marketers think that it means data mining. The truth is, you don't have to use an elaborate data-mining tool to get started with data analysis. In fact, one of the handiest data-analysis tools is Microsoft Excel. The statistical functions in Excel are relatively easy to use, and they provide a quick way to do an initial analysis.
One statistical term that has found its way into popular business language is "correlation." We frequently hear phrases like "the number of visits is correlated with order size." Most of the time, though, there has been no measurement of just how much correlation exists.
Correlation measures the relationship between two sets of data on a scale of 0.0 (no correlation) to 1.0 (100 percent positive correlation) or -1.0 (100 percent negative correlation).
There are many ways to use this technique to learn about the people who visit your Web site.
One way is to analyze how the number of visits to a site correlates with order size. If sales tend to be made on repeat visits, then it's important to make sure marketing vehicles such as e-mail newsletters are being used to bring people back to the site. On the other hand, if the number of visits is not highly correlated to sales, then it's more important to motivate visitors to take action during every visit.
Here is a hypothetical set of data showing the number of times nine customers visited a site and the size of their order:
|Visitor Number || |
Number of Visits
|1 || |
|2 || |
|3 || |
|4 || |
|5 || |
|6 || |
|7 || |
|8 || |
|9 || |
It's clear from this data that the people who visited the site only once placed small orders, but it's harder to discern a relationship between sales and more frequent visits. Measuring the relationship between these two sets of data is easy. Excel quickly calculates a correlation coefficient of 0.84, which indicates a strong relationship between the number of visits and the size of orders.
Correlation can be used to analyze a wide range of data, including the following:
- Inquiries and sales
- Number and length of visits
- Number of pages visited
- Sequence of pages visited
- Demographics and purchase history
- Buying motives and decision criteria
- Click-through data from newsletters
For companies using a customer relationship management system, both online and offline data are available for data analyses that can help marketers spot opportunities.
While correlation does not actually identify the cause of a particular behavior, it does help identify when two behaviors are likely to occur together. As a practical matter, then, if you can elicit visitor behavior that is highly correlated with sales, then you're likely to increase revenue.
Cliff Allen is the co-author of the book One-to-One Web Marketing; 2nd Ed., published by John Wiley & Sons, and has consulted with companies on strategic marketing for 20 years.