We have had inquiries regarding behavioral segmentation – requests for more detailed guidance. The summer has been really busy and I did not want to give your questions short
shrift so I am a bit tardy in responding to your comments. Here are some thoughts regarding the methods that would apply to a hotel chain.

A hotel chain may wish to utilize the current in-house data on its customers. This may not be perfect, but it is a good, reliable and inexpensive place to start.
Let me begin by noting what you can expect to ascertain from the analysis.

1. You will be able to determine key segments and the percentage of your customer base in each segment. I can tell you that most of these segments will be familiar to you because of your knowledge of the business. Some segments which you may hypothesize exist will probably be subsets of others. It will provide targeting information and opportunities. There will be several segments of which three will be: (a) loyal frequent customers; (b) loyal less frequent customers; and (c) one time visitors who you will never see again
   and with whom you should not spend any time.
2. You will understand the characteristics of the segments for further use.
3. As you act on the segments, you will be able to measure movement across segments over time.
4. Don’t forget that since this is your data, no one else will be able to understand your customers as well as you can understand them.
5. You will be able to use any questionnaire research to hone your understanding given what you know from the customer behavior. That is, you can also target your marketing research given what you find from the analysis of your data.
6. You can also utilize the data to compare with other studies you have performed although this can be a bit tricky sometimes.

In addition, when you have finished the analysis, you are likely to be able to utilize the analytic output together with other information, such a census data, to target potential new customers immediately. You can always go back and refine your data collection and analyses later if desired.

The first step is to assemble the customer data you have on customer demographics (address, affiliations, and the like), visit information (dates, length of time, amount spent – which can be used to examine flow patterns), response to promotion programs and whatever else you might have.

From a statistical standpoint, there are several methods which could be used:

1. Cluster Analysis – K-means or hierarchical method which is generally preferred in the industry;
2. Q-Factor analysis which I prefer because of the reliability of reading the principal components and the rotated structures;
3. although with very large samples or 10,000 or more, R-Factor analysis gives quite satisfactory results.
4. Discriminant analysis which is appropriate with some data.

The key is to have someone perform the analysis who:

1. takes time to understand the data;
2. thoroughly understands the technique and nuances
   of interpreting the output from the analysis;
3. can help you apply the output to your needs.
4. Remember, another advantage of this approach is
   that it utilizes your proprietary data and is doing so gives you a strategic advantage over other approaches that tend to use information gathering templates and unreliable, ineffectual data gathering.

If your support cannot do all four of these, it is generally better to do something else. Using this approach correctly will provide insights with which you will certainly be pleased. However, you need to realize that the findings will not be rocket science as one of our European clients once said. Then, he added, “Of course, if it were rocket science, it would provide information which I wouldn’t believe”.

While we like to utilize three or four years of data, we find that even six months provides extremely useful insights allowing for actions when debate over actions has been the previous norm.

In terms of expectations, it is important to realize that there are likely to be few revolutionary outputs from the analysis. Rather, the analysis is more likely to define the six or eight critical issues for meeting customer needs as opposed to the 20 or more issues you currently feel you are facing.

You may have someone on staff or readily available who could help you with this. If you need outside assistance, it is our business to be able to help and we can so do as either project consultant providing guidance or as the principal consultant/analyst

As an aside, I think that you will find that the results are far more functional than results of surveys. Our experience (and the experience of most knowledgeable statisticians) with survey techniques is that the sampling frame can be perfectly designed and the output can be incredibly biased because of response rates. For an individual company, statistical techniques are usually less important because you possess information on a universe of your customers, not a sample.

I am confident that once you work with behavioral segments, you will see it provides more value and real feedback on any changing dynamics of your customer base than you have gained through other approaches.

Good luck if you pursue behavioral segmentation to understand your customers better. If we can be of assistance, that is what we love to do.