|
|
Segmentation is an often used word in marketing circles. It generally refers to the division of a market or customer base into uniform groups that react differently to marketing and sales activities such as promotions, communications, and advertising. The idea is to determine the best groups or segments and a set of characteristics about each of these groups, so you can target them with a finely tuned mix of marketing and sales programs. The trick however is figuring out what these segments should be.
Despite the large number of tools and methods out there for segmentation analysis, they can be broken down into two fundamental types as described below:
- Choose the segments, then fit the data
This technique is pretty much what it sounds like. By whatever means, the segments are chosen and then each data record is assigned into one segment based on the boundaries established for each segment. Often these segments are decided by intuition, analysis of secondary data sources, analysis of internal customer databases, or similar means. For example, segments could be:
- Customers who ordered less than 12 months ago vs. those who have not
- Single purchasers vs. multi-purchasers
- Customer lifetime value greater than $1,000 vs. those less than $1,000
- North vs. South Regions
- Analyze the data to determine the segments
With this approach no predetermined groupings are used. Instead, analytical techniques are used to sift through the data and determine the best segments (often called clusters). We’ll talk more later about a few of these techniques, as well as ASA’s approach to segmentation.
There are many long-term benefits derived from proper segmentation. Here are a few of more impactful ones:
- Tailored marketing generates better returns. Knowing how each segment behaves allows you to tailor messages and promotions specifically that group. As a result, response rates go up and the total revenue dollar impact can be significant.
- Cost Savings. Think about how much wasted mail you receive each day because you are not really a prime target for the promotion. Segmentation can help you determine which groups are the best targets for promotional material and which groups just represent wasted dollars. Segmentation analysis can cut out a great deal of waste in the marketing plan.
- Higher Operating Efficiency. Segmentation can have a rippling effect on the efficiency of your organization. For example, are account managers spending equal time on all accounts or are the “high potentials” receiving the special attention they deserve? Once you know the characteristics of each segment, making optimum resource decisions become straightforward.
- Higher Customer Satisfaction. Many studies have shown that messages and programs that are highly relevant to a customers interest shows your company’s value and improves customer satisfaction. In addition, you’re less likely to irritate certain customer groups with irrelevant promotions.
A little research will quickly reveal that many techniques and variations are available. As a result, it can be difficult to choose the best technique for every situation. Each technique inherently has its pros and cons. Below is a brief description of two of the more widely accepted techniques:
- Cluster Analysis – this approach categorizes data into segments such that members within each segment have similar characteristics and segments differ from each other as much as possible. Essentially it is the “birds of a feather” concept. Common methods used to perform cluster analysis are K-means and self-organizing Kohenen maps.
- Classification trees trees – with this method you select a dependent variable or set of variables, depending on the specific method. Through a series of splitting steps, groups are continually subdivided until the process reaches a termination point. Common methods for classification are CHAID and CART.
- Other, less common methods you may encounter are:
- Factor analysis
- Conjoint Analysis
- Discriminant Analysis
ASA uses a clustering approach to segmentation utilizing a organizing Kohenen methodology. This approach has proven to be a very robust way to determine market and customer segments. In many cases, outcomes from cluster analysis challenge long held beliefs about a company’s customer or market segments. Below are several benefits associated with ASA’s approach:
- Segments are determined from analysis of the data, so outcomes are not biased by pre-determined organizational beliefs.
- A Kohenen self organizing map inherently does a good job of ignoring noisy data. Noisy data is often a practical reality of the business world. Techniques that perform well using noisy as well as missing data can save you hours of painful data clean up and preparation time.
- Clustering works well with data you already have in house. Later, we discuss the many different types of data that could be used in segmentation analysis, however techniques that make the most of out the data your company collects can yield significant cost savings over purchased or appended data.
- Decision tree methods often require users to work out complex splitting, tree growing, and tree pruning rules.
Depending on the segmentation technique chosen, the data requirements may differ vary slightly. A note to readers; many data services will tell you that you need to purchase or append their data to your records before you can do an adequate job of segmentation. This isn’t necessarily the case.
ASA tends to look at data in two ways; that is “data you probably collect”, and “data you probably don’t collect”. Before you seek out third-party data sources, we recommend that you perform the segmentation analysis with the data you already collect. In the long run this approach is less costly and more efficient for deploying segmentation in day-to-day operations.
Data you probably “do” collect
- Transaction histories such as purchase date, items purchased, quantities, method of purchase (mail, internet, call in)
- Lifetime information such as time on file, total sales, number of product categories purchased, lifetime averages such as order size and time between orders
- Customer name, address, residence type
- Source code of how the customer was acquired
Data you probably “don’t” collect
- Demographics such as age, income, gender, marital status, number of children, education level, and occupation
- Geographics such as population density, rural or urban location, and climate data
- Psychographics such attitude and lifestyle indexes, hobbies, and magazines titles read
This is a common question for which unfortunately there is no one right answer. Our advice is “as many segments as you can practically handle”. Here is the logic behind this statement. The more segments you have, the smaller and more precise you can be about characterizing that group. Therefore you can do a better job of targeting. The ultimate is one customer per segment. Of course, segments of one would be impractical for most companies. So how many segments are practical? Generally we find that trying to work with less than 10 segments is practical.
|
|
|
|
|
|
|
|
 |
 |
